Understanding i-Tree: Summary of programs and methods

i-Tree is a suite of computer software tools developed through a collaborative publicprivate partnership. These tools are designed to assess and value the urban forest resource, understand forest risk, and develop sustainable forest management plans to improve environmental quality and human health. This report provides details about the underlying methods and calculations of these tools, as well their potential limitations. Also discussed are the history of i-Tree, its future goals, and opportunities to facilitate new science and international collaboration.


EXECUTIVE SUMMARY
i-Tree is a suite of computer software tools developed through a collaborative publicprivate partnership. These tools are designed to engage urban and rural populations in assessing and valuing their forest resource, understanding forest risk, and developing sustainable forest management plans to improve environmental quality and human health. The tools can assess individual trees and forests in both urban and rural areas. i-Tree's vision is to improve forest and human health through freely available, userfriendly technology that engages people throughout the world in enhancing forest management and resiliency. While the science and models at the core of i-Tree have been in development since the mid-1990s, the i-Tree program suite was first released in 2006 as a framework for science delivery.
i-Tree programs engage citizens, youth, and land managers in understanding their local forest resource, its value, and risks to sustaining forest health. The tools aid in forest management decisions to improve forest and human health and well-being. Though originally developed as tools to aid in urban forest management, the i-Tree suite is being used in rural forests and is being integrated within the USDA Forest Service's Forest Inventory and Analysis program.
Over the past decade, the i-Tree suite's usage and capabilities have increased tremendously.
As of the end of 2019, there have been more than 410,000 users of i-Tree tools in over 130 countries. A nonexhaustive literature search revealed over 900 articles 1 that either use i-Tree, developed methods for i-Tree, or make reference to i-Tree or precursor programs (i.e., UFORE and STRATUM).
The goal of this document is to improve the understanding of i-Tree by summarizing i-Tree's: History and future goals Program methods and potential limitations Opportunities to facilitate new science and international collaboration 1 Searching the phrase "i-Tree urban forest" in Google Scholar in December 2017 produced 2,100 results. The 900-plus articles in the i-Tree references section of this document represents only the first 100 pages of the search results (see appendix 14).

INTRODUCTION
i-Tree is a suite of computer software tools developed through a collaborative publicprivate partnership. The tools are freely available from i-Tree website (www.itreetools. org) and are designed to be accessible and user-friendly and to aid urban and rural land managers and the general public in: Assessing and monitoring their local forest resource Understanding the services and values provided by trees and forests Evaluating risk to forest and human populations Developing sustainable forest management plans Improving environmental quality and human health Selecting appropriate tree species and locations Engaging stakeholders and public audiences While development started in the mid-1990s, the i-Tree suite was first released in 2006.
Over the years the i-Tree suite has grown, adding new tools, methods, and publications. Due to i-Tree's growth in use and capabilities, this document is designed to help readers understand: What is i-Tree?
Where is i-Tree is headed?
What are the current methods and limitations related to calculating ecosystem services and values?
Through a better understanding of i-Tree, international collaboration can be enhanced to help improve these tools for forest managers across the globe.

WHAT IS I-TREE?
i-Tree is a suite of computer programs designed to estimate forest structure, ecosystem services, values, and risks related to forests and people. It was originally developed in the 1990s as the Urban Forest Effects model (UFORE), with a street tree assessment tool (STRATUM) added in the early 2000s. Today i-Tree includes a variety of tools that are used globally and encompass trees on all urban and rural lands.

Vision
To improve forest and human health through user-friendly technology that engages people around the world in enhancing forest management and resiliency.

Goals
To attain this vision, i-Tree continues to develop tools with the ultimate goals of allowing anyone to: Assess local forest conditions

Quantify multiple ecosystem services and values derived from forests
Determine local risks to forest and human health Calculate how changes in forest structure will lead to changes and tradeoffs among ecosystem services and values Develop best forest management strategies to enhance desired ecosystem services and forest and human health Determine the best tree species, locations, and planting rates to optimize ecosystem services and values through time and across space to enhance human health and well-being By achieving these goals, i-Tree users can create healthy, sustainable, and resilient forest landscapes across the urban-to-rural continuum.

Tools
i-Tree tools aid in forest management decisions, education, and advocacy by working at various scales: from the individual tree, to the property, neighborhood, city, and landscape levels. The tools are divided into five categories: Core Programs -fundamental tools that assess forest tree populations, ecosystem services, and values. These main programs are designed to provide forest data and aid in improving forest management decisions. The core programs 2 are known as: i Partner Tools Powered by i-Tree -these are tools developed for specific partners that available for all to use. These tools are: County Tree Benefits 2 Core programs and utilities listed here are current as of January 2020.

GHG Planting Calculator
Harvest Carbon Calculator

Wood Marketplace
Research Programs -the code to research programs are available for advanced, technical users. Results from these tools are often built into core i-Tree tools to make this research more accessible to broad audiences. Technical support is not available for these tools. Research tools are: i-Tree Cool River i-Tree Hydro+ Legacy Programs -programs that have been incorporated within other tools or utilities, or are no longer being updated by i-Tree. Original versions of these programs, except for the Mobile Community Tree Inventory (MCTI) program, can still be used and are available on the i-Tree Website: i-Tree Streets (incorporated within i-Tree Eco) i-Tree Vue (superseded by i-Tree Landscape) Mobile Community Tree Inventory (MCTI) (incorporated within i-Tree Eco) i-Tree Storm The programs and utilities are comprised of desktop (DT) and web-based (web) applications.

CORE PROGRAMS
Eco (DT) uses sample or inventory data collected in the field along with local hourly air pollution and meteorological data to assess forest structure, health, threats, and ecosystem services and values for a tree population. Information provided to the user includes number of trees, diameter distribution, species diversity, potential pest risk, invasive species, air pollution removal and health effects, carbon storage and sequestration, storm water runoff reduction, VOC emissions, and effects on buildings' energy use. Features of i-Tree Eco include: Plot selection programs.
Manual data entry or mobile device data entry.
Existing inventory import capability. Automatic report generation.
Ability to forecast future tree population totals, canopy cover, tree diversity, tree diameter distribution, and ecosystem services and values by species based on user-defined planting rates and default or user-defined mortality rates (e.g., a user can simulate the effects of emerald ash borer by specifically killing off ash trees).
A pest detection protocol for long-term pest detection and monitoring that allows users to input pest and disease signs and symptoms of their trees to produce indications of threats to their forest.
Design (web) links to Google maps and allows users to see how the trees around their home affect various environmental services. Users can use this tool to assess which locations and tree species will provide the highest level of benefits. This tool is geared toward homeowners, students, or anyone interested in tree benefits. Additional capabilities include multiple tree and building placements, projecting future and past benefits, and displaying preferred planting zones.
Landscape (web) integrates national land cover, population, and environmental data to aid in forest management and planning. This tool features: National data layers related to landscape structure Users can select their area of interest to explore these multiple datasets and weight the data to prioritize areas for management actions (e.g., most important areas to plant or protect trees). i-Tree Landscape is a web-based tool that can improve forest management and planning, and engage stakeholders in management prioritizations and decisions to improve forest and human health and sustainability.
Hydro (DT) simulates the effects on hourly stream flow and water quality due to changes in tree cover and impervious cover within a watershed. It contains auto-calibration routines to help match model estimates with measured hourly stream flow.
Canopy (web) allows users to easily photo-interpret Google aerial images to produce statistically reliable estimates of tree and other cover types along with calculations of the uncertainty of their estimates. This tool provides a quick and inexpensive means for cities and forest managers to accurately estimate their tree and other cover types. Canopy can be used anywhere in the world where high-resolution, cloud-free Google images exist (most areas). Use of historical imagery can also be used to aid in change analyses.
MyTree (web) is a mobile phone application (app) that allows users to rapidly quantify the benefits and values of individual or small groups of trees ( Fig. 1).

UTILITIES
Species (web) helps users select the most appropriate tree species based on desired environmental functions and geographic area. The program calculates the best tree species based on user-provided weighting of desired environmental benefits at maturity.

LEGACY PROGRAMS
Streets (DT) is similar to Eco but focuses on street tree populations using regional reference city street tree data. Features of this program have been incorporated within i-Tree Eco. Street tree analyses within Eco do not use reference city data but rather use local pollution and weather data to estimate ecosystem services.
Vue (DT) is the precursor program to i-Tree Landscape. It has now been superseded by the i-Tree Landscape program, which is more accessible and uses newer data with more modeling capabilities.
Mobile Community Tree Inventory (DT) is a basic tree inventory application designed for use with a personal digital assistant (PDA). This program has been incorporated within the mobile data collection feature of i-Tree Eco. This original PDA application is no longer distributed or supported.
Storm (DT) helps assess widespread street tree damage in a simple and efficient manner immediately after a severe storm. It is adaptable to various community types and sizes and provides information on the time and funds needed to mitigate storm damage.
All programs, except for MCTI, are available and free, including technical support for all core programs, online training, user's manuals and resources, and a moderated peer-to-peer online forum.

Partnerships
The core partners in i-Tree development, dissemination, and support are the USDA Forest Service, Davey Tree Expert Company (Davey Tree), Arbor Day Foundation, Society of Municipal Arborists, International Society of Arboriculture, Casey Trees, and SUNY College of Environmental Science and Forestry. In addition, several Federal and university personnel are involved with developing new parts to i-Tree and/or providing data and information used within i-Tree.
For more than 20 years, the USDA Forest Service and Davey Tree have collaborated on cooperative research and development to improve understanding and management of trees and forests. Davey Tree matches USDA Forest Service funding and has jointly-funded employees within USDA Forest Service and Davey Tree offices who work on i-Tree development, dissemination, and support. Through this relationship, and with the help of numerous other cooperators, the i-Tree software suite is developed and now serves as a premier, state-of-theart, peer-reviewed toolset for assessing forest ecosystem services and values (Fig. 2).
As integral partners, the USDA Forest Service and Davey Tree signed a Memorandum of Understanding in 2015 to strengthen their cooperation through an expanded relationship with the following goals: Continue cooperating to develop, maintain, disseminate, and support the i-Tree suite of assessment tools.
Expand the geographic focus of i-Tree to include all landscapes, both public and private, and rural and urban.
Contribute to a collaborative discussion that aims to improve the health of our urban forests and our nation's forests as a whole. More effectively deliver a growing scientific-based body of knowledge of trees and ecosystem services to a broader, more diverse set of stakeholders and geographies.
Explore how mutually beneficial research and development can be better utilized in traditional forest management and in managing emerging threats such as climate change, invasive pests, and increasing wildfire size and severity.
Engage youth and educators in the use of i-Tree to learn about forest dynamics, ecosystem services and their value, the wonder of nature, the importance of stewardship, and to encourage a greater appreciation for the natural environment within our next generation of conservation leaders.

History
The origin of the i-Tree program is the USDA Forest Service's Urban Forest Effects (UFORE) model, which was developed in the 1990s, and the Street Tree Resource Assessment Tool for Urban Forest Managers (STRATUM) model, which was released in the early 2000s. Both of these programs were built from research originally developed in the USDA Forest Service's Chicago Urban Forest Climate Project (McPherson et al. 1994;Nowak 1994aNowak , 1994bNowak , 1994c. UFORE also had its origin from work in Oakland, CA (Nowak 1991). Due to user confusion between these closely related programs, a decision was made to integrate USDA Forest Service urban forestry tools within one program. This program was named "i-Tree" 3 and first released in 2006. The programs were originally designed to aid in urban forest management but have since been expanded to meet the needs of multiple constituents, including rural forest managers, consultants, planners, homeowners, educators, and landscape professionals. The research and work of numerous people were instrumental in the evolution of i-Tree (appendix 1).

i-Tree 2019
With the expansion of programs and partners, i-Tree usage has substantially grown both nationally and internationally. As of the end of 2019, there have been more than 410,000 users of i-Tree products in over 130 countries (Figs. 3 and 4). Usership increased 28 percent (90,000 new users) in 2019 (Fig. 5). The i-Tree website supports approximately 12,000 unique visitors each month (Fig. 6).  The information provided by i-Tree software has been used to influence management and policies throughout the United States and the world in relation to urban forestry. i-Tree results have been used by consultants, managers, and local citizens to guide management and policy decisions related to issues such as: emerald ash borer protection; building financial support for urban forestry programs, tree planting, and management; linking local tree data with the U.S. Conference of Mayors Climate Protection Agreement; public outreach campaigns (e.g., billboards, "price tags"; see Fig. 2) on tree benefits; developing Urban Forest Strategic Management Plans; and focusing funds to improve stewardship (Driscoll et al. 2012). The i-Tree team released a survey (https://www.surveymonkey. com/r/iTreeSurvey) for users to help improve the program and to learn more about success stories related to using i-Tree information. Numerous links and dependencies to i-Tree programs exist (appendix 2).

i-Tree and Urban FIA
The USDA Forest Service's Forest Inventory and Analysis (FIA) Program provides critical information needed to assess the status, trends, and sustainability of the nation's forests. In 2014, a national Urban FIA (UFIA; https://www.nrs.fs.fed.us/fia/urban/) program began by integrating i-Tree into FIA. Since both i-Tree and UFIA fall within the bounds of the overall FIA program, there can be confusion between the two programs.
As stated previously, i-Tree is designed to estimate structure, ecosystem services, values, and risks to forests and people. Urban FIA is a national natural resources inventory system designed to monitor the quality, health, composition, and benefits of trees and forests within our nation's urban land. UFIA has a goal of monitoring citywide tree data for approximately 100 of the nation's most populous cities. To date, 35 cities have started monitoring (Fig. 7). More cities will be monitored in the future. Cities are monitored based on a panel system, where a proportion of total plots are analyzed annually. For example, in a 7-year panel, 1/7 of the total plots (~29 out of 200 plots) are analyzed each year, such that after year 7, all plots are measured. Year 8 remeasures year 1 plots, year 9 remeasures year 2 plots, and so on. Some cities, such as Austin ) and Boston '20  Houston, Texas , have opted to collect all plots in year 1 and then start the panel-based resampling in year 2 to produce an assessment with all plots after the first year. Current cities being analyzed, by starting year, are: These cities are selected based on partnerships with local municipalities or states.
Urban FIA analyses are not limited to city boundaries, but also include the larger surrounding metropolitan landscape, complementing regional and local efforts to provide a cohesive picture of urban forest conditions across the United States. Urban FIA provides quality, publically available data for trees and seedlings that can be compared with rural forests regarding: Ecosystem services and values (via i-Tree) Tree volume, biomass, carbon, merchantability, and cull estimates Forest health (e.g., crown conditions and damage to trees) Forest change through time Related to UFIA, i-Tree Eco: Was used in pilot testing of UFIA analyses since the late 1990s (Buckelew-Cumming et al. 2008;Nowak et al. 2007a.
Analyzes the ecosystem services and values derived urban forest data collected in the UFIA program.
Is being seamlessly integrated within FIA to optimize data processing and reporting of Urban FIA data and to allow for Urban FIA and i-Tree to produce similar outputs.
While i-Tree helps with national urban forest monitoring through FIA, it also helps anyone in conducting local forest analyses. If a community's urban forest is being assessed by UFIA, then that community will not need to run i-Tree Eco because i-Tree will be used on the FIA data to estimate these services and values. If UFIA is not collecting data in a community, i-Tree Eco can be used by a community to collect information on their local urban or rural forest and have results similar to Urban FIA analyses. There are also additional i-Tree tools that a community or individuals can use to conduct more localized assessments (e.g., rural forests, street trees, parks, back yard, single trees). Users can choose what i-Tree tool to use to best meet their local needs.

FUTURE GOALS
The overarching goal of the i-Tree program is to develop best management practice prescriptions based on local environmental and forest data. These prescriptions will aid managers in sustaining healthy and functional forests to improve human health and well-being. These management prescriptions will detail best tree species, locations, and planting rates to attain desired outcomes. The specifics of each prescription depend upon local site conditions and issues, and user's preferences. To attain this goal, the i-Tree team is focused on developing both i-Tree Eco and i-Tree Design as the tools that integrate the science and information from numerous sources in one system to aid managers (Fig. 8).
To reach this goal, the i-Tree team is currently focusing on developing more national map-layer information into i-Tree Landscape and more tree species-specific information within i-Tree Species. Upon completion of these two tools, the information in these tools will feed environmental quality and risk information (from Landscape) and species' abilities to provide environmental benefits (from Species) into Eco and Design to develop forest management guidelines. The suggested management prescriptions will include species recommendations, optimal locations, and planting rates needed to best address local environmental issues and risks, and to sustain a healthy forest into the future. For example, when a user enters data from a specific location, i-Tree will search the Landscape database for the area to determine the environmental issues and risks, potentially including poor air quality, high UV radiation, insect and disease threats, and future species performances based on climate change. Knowing these issues, i-Tree will search the species database and suggest the best species given the current issues and risks. For example, if the area has poor air quality and Emerald ash borer, the prescription will suggest the best species for improving air quality that are not ash trees. The prescription will also suggest landscape designs, planting locations and planting rates to sustain current or desired levels of tree cover throughout the landscape. Future development and integration of tools is critical to meeting this management prescription goal.
While this management prescription tool is the overarching goal, there are 10 specific goals that i-Tree intends to attain over the next 5 years (2020 to 2025): 1) Maintain quality programs -This work is essential and involves keeping software and servers functional and bug-free. With growing programs, usage, and constantly changing software (e.g., Microsoft® Windows operating systems, Internet browsers), continuous Arrows indicate the direction of passage of information or routines to subsequent programs. Dotted-line arrows indicate future linkages. i-Tree Design and i-Tree Eco will ultimately make forest management recommendations related to best tree species and locations to improve forest and human health given local conditions. maintenance is essential. The code for older tools will be released through an i-Tree legacy site and will no longer be supported; core tools and utilities will be maintained on the i-Tree site; and new tools (see #2 below) will be released on an i-Tree Research site once these tools have been tested and passed peer review.
2) Integrate more ecosystem services and values -Currently many new ecosystem service calculations are in development, including tree effects on air temperatures , human comfort, green infrastructure effects, forest bathing, wood products, and pollen. Once research on these new services passes peer review, they will be incorporated within the core i-Tree tools or released through the i-Tree Research development site as appropriate. This work will be useful for ecosystem/natural capital accounting.
3) Increase connections to human health and environmental regulations -Links with air quality and human health already exist in i-Tree; working with numerous cooperators, the i-Tree team is working to add additional connections to human health related to air temperature, ultraviolet (UV) radiation, and stress reduction. i-Tree staff continues to try and link i-Tree outputs with environmental regulations. 4) Enhance i-Tree Landscape -Based on numerous national datasets from several cooperators, this tool will be adding many features, including national map data related to:

Forest fragmentation
Heat islands Potential hurricane damage To create healthy and resilient forests that provide optimal services for people, it is important to determine the best locations for trees within the landscape. To do this, four factors are being considered: Human and forest population distribution Locational differences in ecosystem service production Risks to human health (e.g., air and water pollution, temperature, climate change, flooding, UV radiation exposure) Risks to forest health (e.g., insect and diseases, climate change, urban development, fire) In addition to data that aid in assessing locations with the highest risks to forests and/ or humans, other new features will be added, including new querying and prioritization capabilities and increased map resolution. 5) Improve capabilities of i-Tree Species -This program is designed to determine the best tree species for desired ecosystem services. This program will be enhanced by expanding and enhancing species information to aid users in making species selections and improving environmental quality. 6) Increase functionality of i-Tree Design/My Tree -These programs are designed to easily calculate effects of small tree populations by allowing users to estimate the benefits and values of individual trees. i-Tree Design will be enhanced by linking to data in i-Tree Landscape and new ecosystem services in i-Tree Eco that are being developed. The MyTree phone app ( Fig. 9) will be enhanced by adding in cumulative tree benefits of the tree's life span and allowing users to pin the tree locations on Google Maps to create a global tree map. 7) Full integration of i-Tree tools within FIA -The i-Tree team, which is part of the USDA Forest Service's Forest Inventory and Analysis (FIA) program, is working closely with the Urban FIA team and other FIA groups to integrate i-Tree seamlessly within FIA processes. This integration will allow for rural forest estimation of many ecosystem services and values that are now only typically available for urban forests (e.g., air pollution removal).

8) Increase utility of tools for National Forests and
State Forests -i-Tree Landscape and other tools have capabilities to aid in national and state forest management. By working with national and state forest staff, tools will be enhanced or modified to address regional scale forest issues and management needs. 9) Enhance School Educational Curriculum using i-Tree -i-Tree staff are working with USDA Forest Service Conservation Education, Project Learning Tree, and others to develop student and teacher guide books and curriculum (www.itreetools.org/support/ resources-overview/welcome-i-tree-teaching-resources) to facilitate the use of i-Tree to teach science in schools.
10) Expand international partnerships -i-Tree tools have the capability and are being used internationally, but the tools are often not specifically designed to work internationally. Thus, developing international partnerships (https://www.itreetools.org/ support/resources-overview/i-tree-international) to facilitate the development and use of i-Tree is critical to sharing and enhancing the global use of the tools.
The development of i-Tree has expedited urban and rural forest data collection and analysis across the world (Fig. 10). This international data collection is providing valuable new data to aid research in understanding patterns and variations in forests and the ecosystem services and values they provide. New i-Tree versions are being developed for New Zealand and for select cities in other countries via i-Tree Database. i-Tree serving as a global standard on data collection and analyses will facilitate further research and comparison on forests worldwide.

METHODS, ADVANTAGES, AND LIMITATIONS
The premise behind i-Tree is shown in Figure 11. Structure is the basic information on the physical forest resource (e.g., number of trees, species composition, tree sizes and locations, leaf area, etc.). The attributes are directly measured by users or estimated (e.g., leaf area) by i-Tree based on direct measures of structure. From the structure data, along with local environmental data (e.g., weather data), various tree functions (e.g., gas exchange, tree growth) are estimated. These functions are then converted to various services (e.g., pollution removal) based on other local data (e.g., pollution concentrations). These services are then converted to benefits (e.g., cleaner air, impacts on human health) based on other data (e.g., local atmospheric conditions, human population data). Finally, the benefits are converted to values based on various economic procedures.
Forest management objectives often seek to improve environmental or human health, or the value of the forests. Forest managers do not directly manage functions or values, rather they manage forest structure (i.e., plant and remove trees, protect existing forests, and select species and locations) to optimize services and values. However, managers are not the only force influencing forest structure. The existing forest and the environment have a substantial influence on forest structure through natural regeneration, tree growth and mortality, storms, insects and diseases, invasive plants, and other factors. Included in this environmental influence are unintended consequences of human actions and development (e.g., introduction of exotic plants, insects and diseases, increased air temperatures and pollution, climate change, etc.). People and nature act to alter forest structure and consequently forest services and values. i-Tree is designed to aid managers by providing basic data on forest structure, services, and values. The i-Tree team is also working on providing management guidance related to the best species and locations to sustain or enhance forest health, services, and values through time. There are four main steps needed to quantify the services and values from forests (Nowak 2018): 1) Quantify the forest structural attributes that provide the service for the area of interest (e.g., number of trees, tree cover). These structural data are essential as they quantify the resource attributes that provide the services.
2) Quantify how the structure functions and influences the ecosystem service (e.g., tree density, tree sizes, and forest species composition are significant drivers for estimating carbon storage).
3) Quantify the benefit of the selected ecosystem service. In many cases, it is not the service itself that is important, rather the impact that the service has on human health or other environmental attributes that provide value to society. 4) Quantify the economic value of the impact provided by the ecosystem service.
Given the framework illustrated in Figure 11, it is imperative that forest structure be accurately assessed. Inaccurate measurement of structure will lead to inaccurate estimates of subsequent services and values. Similarly, the validity of the models and data that are used to estimate services and values will impact that accuracy of these estimates. As with any assessment and modeling framework, there are tradeoffs among selected methods in terms of efficiency, cost, practicality, and accuracy. All current estimates and means of estimation can be improved to varying degrees.
The following section will discuss the basic methods used to assess forest structure in i-Tree along with advantages and limitations of the approaches used. Additional sections will introduce the methods used convert those structural assessments into estimates of services and values. This report may not provide full details but will reference documents and research that contain the full methods. Additional sections will also detail how various i-Tree tools beyond i-Tree Eco assess structure, services, and values. The methods detailed here correspond to i-Tree 2020 (version 6.1.31).

Assessing Forest Structure in i-Tree Eco
Good forest structure data are critical to i-Tree. These data can vary from information about an individual tree to consolidated information about the entire urban forest. Forest population data can be derived either from an inventory of the population, where all trees are measured, or from a sample of the population, where some subsets of the population are randomly selected and the population total is estimated from this sample.
Before taking the sample, land managers can decide whether the data will be collected with a certain number of plots in each land use type (e.g., prestratified), or collected randomly from plots and later categorized by land use type (e.g., post-stratified). Prestratification can reduce the statistical variance and allow users to put more plots in desired strata (e.g., putting more plots within forested urban areas). However estimates of change through long-term monitoring of the plots is complicated as the strata can change through time.
With post-stratification, the randomly located plots can be reclassified into most any strata classification after data collection (e.g., land use, management zones). This post-stratification may not be the most efficient in reducing variance but allows for easier assessments of change.
With an inventory of all trees, there is no estimate of variance or sampling error in the population total as the entire population is inventoried. With sampling of a subset of trees, the standard error of the population estimate is calculated.
Users can select which assessment method is most appropriate for their population (i.e., inventory, prestratified sample, or random sample with post-stratification). Inventories likely work best for small populations of trees, or populations where intensive management is conducted on each tree (e.g., street tree populations). Quality assurance guidelines are provided in the i-Tree field manual (i-Tree Team 2019a) to help ensure that tree measurements are conducted and recorded properly in the field.
This section will detail what measurements are made or derived for each tree, how the estimate is derived, the uncertainty of the measure, and how the standard errors are calculated for population samples. This section and the next section, "Assessing Ecosystem Services" focus on methods used in i-Tree Eco, which is the core i-Tree computer program. Other i-Tree tools use derivatives of i-Tree Eco outputs (results) to assess ecosystem services and these derivative procedures are described later in this report. Field and user's manuals are available in several languages to guide users in establishing projects, collecting field data, and operating the program. These manuals are updated regularly and available online. In this report, we refer to version 6.0 (7.2.2019) of the field and user's manual (i-Tree Team 2019a, 2019c).

TREE MEASUREMENTS
i-Tree defines a tree as any woody plant with a diameter at breast height (d.b.h.; defined as the measurement at 1.37 m [4.5 ft]) greater than or equal to 2.54 cm (1 inch). A shrub is defined as any woody plant with a d.b.h. less than 2.54 cm. However, users can set their own classification of tree versus shrub (e.g., based on species). Table 1 summarizes the core variables directly measured for each tree. Other optional tree or site variables are detailed in the field manual (i-Tree Team 2019a). Where: BA = the total basal area of the stems (ft 2 ), and d.b.h. is measured in inches.
The tree characteristics are measured in the field and assumed to be measured without error. The i-Tree Eco User's Manual (iTree 2019c) describes how to collect these variables along with quality assurance procedures.
These measured characteristics are used to derive secondary structural variables and estimate various ecosystem services ( Table 2). The secondary structural variables derived in i-Tree are leaf area, leaf biomass, leaf area index, and total tree biomass. Subsequently, the directly measured characteristics and derived variables are used to estimate ecosystem services (Table 2).

LEAF AREA AND LEAF AREA INDEX
The cumulative leaf area in an urban forest canopy is an important variable influencing estimates of biomass, air pollution removal, carbon storage and sequestration, and other ecosystem services.
Leaf area is defined simply as the amount of surface area (one-sided) of leaves on a tree. Leaf area measurements are "scaled up" to cover an entire urban forest. The cumulative amount of leaf area per unit of projected ground surface area is known as the leaf area index (LAI = leaf area (m 2 ) / ground area (m 2 )).
Leaf area of individual open-grown, deciduous trees 5 is calculated using a regression equation 6 (Nowak 1996): ln Y = -4.3309 + 0.2942H + 0.7312D + 5.7217S + -0.0148C 4 Basal area is defined in this case as the total cross-sectional area of all stems at d.b.h. in the cluster. 5 Open-grown trees are defined as trees with a crown light exposure of 4 or 5 (see Tree Measurements). 6 Regression equation is used for deciduous trees with a crown height-to-width ratio of between 0.5 and 2.0; crown heights between 1 and 12 m, and crown widths between 1 and 14 m (see Tree Measurements). For trees with a crown height-to-width ratio (HWR) >2, the leaf area estimate is calculated by the regression equation based on the tree crown width and a HWR = 2. The tree's total leaf area is then calculated as the leaf area estimate × the tree's HRW/2. For trees with a crown HWR <0.5, a tree leaf area estimate is calculated by the regression equation based on the tree height and a HWR = 0.5. The tree's total leaf area is then calculated as the leaf area estimate × 0.5/HRW. For trees with dimensions beyond the end of the regression equation range (crown heights <1 m or >12 m or crown widths <1 m or >14 m), the leaf area is calculated by multiplying the ground area of the crown by the leaf area index (LAI) based on tree measurement at the end of the regression equation range. For example, once a crown height exceeds 12 m, the LAI for a 12 m crown height is determined based on the tree's actual crown width, height to width ratio, and shading coefficient; and this LAI is applied to ground area of the crown to determine the total leaf area.  The shading coefficient is an estimate of the percentage of sunlight striking a tree crown that is not transmitted through gaps to the ground (appendix 3). Trees with very dense crowns would permit less light to reach the ground, and thus have a high shading coefficient. If no shading coefficient is available for an individual species, the average of shading coefficients from the other species' in the same genus is used. If no genus coefficient exists, then average family, order, subclass, or class coefficients are used. The coefficients, listed in appendix 3, will be updated based on new data from McPherson et al. (2018).
To estimate the leaf area for a conifer tree (class Pinopsida, excluding pines), an average shading coefficient of 0.91 is used. 7 A shading coefficient of 0.83 is used for pines (Pinus spp.). 8 The leaf area for trees in more light-limited positions (crown light exposure = 0 or 1), such as those found in a dense forest setting, is calculated using a leaf area formula for forests based on the Beer-Lambert Law: Where x s = shading coefficient, k = 0.52 for conifers and 0.65 for hardwoods (Jarvis and Leverenz 1983), and r = crown radius.
For trees with crown light exposure = 2-3, leaf area is calculated as the average of the leaf area from the open-grown (crown light exposure = 4-5) and closed canopy equations (crown light exposure = 0-1).
When crown dieback is present, estimates of leaf area are adjusted downward by multiplying by: 1 -percent crown dieback (0-1) More detailed methods are given in Nowak (2005) and Nowak et al. (2002bNowak et al. ( , 2008.
For a tree population assessment, two types of leaf area indices are calculated: 1) LAI p : total population leaf area divided by total study area size, and 2) LAI c : leaf area standardized per unit tree cover: LAI c = LAI p / percentage tree cover.

LEAF BIOMASS
Tree leaf biomass (weight of dry leaves) is calculated from leaf area estimates using species-specific conversion factors (appendix 4). Leaf biomass of shrubs is calculated as the product of the shrub crown volume occupied by leaves (m 3 ) and measured leaf 7 The 0.91 shading coefficient class is believed to be the best class to represent conifers as conifer forests typically have about 1.5 times more LAI than deciduous forests (Barbour et al. 1980), the average shading coefficient for deciduous trees is 0.83 (Nowak 1996) biomass factors (g/m 3 ) for individual species (appendix 5). Shrub leaf area is calculated by converting leaf biomass to leaf area based on measured species conversion ratios (m 2 /g). Due to limitations in estimating shrub leaf area by the crown-volume approach, shrub leaf area is not allowed to exceed an LAI of 18. More detailed methods are given in Nowak et al. (2002bNowak et al. ( , 2008.
If no shading coefficient is found for an individual shrub species, the average coefficient from the species' genus is used. If no genus coefficients exist, then next higher phylogenetic level average is used as available.

TREE BIOMASS
Total tree dry-weight biomass for each measured tree is calculated using allometric equations from the literature (see Nowak 1994b andNowak et al. 2002b). Equations that predict aboveground biomass are converted to whole tree biomass based on root-to-shoot ratio of 0.26 (Cairns et al. 1997). Equations that compute fresh-weight (also referred to as "green weight") biomass are multiplied by species-or genus-specific conversion factors to yield dry-weight biomass. These conversion factors, derived from average moisture contents of species reported in scientific literature, averaged 0.48 for conifers and 0.56 for hardwoods (see Nowak et al. 2002b).
Open-grown, maintained trees tend to have less aboveground biomass than predicted by forest-derived biomass equations for trees of the same d.b.h. (Nowak 1994b). To adjust for this difference, biomass results for open grown trees with a crown light exposure value of 4-5 are multiplied by a factor 0.8 (Nowak 1994b). Since deciduous trees drop their leaves annually, only carbon stored in wood biomass is calculated for these trees.
Where there are multiple equations available for individual species, they were combined to produce one predictive equation for a maximum range of diameters. These combined species equations produced results that are typically within 2 percent of the project-wide estimates obtained by using the separate equations with more limited diameter ranges. Formulas were combined to prevent large changes in sequestration or storage estimates for consecutive diameter classes, which can occur when changing biomass equations. These splined equations are provided in Table 3. If no biomass equation is found for an individual species, the average of results from equations of the same genus is used. If no genus coefficients exist, then next phylogenetic level average is used as available.

UNCERTAINTY IN TREE AND LEAF ESTIMATES
The estimation error in calculating leaf area, leaf biomass, and tree biomass is unknown, but model-derived estimates fall within range of measured values for forests and trees (e.g., McPherson 1998, 2003;Nowak et al. 2013a). The LAI p values for various cities typically range between 0.3 and 2.2 (average = 1.1) ). The LAI c average is 4.8 among measured cities (Nowak and Greenfield 2018). Typical LAI values for natural forests are 5-8 for deciduous forests and 9-11 for boreal conifer forests (Barbour et al. 1980). As urban forests are typically deciduous, 4.8 is a reasonable LAI value given the urban forest's more open structure (less vertical layering of trees, which decreases LAI). More leaf area to biomass conversion factors will be added to the i-Tree database based on McPherson et al. (2016). Leaf area estimation can likely be improved with more speciesspecific shading coefficients.   The biomass equations used in i-Tree have similar form and coefficient values to other carbon/biomass equations (Fig. 12). The average carbon density per unit of canopy cover in urban areas is 7.69 kg C/m 2 , slightly larger than in forest lands (7.24 kg C/m 2 ). However, the forest land estimate assumes 100 percent tree cover, which likely leads to an underestimate of carbon storage per unit of tree cover. i-Tree carbon densities are commensurate with estimates from numerous other studies (Nowak et al. 2013a).
As there are currently no urban tree biomass equations, forest-derived equations are used as proxies and should provide reasonable estimates as the tree diameter and height are measured. There are likely some differences in biomass between urban and forest-grown trees (e.g., McHale et al. 2009, Nowak 1994b, but more research is needed to determine actual differences or to develop specific equations for urban grown trees. Within i-Tree Eco, standard errors are reported for variables (leaf area, leaf biomass, tree biomass) that represent sampling error rather than error of estimation. As such, these sampling errors underestimate the total error of the estimates. Lack of information regarding errors in equations used to derive the secondary variables makes it impossible to fully account for estimation errors.

Planned Improvements
Add more biomass equations and species adjustments based on wood density (see planned future improvements in Carbon Storage and Sequestration).
Update shading coefficients based on new research from McPherson et al. (2018).
Remove Beer-Lambert equations for leaf area when CLE <2 and only use the regression approach. This change will be necessary for monitoring changes in leaf area. Only one approach can be used to avoid shifting equations when trees change CLE classes.
Add leaf biomass and nutrient estimates and values of annual leaf drop ).

PLOT MEASUREMENTS
In addition to measurements on each tree, other measurements are taken for each sampled plot, including: Portion of plot measured (%) Tree cover (%) Shrub cover (%)

Land use
Ground cover types (%) The percentage cover types are estimated to the nearest 5 percent class (e.g., 0-5%, 6-10%). While these estimates are fairly coarse, the precision of tree and other cover estimates can be improved through photo interpretation using i-Tree Canopy. This type of photointerpretation is often used instead of field plot estimates due to the increased precision. With 1,000 random points, which can be interpreted in about 1 day, the standard error of the cover estimate will be less than 1.6 percent. Details on calculating sampling error from photo-interpretation are given in Nowak (2011) and in the i-Tree Canopy section. More detailed information on plot sampling and measurements are given in the i-Tree Field Guide and User's Manual (i-Tree 2019a, 2019c).

SAMPLE TOTALS AND ERRORS
Sample totals and errors for metrics measured or derived from plot variables are based on standard statistical estimates of totals and variance from the plot data based on formulas and procedures detailed by a USDA Forest Service statistician. 9

Assessing Structural Metrics and Risks in i-Tree Eco TREE SPECIES DIVERSITY AND RANGE
i-Tree Eco estimates various tree species diversity indices and takes into account the native range of species.

Required Inputs
Tree species

Methods Overview
The following indices are calculated for the living tree population in plot sample projects: Shannon-Wiener Diversity Index is a widely used measure of diversity, based on information theory. This index assumes that the individual trees are randomly sampled from an infinite population. It also assumes that all species within a land use type or city have been sampled. This index is an indicator of species richness and has a moderate sensitivity to sample size (Magurran 1988) and, therefore, land uses and/or cities may not be comparable. The index is calculated as: Where p i is the proportion of trees in species i estimated as n N i . n i , and N is the total number of trees measured.

Menhinick's Diversity Index (D Mn )
) is an indicator of species richness and has high sensitivity to sample size (Magurran 1988) and, therefore, may not be appropriate for comparison between strata with widely different numbers of plots. The index is calculated as: S � = number of species recorded, and N = total number of trees measured.
Simpson's Diversity Index is an indicator of species dominance and has a low sensitivity to sample size (Magurran 1988) and, therefore, may be more appropriate for comparison among strata. The index is calculated as: Shannon-Wiener's Evenness Index assumes that all species within a land use type or city have been sampled. It is an indicator of species evenness and has a moderate sensitivity to sample size (Magurran 1988) and, therefore, may not be useful for comparisons among strata or cities.

E H S c / ln
Where H ' = Shannon-Wiener diversity index, and S = number of species.
Sanders' Rarefaction is a mathematical technique to standardize species richness from samples of several different populations with unequal sample sizes (e.g., different land use types within a single city) (Magurran 1988). This process is performed using rarefaction curves, i.e., scatter-plots of the number of species as a function of the sample size (i.e., the number of plots).
In i-Tree, strata (e.g., land use categories) represent the different populations, where each strata contains a different number of trees sampled. Rarefaction is stated as the number of tree species found within a standardized sample size over all strata (x species found from a sample of y individual trees). Typically, the standardized sample size is identical for each individual stratum (e.g., 10), but larger for the entire study area (e.g., 250). Sanders' Rarefaction is calculated as:

Importance Value
Each tree species' importance value (IV) is calculated as the sum of the species contribution to the total tree population and leaf area.
IV = (percent of total number of trees comprised by a species x 100) + (percent of total population leaf area comprised by a species x 100).

Native Status
The proportion of the tree population that originated from different parts of the United States and world was calculated based on the native range of each species (e.g., Burns and Honkala 1990a, b;Gleason and Cronquist 1991;Grimm 1962;Hough 1907;Little 1971Little , 1976Little , 1977Little , 1978Platt 1968;Preston 1976;Sunset Books 1979;Viereck and Little 1975).

Economic Valuation
None. More research is needed on the value of species diversity and how it can be modeled and assessed in a local context.

Advantages, Uncertainties, and Limitations
Tree species diversity indices provide an indication of the diversity of the tree population only. They do not account for total biodiversity.
Comparing diversity indices among strata is fraught with pitfalls. Unequal sampling effort is a prime issue, where each strata not only has a different number of plots, but also different number of individual trees sampled. Many of these diversity indices show at least a moderate sensitivity to sampling intensity at both the area and individual scale (Magurran 1988). While rarefaction attempts to ameliorate this issue of sampling sensitivity, other techniques, such as jackknifing and the pooled quadrat method, may be used in the future. In addition, though each strata is randomly sampled with plots, there is no guarantee of randomness for the individual trees on those plots. Each individual diversity index has its own assumptions (listed under each diversity index above) that may be violated by i-Tree Eco as well.

Planned Improvements
None.

POTENTIAL INSECT AND DISEASE EFFECTS
The risks to urban trees posed by 36 different insects and disease are estimated based on host-tree data and pest range maps in the United States. The 36 insects and diseases are listed in Table 4.
In addition to the risk the insect or disease poses to the tree's life, i-Tree Eco also estimates the impact to carbon storage and sequestration, leaf area and biomass, and structural value ( Table 4). The model calculates the potential maximum impact on these tree variables and also categorizes the pests based on how far the pest is away from the study area.

Required Inputs
Tree species   (Liebhold et al.1995, Maclure et al. 2001, McCambridge and Trostle 1970, Riffle and Peterson 1986, Sawyer 2011, Society of America Foresters 2011, USDA Forest Service 1985. Range maps for each of the 36 pests were obtained from the USDA Forest Service Forest Health Technology Enterprise Team (USDA Forest Service 2017b). For each U.S. county, a pest is classified as being present: a) within the county, b) within 250 miles of the county, c) between 250 and 750 miles from the county, or d) greater than 750 miles away.
i-Tree Eco reports the proportion and number of live trees susceptible to each pest that is found in the study area as well as the subsequent impacts on carbon storage and sequestration, leaf area and biomass, and structural value of the potentially affected trees.

Economic Valuation
The structural value of susceptible tree species for each pest is calculated (see Structural Valuation).

Advantages, Uncertainties, and Limitations
The potential impacts of numerous insects and diseases give an indication of risk to tree species, the tree population at large, and the services trees provide. The values given are the maximum potential impact, not the actual impact, which is likely less than the potential maximum. i-Tree Eco also provides a matrix of tree species and pests showing known risks to tree species, the weighted susceptibility of a tree species to the numerous pests, and distance of the pest from the study area/county.

Planned Future Improvements
Update species range maps and add new pests as information becomes available.

INVASIVE SPECIES
i-Tree Eco reports the total number and percentage of the tree population classified as an invasive species in the United States, as well as the associated leaf area.

Required inputs
Tree species Total height Height to crown base Crown width Percentage canopy missing: percentage of crown volume not occupied by branches or leaves

Methods Overview
Tree species data is cross-referenced to state invasive species lists (appendix 6); for states with no data, data from the n8earest state are used.

Economic Valuation
None. Research is needed on how the negative values associated with invasive species can be modeled and assessed across varying landscape types.

Advantages, Uncertainties, and Limitations
Invasiveness is dependent upon state lists; classifications vary among states.

Planned Improvements
Add invasive species summaries; global species invasive list is being compiled by Naomi Zurcher of Arbor Aegis.
Update state invasive species lists.

STRUCTURAL VALUATION
The structural valuation of trees is based on the trunk formula method of the Council of Tree and Landscape Appraisers (CTLA 1992). Structural value is based on the estimated cost of replacing a small tree or extending that cost to larger trees. This value is used for monetary settlement of compensation for damage or death of plants in litigation, insurance claims of direct payment, and loss of property value for tax implications.

Required Inputs
Tree species Land use

Methods Overview
Structural valuation is based on four tree/site characteristics: trunk area (cross-sectional area at d.b.h.), species, condition, and location. Trunk area and species are used to determine the basic value, which is then multiplied by condition (0-1) and location ratings (0-1) to determine the final tree structural value.
For transplantable trees, average replacement cost and transplantable size were obtained from International Society of Arboriculture (ISA) publications or contacting local ISA Chapters (ACRT 1997) to determine the basic replacement price ($/cm 2 of crosssectional area) for the tree. The basic replacement price from the state (or nearest state if no state data were available) is multiplied by trunk area and species factor (0-1) to determine a tree's basic value. The minimum basic value for a tree prior to species adjustment is set at $150. Local species factors were also obtained from ISA publications (appendix 7). If no species data are available for the state, data from the nearest state are used. If no species value is found for an individual species then the average from the same genus is used. If no genus value is found, the average of results from all broadleaf or conifer equations is used.
For trees larger than transplantable size the basic value (BV) is:

BV = RC + (BP x [TA A -TA R ] x SF)
Where RC (replacement cost) = cost of a tree at the largest transplantable size, BP (basic price) = local average cost per unit trunk area (dollars/cm 2 ), TA A = trunk area (based on d.b.h.) of the tree being appraised, TA R = trunk area of the largest transplantable tree, and SF = local species factor (0-1).
For trees larger than 76.2 cm in d.b.h., trunk area is adjusted downward based on the premise that a large mature tree would not increase in value as rapidly as its truck area.
The following adjusted trunk-area formula is determined based on the perceived increase in tree size, expected longevity, anticipated maintenance, and structural safety (CTLA 1992): More detailed methods are given in Council of Tree and Landscape Appraisers (1992) and Nowak et al. (2002a).

Non-U.S. Estimates of Structural Value
The CTLA process was also used for estimates in areas outside for the United States. However, this process required non-U.S. country specific species values and costs. In some countries, local estimates were provided by international partners, in other countries, average U.S. values were used. Appendix 7 details specifics on how non-U.S. countries were addressed in estimating compensatory values.

Advantages, Uncertainties, and Limitations
The estimates of structural value are not based on CTLA's latest procedures and thus should not be used for purposes apart from i-Tree assessments. A local tree professional should be used to determine local tree values for specific cases, including potential litigation. However this adaptation of the CTLA method provides reasonable estimates of structural values based on CTLA procedures. Tree and condition factors used in the process link to i-Tree condition classes and land uses. The condition classes are not the same as identified by CTLA but are a general proxy. The land use classes used relate to older versions of the CTLA process and not the most recent guides.

Planned Improvements
Update species and basic price values from state ISA organizations.
Update basic price and transplantable tree prices based on producer price index.

Assessing Ecosystem Services and Values in i-Tree Eco
i-Tree Eco estimates the effects of urban forests and trees on pollution removal and human health impacts, building energy use, carbon sequestration and storage, oxygen production, ultraviolet radiation mitigation, hydrology, and wildlife suitability. To see which tree field variables are used to estimate various ecosystem services and values, see Table 2.

AIR POLLUTION REMOVAL
Air pollution removal estimates are based on modeling of gas exchange and particulate matter interception by trees, shrubs, and grass for carbon monoxide (CO), nitrogen dioxide (NO 2 ), ozone (O 3 ), particulate matter less than 10 microns (PM 10 ), particulate matter less than 2.5 microns (PM 2.5 ), and sulfur dioxide (SO 2 ).

Required inputs
Tree, shrub, and grass cover Tree species

Methods Overview
i-Tree Eco estimates the hourly dry deposition of O 3 , SO 2 , NO 2 , CO, PM 10 and PM 2.5 throughout the year based on tree, shrub and grass cover data, hourly National Climatic Data Center (NCDC) weather data and U.S. Environmental Protection Agency (U.S. EPA) pollution-concentration monitoring data. Missing hourly pollution data are filled in based on procedures detailed in Hirabayashi and Endreny (2016). Weather station data quality information is detailed in Hirabayashi (2017). Daily particulate matter data are used as hourly inputs (i.e., daily average is used for each hour of the corresponding day). If multiple monitors exist, the average of all monitor data are used. Missing hourly pollution data are filled in based on procedures detailed in .
Pollution removal, or downward pollutant flux (F; in g/m 2 /s), is calculated as: Deposition velocity is calculated as (Baldocchi et al. 1987): Where R a = sum of the aerodynamic boundary layer, R b = quasi-laminar boundary layer, and (R c ) = canopy resistances.
Hourly estimates of R a and R b are calculated using standard resistance formulas (Killus et al. 1984, Nowak et al. 1998, Pederson et al. 1995 and hourly weather data. R a and R b effects are relatively small compared to R c effects. Hourly canopy resistance values for O 3 , SO 2 , and NO 2 are calculated based on a modified hybrid of big-leaf and multilayer canopy deposition models (Baldocchi 1988, Baldocchi et al. 1987. Canopy resistance (R c ) has three components: stomatal resistance (r s ), mesophyll resistance (r m ), and cuticular resistance (r t ), such that: 1/R c = 1/(r s +r m )+1/r t Mesophyll resistance is set to 0 seconds/meter (s/m) for SO 2 (Wesely 1989) and 10 s/m for O 3 (Hosker and Lindberg 1982). Mesophyll resistance is set to 100 s/m for NO 2 to account for the difference between transport of water and NO 2 in the leaf interior and to bring the computed deposition velocities in the range typically exhibited for NO 2 (Lovett 1994). Base cuticular resistances are set at 8,000 s/m for SO 2 , 10,000 s/m for O 3 , and 20,000 s/m for NO 2 to account for the typical variation in r t exhibited among the pollutants (Lovett 1994).
Hourly inputs to calculate canopy resistance are photosynthetic active radiation (PAR; μE/m 2 /s), air temperature (K o ), windspeed (m/s), frictional velocity (u * ) (m/s), CO 2 concentration (set to 390 ppm), and absolute humidity (kg/m 3 ). Air temperature, windspeed, u * , and absolute humidity are measured directly or calculated from measured hourly NCDC meteorological data. Total solar radiation is calculated based on the meteorological/statistical (METSTAT) model with inputs from NCDC data (Maxwell 1994). PAR is calculated as 46 percent of total solar radiation input (Monteith and Unsworth 1990).
As removal of CO and particulate matter by vegetation is not directly related to transpiration, R c for CO is set to a constant for in-leaf season (50,000 s/m) and leaf-off season (1,000,000 s/m) based on data from Bidwell and Fraser (1972). For PM 10 , the median deposition velocity from the literature (Lovett 1994) is 0.0128 m/s for the in-leaf season. Base particle V d is set to 0.064 based on a LAI of 6 and a 50 percent resuspension rate of particles back to the atmosphere (Zinke 1967). The base V d is adjusted according to actual LAI and in-leaf vs. leaf-off season parameters. For PM 2.5, hourly deposition velocities and resuspension rates vary with wind speed as detailed in Nowak et al. (2013b).
To limit deposition estimates to periods of dry deposition, deposition velocities are set to 0 during periods of precipitation. The model is run at the population scale to estimate pollution effects. Average hourly pollutant flux (g/m 2 of tree canopy coverage) among the pollutant monitor sites is multiplied by total tree-canopy coverage (m 2 ) to estimate total hourly pollutant removal by trees across the study area.
Daily leaf area is determined based on: Percentage evergreen tree species Leaf-on / leaf-off dates -based on first and last frost dates from the NCDC (2005) for the United States and Weather Online (2016) for international locations.
Daily LAI -maximum LAI is derived from field data. Daily estimates combined LAI values with percentage evergreen information and local leaf-on and leaf-off to estimate total daily leaf surface area assuming a 4-week transition period centered on leaf-on and leaf-off dates for spring and autumn, respectively. Detailed methods are given in Hirabayashi and Nowak (2016).
Limits of total tree removal of O 3 , NO 2 , SO 2 , and PM 10 are estimated using the typical range of published in-leaf dry deposition velocities (Lovett 1994). The ability of individual trees to remove pollutants is estimated for each diameter class using the formula (Nowak 1994a): Where I x = pollution removal by individual tree x (kg), R t = total pollution removed for all trees (kg), LA x = total leaf area of tree x (m 2 ), and LA t = total leaf area of all trees (m 2 ).
This formula yields an estimate of pollution removal by individual trees based on leaf surface area (LA), the major surface for pollutant removal.
To estimate percentage air quality improvement due to the removal of dry deposition (Nowak et al. 2000), mixing height data 10 from closest radiosonde station (from NOAA's Earth System Research Laboratory) are used in conjunction with local deposition velocities. Daily morning and afternoon mixing heights from the closest station is interpolated to produce hourly values using the U.S. EPA's PCRAMMIT program (U.S. EPA 1999). Minimum boundary-layer heights are set to 150 m during the night and 250 m during the day based on estimated minimum boundary-layer heights in cities. Hourly mixing heights (in meters) are used in conjunction with pollution concentrations (μg/m 3 ) to calculate the amount of pollution within the mixing layer (μg/m 2 ). This extrapolation from ground-layer concentration to total pollution within the boundary layer assumes a well-mixed boundary layer, which is common in daytime (unstable conditions) (Colbeck and Harrison 1985). Hourly change in pollution concentration is calculated as: Percentage air quality improvement is calculated as:
Health benefits, and the associated economic value of these benefits, from the removal of pollutants NO 2 , SO 2 , O 3 , and PM 2.5 , are made based on the U.S. EPA's Environmental Benefits Mapping and Analysis Program (BenMAP) model (U.S. EPA 2012) with methods detailed in Nowak et al. (2013a. Based on BenMAP, various standardized health impacts and dollar values were calculated for each U.S. county. The standardized values were calculated for each pollutant as the impact or value/person/pollutant concentration change using local pollution and population data. These values are multiplied by the corresponding local population total with a county from U.S. Census data and pollution concentration change due to trees and other vegetation in the study area to determine health impacts and associated dollar values. More detailed documentation and formulas are given in Hirabayashi et al. (2015) and Hirabayashi (2016).

Economic Valuation
Economic valuation of pollution removal is estimated in one of two ways: 1) Externality values. These values can be considered the estimated cost of pollution to society that is not accounted for in the market price of the goods or services that produced the pollution. i-Tree Eco uses estimates of externality values (Murray et al. 1994, Ottinger et al. 1990) for the valuation of CO ($1,599/tonne 11 in 2011 dollars); these values are updated based on the producer price index (U.S. Bureau of Labor Statistics 2017).
2) Health values. i-Tree estimates the number of incidents avoided and the total dollar value of several health factors related to four other pollutants: NO 2 , SO 2 , O 3 , and PM 2.5 . The estimates are based on health-care expenses (i.e., cost of illness), productivity losses associated with specific adverse health events, and on the value of a statistical life in the case of mortality as derived from the U.S. EPA BenMAP model , U.S. EPA 2012 For international estimates, regression equations ) based on population density can be used to estimate a dollar value per ton of pollution removal.

Advantages, Uncertainties, and Limitations
The i-Tree model produces deposition velocities and daily flux patterns that coincide with measured field data (e.g., Lovett 1994, Morani et al. 2014). The hourly routines allow for interactions between pollution concentrations and deposition velocities to be quantified, which are missed using annual, monthly, or daily average data. Sensitivity analyses of dry deposition velocity reveals that estimates of these velocities are most sensitive to the inputs of air temperature and leaf area index (Hirabayashi et al. 2011).
Pollution removal is modeled based on hourly measured pollution concentrations. The concentrations often come from only one local monitor, or the average of several monitors. Thus only one concentration value per hour is used to represent the entire study area. This approach has limited spatial resolution of concentration data in the study area, which is necessitated by limited spatially specific pollution concentration data. If more spatially refined pollution data exist, the model can run multiple times using various zones within the city with different pollution concentration data.
Health impacts are dependent on the estimations based on the U.S. EPA BenMAP model, which utilizes epidemiological data to estimate health impacts due to changes in pollution concentrations. Coupling BenMAP with i-Tree estimates of air pollution removal by trees is the only known program that estimates health benefits of trees' ecosystem service of pollution removal. Health estimates can be estimated for any area in which local population data by age class and tree cover data exists. These health-benefit estimates would be locally specific, except for pollution concentration data, which may be developed from a larger geographic area (i.e., concentration data are necessarily specific to each local geography).
This process accurately estimates pollution removal. However, it only estimates pollution removal and the removal effect on local pollutant concentrations; it does not account for tree impacts regarding: Local scale interactions with wind. Trees alter wind patterns that can either increase or decrease concentrations at the local site scale.
Potential pollution formation due to volatile organic compounds emissions.
Potential health impacts due to tree pollen.
Drought. i-Tree Eco assumes ample soil moisture so that gas exchange can occur. During drought, the module will overestimate pollution removal of NO 2 , SO 2 , and O 3 . However, even if drought effects are modeled, the important question in urban areas is to know when the trees actually have drought effects. Even though a natural forest may experience drought, urban trees often do not have drought effects due to watering of trees by humans and possible leaky pipes that supplement natural soil moisture.
Estimates of interception and resuspension of PM 2.5 are based on local leaf area, windspeed, precipitation, and pollution concentrations (Nowak et al. 2013b). Estimates of removal of PM 10 and CO are rather rudimentary and based on average deposition velocities or, in the case of CO, very limited research. More research is needed regarding tree impacts on PM 2.5 , PM 10 , and CO concentrations so that estimates can be improved. In addition, more spatially refined pollution concentration data would aid in assessing more locally specific tree effects.

Planned Improvements
Add new weather and pollution data years as they become available.
PM 10 estimation will be included in the future due to limited global PM 2.5 data. PM 10 estimates were dropped in 2016 in favor of using PM 2.5 .
Develop and add a drought routine to account for the limited gas exchange during droughts.
Undertake xylem conduit analyses to help separate species differences in gas exchange to help determine differences in pollution removal among tree species.
Add new externality values (Korzhenevych et al. 2014, van Essen et al. 2011) as an option for economic valuation.
Update economic values to most recent dollar year.
Add new pollen allergy index based on species composition and sizes.
Update CO 2 concentration from 390 ppm.

BUILDING ENERGY USE AND EMISSIONS
i-Tree Eco estimates the effects of trees on energy consumption and pollutant emissions from residential buildings. The effects are broken down by building energy use and energy influence type.

Methods Overview
Methods for energy estimates are based on . For each tree within 18 m of a one-or two-story residential building, information on distance and direction to the building is recorded. Any tree that is less than 6 m in height or farther than 18 m from a building is considered to have no effect on building energy use.
Using the tree size, distance, direction to building, climate region, leaf type (deciduous or evergreen), and percentage cover of buildings and trees on the plot, the amount of carbon avoided from power plants due to the presence of trees is estimated based on methods in  i-Tree adjust for varying energy effects due to poor tree condition. The default energy effect values are adjusted as follows: Energy adjustment = 0.5 + (0.5 x tree condition)
This adjustment factor is applied to all tree energy effects for cooling, but only evergreen trees for the effects for heating.
As tree cover increases in an area, the individual tree effect on climate diminishes, though the cumulative effect of all trees can increase. Base climate effect values for a tree are given for plots of 10, 30 and 60 percent cover (McPherson and Simpson 1999) and interpolation formulas are used to determine the actual tree value based on the percentage of tree and building cover on a specific plot. For plots with less than 10 percent cover, the slope between the 10 and 30 percent cover values is used for the interpolation. Plots with percent cover greater than 60 percent used the slope between 30 and 60 percent cover with a minimum individual tree climate effect of one-third the effect at 60 percent cover. This minimum is set to prevent a tree from obtaining a zero or negative effect at high cover (100 percent).   to estimate total emissions associated with changes in energy use. Plot effects are combined to yield the total energy and associated emissions effects due to the urban forest.

Economic Valuation
The Various approaches are used to estimate the values of the changes in emissions. The CO 2 value varies through time based on year and adjustments to the dollar value based on the producer price index. The current CO 2 value is estimated at $51.23 per tonne based on the estimated social costs of carbon for 2020 with a 3 percent discount rate updated to 2018 dollars (Interagency Working Group 2016). The CH 4 value is estimated at $980 per tonne based on the ratio of the estimated social costs of methane to carbon dioxide (ratio =24.5) for 2010 with a 3 percent discount rate (Marten and Newbold 2011). This ratio is applied to the most recent CO 2 value to update the value for CH 4 . Social costs estimate the monetized damages associated with incremental increases in emissions and include changes in net agricultural productivity, human health, property damages from increased flood risk, and the value of ecosystem services due to climate change (Interagency Working Group 2016).
Pollution removal value for CO is estimated using national median externality values ($1,599/tonne in 2011 dollars) (Murray et al. 1994). Externality values can be considered the estimated cost of pollution to society that is not accounted for in the market price of the goods or services that produced the pollution.

Non-U.S. Estimates of Building Energy Effects
The i-Tree process for estimating tree effects on building energy use were developed for U.S. climate, buildings, and fuel use. Using this procedure outside the United States has significant limitations. However, numerous users insist on some type of estimation of energy impacts from i-Tree outside the United States. Thus, the i-Tree model has been applied in other countries, but with the following warning: Because this model component is designed specifically for the U.S., its utility is limited in international applications. International users will receive energy results that are based on the characteristics of the user-defined U.S. climate region, including emission factors, typical construction practices and building characteristics, and energy composition (i.e., type of and amount used). Therefore, results should be used with caution as they assume that the building types, energy use, and emission factors of the U.S. are the same as those internationally (i-Tree 2019b).
The only local values used in the estimates outside the United States are electricity and fuel costs. The remainder of the estimation is based U.S. conditions from the assigned climate zone. Details on local energy values and the comparisons between international areas and U.S. climate zones is given in appendix 9.

Advantages, Uncertainties, and Limitations
i-Tree Eco uses results from building energy simulations from numerous climates across the United States with tree results varying based on tree size and position around the building. i-Tree Eco relies on state-specific energy and emission values. However, it is unknown how well these values represent actual energy savings, so it is difficult to validate tree effects on energy use, leading to uncertainties. The results are likely reasonable as the model simulations appear reasonable, but the uncertainty of the estimates is unknown.

Planned Improvements
Add emission amounts and values for CH 4 , CO, NO x , SO 2 , VOCs, PM 10 , and PM 2.5 .
Update energy values.
Work with SUNY College of Environmental Science and Forestry to develop a new, more robust energy program.
Add local health values based on county-based BenMAP values.

CARBON STORAGE AND SEQUESTRATION
Carbon dioxide (CO 2 ) in the atmosphere is necessary for plants and trees to grow. Trees absorb carbon dioxide during photosynthesis, storing carbon and producing oxygen as a byproduct of photosynthesis. Carbon sequestration is the process of removing carbon from the atmosphere and storing it in a physical element (e.g., a tree). i-Tree Eco estimates carbon storage in trees, annual carbon sequestration, and emission of carbon via tree decomposition.

Required inputs
Tree species D.b.h.

Total tree height
Crown dieback Crown light exposure

Methods Overview
Carbon storage is estimated by multiplying tree biomass by 0.5 (Chow and Rolfe 1989).
To prevent carbon storage from overestimation for very large trees, total carbon storage is capped 7,500 kg of carbon in i-Tree Eco and Forecast. To estimate annual gross carbon sequestration, the tree d.b.h. is incrementally increased in the computer model based on an estimated annual growth rate. The carbon storage in the current year (year 0) is then contrasted with carbon storage in the next year (year 1) to estimate the annual sequestration. If a tree's biomass estimate was capped at 7,500 kg and the tree is alive, carbon sequestration for these large trees is estimated at 25 kg/year.

Tree Growth
Annual tree diameter growth is estimated for the study area based on: 1. Base growth rate-Open grown tree growth rates were based on measured street tree growth (Fleming 1988, Frelich 1992, Nowak 1994b, which were standardized to a 153 day frost free length as follows: standard growth = measured growth × (153/number of frost-free days of measurement).
The average diameter growth rate for open-grown trees with 153 frost free days is 0.33 in/yr.

2.
Length of growing season-To determine a local base growth rate, the standard growth rate was adjusted based the local length of growing as follows: base growth = standard growth × (number of frost-free days in area/153).
3. Species growth rates-Based on these data, the average standardized diameter growth rates for open-grown trees with 153 frost free days are set to 0.23 in/yr for slow growing species, 0.33 in/yr for moderate growing species, and 0.43 in/yr for fast growing species (species information can be seen and input via i-Tree Database, (https://database. itreetools.org/#/splash). There are limited measured data on urban tree growth for slow, moderate, or fast-growing tree species, so the growth rates used in i-Tree Forecast are estimates.
4. Tree competition-Crown light exposure (CLE) measurements are used to represent tree competition. CLE measurements for each tree are based on the number of sides and/ or top of tree exposed to sunlight. A CLE of 0-1 represents forest conditions with a closed, or nearly closed canopy, where none or one side of the tree is exposed to sunlight. CLE of 2-3 represents park conditions and 4-5 represents open-grown conditions. Based on a comparison of species growth rates between street trees (CLE 4-5), park trees (CLE 2-3) and forest-grown trees (CLE 0-1) (deVries 1987, Fleming 1988, Frelich 1992, Nowak 1994b, Smith and Shifley 1984, the base growth for trees are as follows: trees with CLE 0-1 = Standardized growth (SG) × 0.44; trees with CLE 2-3 = SG × 0.56, and trees with CLE 4-5 = SG × 1.
5. Tree condition-Growth rates are adjusted for tree condition based on percentage crown dieback. Base growth rates are multiplied by 1 -percentage dieback. For example, a tree with 40 percent dieback, base growth rates are multiplied by 0.6. 6. Tree height-As a tree approaches "maximum" height, growth rate decreases. Thus, the species growth rates as described above are adjusted based on the ratio between the current height of the tree and the average height at maturity for the species. The estimated tree height at maturity is derived from the literature. When a tree's height is more than 80 percent of its average height at maturity, the annual diameter growth is proportionally reduced from full growth at 80 percent of maximum height to 2.22 percent of full growth at 125 percent of height at maturity.

Decomposition
An estimation of the carbon lost due to more rapid carbon release (e.g., mulching of tree components and burning) and delayed release (e.g., decomposition) is calculated and subtracted from the gross sequestration to estimate net sequestration. To estimate carbon release, various assumptions are made related to probability of mortality, probability of recording a dead tree, and decomposition rates. More detailed information on storage, gross and net sequestration, and tree growth can be found in Nowak et al. 2002bNowak et al. , 2008.

Economic Valuation
The value of carbon storage and sequestration is based on the social cost of carbon as reported by the Interagency Working Group on Social Cost of Carbon (2016). Social cost associated with a pollutant (e.g., CO 2 ) refers to an estimate of total (global) economic damage attributable to incremental increase in the level of that particular pollutant in a given year. The current CO 2 value is estimated at $51.23 per tonne based on the estimated social costs of carbon for 2020 with a 3 percent discount rate to reflect 2018 dollars (Interagency Working Group 2016). Users can adjust this value, if they so desire, by taking a ratio of the desired value (DR) per tonne CO 2 to the $51.23/tonne CO 2 (updated value = i-Tree reported value x DR/51.23).

Advantages, Uncertainties, and Limitations
The advantages and limitations associated with carbon storage estimates are the same as with tree biomass estimates (provided earlier) as carbon storage is directly related to biomass. Overall the storage estimates are reasonable and the standardized values per unit tree cover are comparable to estimates for U.S. forests and from other cities around the world (Nowak et al. 2013a). National estimates of urban forest carbon storage and sequestration have been estimated through the years using this procedure (Nowak 1993, Nowak and Crane 2002, Nowak et al. 2013a). Estimates of storage could be improved with additional biomass equations (see planned future improvements below), and specifically, biomass equations developed for urban conditions. Capping of total storage at 7,500 kg for large trees is also a limitation if the tree stores greater than 7,500 kgC, however, the cap prevents extremely large and uncapped estimates from occurring via the logarithmic biomass equations used.
Estimates of gross carbon sequestration are dependent upon good biomass and tree growth equations. Growth rates for urban trees will range between 0 cm d.b.h./yr (dead trees) to 2.54 cm d.b.h./yr for fast-growing, open-grown healthy trees in areas with no frost. Estimated growth rates are average rates where rates of individual trees may be higher or lower than the estimated class average. These growth rates for "moderate" trees are within range for measured urban and forest tree growth (e.g., deVries 1987, Fleming 1988, Frelich 1992, Nowak 1994b, Smith and Shifley 1984, Wood 2010. For i-Tree programs that simulate long-term tree growth and carbon sequestration (i.e., i-Tree Forecast, Planting, Design), growth rate estimates can have a substantial impact on carbon sequestered. The estimated growth is based on the tree species, condition, and crown light exposure of the measured tree. Better long-term growth rates for trees will help improve growth estimates.
Net sequestration is based on gross sequestration minus losses due to decomposition. Decomposition estimates are quite rudimentary and are based on various assumptions of mortality and decomposition rates. Improved research on decomposition rates, how wood is decomposed (e.g., burn, mulch, natural decomposition), and mortality rates for urban trees are needed to enhance the net sequestration estimates.

Planned Improvements
i-Tree is adding 83 new carbon equations from across the globe (appendix 10; Fig. 12) from the GlobAllomeTree (2017) database. Not all carbon equations are used from this database. Only equations that produced whole tree carbon estimates within 2-12 t C at 100 cm d.b.h. were selected to remove outliers (Fig. 13). Wood density data (appendix 11) are also being added to the i-Tree database from the Global Wood Density Database (Chave et al. 2009) and McPherson et al. (2016. The new carbon equations and wood density data will produce carbon estimates based on more equations and a process of weighting wood densities between the actual species measured and the species equation used:

C est = C eq x WD spp / WD eq
Where C est = carbon estimate, C eq = carbon estimates derived from equations, WD spp = wood density of the species measured, and WD eq = average wood density from the carbon equations.
This weighting process and new equations will be incorporated in 2020.
Tropical biomass equations from wet, moist, and dry tropical forests (Chave et al. 2005) will also be added to be used in tropical cities instead of current biomass equations. Potential wood products derived from removed trees is also being incorporated . i-Tree Eco will also incorporate species and height variations as detailed in the tree growth estimates section that is used in other programs to estimate tree growth.

FOOD PROVISIONING SERVICES
Individual species characteristics about fruit, flowers, pollinators, and potential use for pharmaceuticals, fiber, food and fuel are provided for 44 tree species if these species are found within the local assessment. This information is provided to communicate the value of various species and aid in appropriate species selections. These data were gathered by the Community Food Forestry Initiative (Inhabit.earth, n.d.).

Planned Improvements
Add information for more tree species.

OXYGEN PRODUCTION
As trees capture carbon dioxide during the photosynthesis process, they give off oxygen. Likewise, when carbon dioxide is released through decomposition, oxygen is consumed. The oxygen production estimates directly relate to estimates of carbon sequestration. , estimates of carbon range between about 1.2 t C to 5.9 t C, depending upon species equation. The equation estimates will not vary much as most of the urban forest is typically less than 15.2 cm (6 inches). However, for the fewer large trees, the estimates can vary by as much as fivefold.

Required Inputs
Same as for Carbon Storage and Sequestration: Tree species D.b.h.

Total tree height
Crown dieback

Methods Overview
The amount of oxygen produced is estimated from carbon sequestration based on atomic weights:

Economic Valuation
i-Tree assigns the value of oxygen production at $0/tonne. The reason the oxygen production value of trees is insignificant is due to the large amount of oxygen within the atmosphere (approximately 21 percent of the atmosphere's volume is oxygen) and because species of algae are estimated to replace about 90 percent of all oxygen used. Thus, though trees do produce significant amounts of oxygen, it is not a significant ecological benefit (see Nowak et al. 2007b).

Advantages, Uncertainties, and Limitations
Same as for Carbon Storage and Sequestration.

Planned Improvements
None.

STREAM FLOW AND WATER QUALITY
i-Tree Eco estimates the amount of rainfall intercepted, stored, transpired, and evaporated by urban forest tree canopies as well as the volume of runoff avoided due to trees' presence.

Species
Total height Height to crown base Crown width Percentage canopy missing: percentage of crown volume not occupied by branches or leaves Total tree cover

Methods Overview
Based on leaf/bark area data and local hourly weather data, i-Tree Eco estimates hourly rain interception, evaporation from leaf surfaces, potential evapotranspiration, transpiration, and avoided runoff values. These calculations are process-based, meaning each process is simulated individually and then linked with other processes. For example, interception is simulated using an improved Rutter methodology (Valente et al. 1997), and evaporation is simulated based on the work of Deardorff (1978) and Noilhan and Planton (1989). Estimates are generated based on current tree conditions and then without trees in order to estimate the impact of trees on surface runoff. Impervious cover beneath trees is assumed to be 25.5 percent, the national average impervious cover (Nowak and Greenfield 2012). To estimate individual tree effects, the water impacts across the entire population of trees are prorated back to the tree-scale proportional to tree leaf area of the individual tree.

Economic Valuation
i-Tree Eco uses the U.S. national average dollar value of $0.008936/gallon to estimate the value avoided runoff due to trees. The value is based on 16 research studies on costs of storm water control and treatment (McPherson et al. , 2000(McPherson et al. , 2001(McPherson et al. , 2002(McPherson et al. , 2004(McPherson et al. , 2006a(McPherson et al. , 2006b(McPherson et al. , 2006c(McPherson et al. , 2007Peper et al. 2009Peper et al. , 2010Vargas et al. 2007aVargas et al. , 2007bVargas et al. , 2008.

Advantages, Uncertainties, and Limitations
i-Tree Eco offers fast and easy access to robust, process-based hydrology estimates. Hydrologic models that are easier to use often do not have robust, process-based estimates, while models with more robust methods tend to require more technical expertise. Few hydrologic models explicitly simulate eco-hydrology of trees (Coville et al. 2020).
While this is one of the easiest-to-use tree-based simulations of hydrology, it has limitations: it simplifies surface and subsurface hydrology to focus on the effects of trees. i-Tree Hydro provides more comprehensive processing of hydrologic effects than i-Tree Eco. Eco's subsurface routines are simplified and do not take into account varying amounts of impervious cover. Both i-Tree Eco and i-Tree Hydro do not account for the effects of different spatial arrangements of trees or other land cover. They are statistically rather than spatially distributed rainfall-runoff models, accounting for the amount of tree cover relative to other land cover types.
i-Tree Eco also uses default soil and hydrologic parameters (e.g., soil texture class) for the nation. This is suitable for first-order comparisons between different land cover amounts in any given area, but it comes with uncertainty about how well those defaults describe local conditions in various areas.
Economic valuation of reduced runoff is also a limitation with i-Tree Eco as it uses an estimated average storm water control cost, which is a very rough approximation of value.
Approaches and estimates of the economic valuation of change in stream water flow and quality are limited. As better estimates become available, they will be used. Users can also use local values, if known, by using a ratio of the local value to the model default value ($0.008936/gallon runoff).

Planned Future Improvements
Add water quality effects of trees based on runoff volume and local watershed event mean concentration (EMC) values.
Add variable impervious beneath trees.
Update economic values to most recent dollar year.

VOLATILE ORGANIC COMPOUND EMISSIONS
Volatile organic compounds (VOCs) are emitted by trees and are precursor chemicals to ozone and other pollutant formation. Thus, VOC emissions can be considered a disservice via the potential formation of ozone and other pollutants. Species have varying emission rates based on leaf biomass and local meteorological conditions.

Species
Total height Height to crown base Crown width Percentage canopy missing: percentage of crown volume not occupied by branches or leaves

Methods Overview
Tree VOC emissions are computed based on procedures used in the EPA's Biogenic Emissions Inventory System (BEIS) (U.S. EPA 2017a). The VOC emissions depend on tree species, leaf biomass, air temperature, and other environmental factors. i-Tree Eco estimates the hourly emission of isoprene (C 5 H 8 ) and monoterpenes (C 10 terpenoids), the two dominant VOC categories emitted by trees. Leaf biomass estimates are derived from the field data and multiplied by genus-or family-specific emission factors (Nowak et al. 2002b) to produce emission levels standardized to 30 o C and photosynthetically active radiation (PAR) flux of 1,000 µmol/m 2 /s. The genus-or family-specific emission factors were derived from the literature (appendix 12). If genus-specific information is not available, median emission values for the family, order, or superorder are used. Standardized emissions are converted to actual emissions based on light and temperature correction factors (Geron et al. 1994) and local meteorological data.
VOC emission (E) in μgC/tree/hr for isoprene and monoterpenes is estimated as: Where B E is the base genus emission rate in μgC/g leaf dry weight/hr at 30 o C and PAR flux of 1,000 µmol/m 2 /s, B is species leaf dry weight biomass (g), and for isoprene emission is calculated as: Where α = 0.0027, R is the ideal gas constant (8.314 J/K/mol), T(°K) is leaf temperature, which is assumed to be air temperature, T M = 314 °K, C T1 = 95,000 J/mol, and C T2 = 230,000 J/mol (Geron et al. 1994, Guenther 1997, Guenther et al. 1995.
As PAR strongly controls the isoprene emission rate, PAR is estimated at 30 canopy levels as a function of above-canopy PAR using the sunfleck canopy environment model 12 with the LAI derived from the field measurements.
Hourly inputs of air temperature are from measured National Climatic Data Center (NCDC) meteorological data. Total solar radiation is calculated based on the National Renewable Energy Laboratory Meteorological/Statistical Solar Radiation Model (METSTAT) with inputs from the NCDC data set (Maxwell 1994). PAR is calculated as 46 percent of total solar radiation input (Monteith and Unsworth 1990). Hourly VOC estimates are summed to produce a VOC emission rate per genera. To estimate individual tree VOC emissions, the total genera emission is prorated back to the tree or species level proportional to individual tree or species biomass in the genera. More detail methods are given in Hirabayashi (2012Hirabayashi ( , 2016.

Economic Valuation
Not assessed; VOC emissions need to be converted to ozone impacts prior to being valued.

Advantages, Uncertainties, and Limitations
The VOC emission values produced by i-Tree Eco are closely aligned with standardized values produced by U.S. land use inventory (Kinnee et al. 1997). The modeled pattern of photosynthetically active radiation (PAR) also closely matches measures of PAR radiation. However, this module only estimates VOC emissions and does estimate ozone formation from VOC emissions to contrast against ozone removal by trees. Photochemical models are likely the best means to determine the ultimate effects of VOC emissions on ozone formation.

Planned Improvements
Add simplified means of converting VOC emissions to ozone formation.

ULTRAVIOLET RADIATION
i-Tree Eco quantifies urban tree effects on mitigating the intensity of ultraviolet (UV) radiation on the ground within different land use types across a study area. The ultraviolet index or UV index is an international standard measurement of the strength of sunburnproducing UV radiation at a particular place and time. It is directly proportional to the intensity of UV radiation that causes sunburn on human skin. A higher UV index value represents a greater risk of sunburn.

Required Inputs
Percentage tree cover

Methods Overview
To estimate tree effects on reducing UV radiation, four datasets are required: Canopy cover (from field data) UV index values from Tropospheric Emission Monitoring Internet Service (TEMIS 2016) for the years 2008-2013. TEMIS provides near-real time data on total ozone and surface UV data.
Hourly cloud cover from local weather data Solar zenith angle data, calculated internally based on location.
These data are combined with equations (Grant and Heisler 2006, Grant et al. 2002, Heisler et al. 2003a to predict the UV protection factor and changes in the UV index due to trees within each city land use for mid-day conditions for an entire year. Detailed methods are provided in Na et al. (2014).
i-Tree Eco estimates the following factors for each land use and the overall study area based on average tree cover in the land use: Protection factor-a unitless value meant to estimate the UV radiation-blocking capacity of trees. It is comparable to the SPF rating in sunscreen and is calculated as unshaded UV index divided by shaded UV index.
Reduction in UV index-the change in UV index as the result of trees and calculated as unshaded UV index minus shaded or overall UV index.
Percent reduction-the reduction in UV index expressed as a percentage change and calculated as the reduction in UV index divided by unshaded UV index.
Two types of UV indices are produced: Overall UV index-based on a person in tree shade proportional to the amount of tree cover in the area (average effect). The likelihood of being in the shade increases as tree cover increases.
Shaded UV index-based on a person always being in tree shade (maximum effect).

Economic Valuation
Not assessed. Research results on the health values associated with reduced UV radiation need to be incorporated within i-Tree.

Advantages, Uncertainties, and Limitations
i-Tree Eco builds on the equations from Grant et al. (2002) to estimate the below-canopy reduction in UV radiation across a study area. i-Tree Eco provides an easy means to estimate tree impacts on UV radiation that coincides with estimates from other studies and field measurements. However, the uncertainty of the estimates are unknown. Potential limitations are that it assumes the trees are evenly spaced across the landscape, thus shade effects could be overestimated compared to scenarios where all of the tree cover is aggregated together (e.g., in a forest stand). The module also does not include the effects of local aerosols and building shade on altering UV radiation loads in the study area.

Planned Improvements
Add health impacts due to reduced UV radiation.

Required Inputs
Plot land use

Total height
Crown dieback: percentage crown dieback in live crown area

Methods Overview
Bird habitat models developed by Lerman et al. (2014) are incorporated into i-Tree Eco. Each model predicts a habitat suitability index between 0 and 1, where a score of 0 indicates unsuitable habitat conditions (i.e., strong likelihood the species is not present) and a score of 1 indicates the habitat conditions have a strong likelihood of supporting the species. i-Tree Eco uses a suitability index for each bird species listed above (if the location is within the native range of bird species) under current conditions and also with all trees removed to determine what impact the current forest structure has on habitat suitability. Details and equations are given in Lerman et al. (2014).

Economic Valuation
None. Research is needed on how a value from numerous wildlife species can be modeled and assessed in a local context.

Advantages, Uncertainties, and Limitations
Model results have been compared to field data in Baltimore; these results support the efficacy of using the habitat models to predict the habitat quality of urban areas for a variety of species. However, the model could not be validated for the Baltimore oriole and scarlet tanager model at this time. Still, these untested models have greater value than no information regarding these species' habitat relationships (Lerman et al. 2014).
Limitations to the habitat assessment approach are detailed in Lerman et al. (2014).

Planned Improvements
Add more wildlife habitat models to i-Tree Eco. Table 5 summarizes the estimation uncertainty of i-Tree Eco. Table 6 summarizes the economic valuation procedures used in i-Tree. Table 7 summarizes the monetary values used in i-Tree. Not all services are valued or assessed in i-Tree (e.g., noise reduction, aesthetics, temperature reduction), so the valuation is conservative. In many cases, users can adjust these values if they have better or more local values. OK based on comparison with measured leaf area index values and field tests (e.g., McPherson 1998, 2003) Leaf biomass Direct conversion from leaf area using species values (g/m 2 ) leaf area. Unknown uncertainty, but likely minimal as crown parameters are measured. Range of uncertainty of removal is given; produces deposition velocities and daily flux patterns that coincide with measured field data (e.g., Lovett 1994, Morani et al. 2014 , 2004, 2006a, 2006b, 2006cPeper et al. 2009Peper et al. , 2010Vargas et al. 2007aVargas et al. , 2007bVargas et al. , 2008 Other i-Tree Tools i-Tree Eco is the core tool in assessing forest structure, services, and values. However, other i-Tree tools also assess this information, using data derived from i-Tree Eco procedures (Table 8). The following sections will provide a brief overview of how the tools assess forest structure and services.

I-TREE CANOPY
i-Tree Canopy uses Google aerial images of the study area to produce statistical estimates of tree and other land cover types, such as grass, structures, and impervious surfaces. Random points are placed in the defined area of interest and the user identifies the land cover class at the point center. Cover classes are defined by the user. Statistical estimates of area in each cover class (as a percent) are calculated as:

% = n / N
Where n = number of point hitting the cover class, and N = total number of points analyzed among all cover classes.
The standard error (SE) of the estimate is calculated as:
Percentage tree cover is multiplied by the area analyzed to determine the total tree cover area. If the number of points classified in a category (n) is less than 10, a different SE formula (Poisson) is used as the normal approximation cannot be relied upon with a small sample size (<10) (Hodges and Lehmann 1964). In this case:

SE = (√n) / N
More detailed information and examples are given in Nowak (2011).
Estimates of air pollution removal, carbon sequestration, and hydrologic impacts are based on the area of tree cover. A local standardized removal rate (e.g., kg/m 2 of tree cover) is multiplied by local tree cover (m 2 ) from i-Tree Canopy to estimate a total effect from the trees (kg). For air pollution removal, i-Tree Eco used air pollution and weather data to estimate average pollution removal effect per unit of tree cover (g/m 2 or $/m 2 of tree cover) for each U.S. county. Detailed methods regarding this process are given in Hirabayashi and Nowak (2016) and .
Estimates of carbon storage and annual sequestration are based on national and state data and methods from Nowak et al. (2013a). National standardized value of carbon storage per area of tree cover (7.69 kg C/m 2 ) are applied to the tree cover amount to estimate carbon storage. State-specific values (kg C/m 2 tree cover/year) are used to estimate carbon sequestration. These values vary based on length of growing season (Nowak et al. 2013a).
Estimates of avoided runoff, transpiration, and rainfall interception are based on the area of tree cover and local weather data. A local standardized removal rate (e.g., m 3 water/m 2 of tree cover) is multiplied by local tree cover (m 2 ) from i-Tree Canopy to estimate a local total effect from the trees (m 3 ). For hydrologic estimates, i-Tree Eco used weather data to estimate average effects per unit of tree cover (m 3 /m 2 or $/m 2 of tree cover) for each U.S. county. Detailed methods regarding this process are given in Hirabayashi (2015).

Planned Improvements
Update standardized carbon and pollution values.

I-TREE COOL RIVER
i-Tree Cool River is a mechanistic model that simulates the effects of riparian shading and impervious runoff on river temperatures. This model code is available under the research suite and model methods are detailed in Abdi and Endreny 2019 and Abdi et al. 2020.

Planned Improvements
None.

I-TREE COUNTY
i-Tree County is a tool based on data and methods in i-Tree Landscape. This tool allows users to quickly estimate ecosystem services and values from trees in an entire U.S. county or for a smaller area based on inputs of the area of tree cover.

Planned Improvements
Update values as i-Tree Landscape values are updated.

I-TREE DESIGN
i-Tree Design links to Google maps and allows a user to sketch the outline of their home and estimate how the trees around their home affect energy use and savings, and other environmental services of carbon sequestration, air pollution removal, and rainfall interception. Effects are projected over time based on methods from i-Tree Forecast, which projects tree size through time.
Building energy conservation effects and carbon storage/sequestration are based on the methods detailed for i-Tree Eco. Air pollution removal, rainfall interception, and avoided stormwater runoff are calculated based on methods detailed in i-Tree Canopy using the tree's canopy area. In addition, pollution removal also incorporates tree species information (deciduous vs. evergreen) and tree leaf area index calculations to improve the estimates. These variable removal rates by leaf area were determined by running the i-Tree Eco model for every U.S. county using LAI values ranging from 0 to 18, in 0.5 increments. Storm water abatement values ($/ gallon) are calculated from the i-Tree Streets Reference City Community Tree Guides, a collection of documents, primarily published by the Forest Service, available at http://www.itreetools.org/resources/archives.php.

Planned Improvements
Add information related to invasiveness of species and insect and disease risks.
Link to i-Tree Species and i-Tree Landscape to provide management guidance related to tree species selections and locations.

I-TREE FORECAST
i-Tree Forecast simulates future forest structure (e.g., number of trees and sizes) and various ecosystem services based on annual projections from the current forest structure data obtained from i-Tree Eco. There are three main components of i-Tree Forecast: Tree growth: projects annual growth for tree diameter, crown size, and leaf area for each tree.
Tree mortality: projects annual mortality based on user-defined mortality rates.
Tree establishment: projects annual new tree populations as input by users; can be used to illustrate the effect of the new trees or determine how many new trees need to be added annually to sustain a certain level of tree cover or benefits.

Tree Growth
Annual tree diameter growth is estimated for the study area based methods detailed in the carbon storage and sequestration section. i-Tree Forecast estimates tree height, crown width, crown height, and leaf area based on tree diameter for each future year. Tree height, crown height, and crown width are calculated using species, genus, order, and family specific equations that were derived from measurements from urban tree data (appendix 13). If there is no equation for a particular species, then the genus equation is used, followed by the family and order equations, if necessary. If no order equation is available, we use an equation for all trees to estimate these parameters. Projected leaf area is calculated from the crown height, tree height, and crown width estimates based on i-Tree methods (see Leaf Area and Leaf Area Index).
Projected canopy cover is calculated by summing the two-dimensional crown area of each tree in the population.

Tree Mortality Rate
Canopy dieback is the first indicator for tree mortality. Trees with 50 to 75 percent crown dieback have an annual mortality rate of 13.1 percent, trees with 76 to 99 percent dieback have a 50 percent annual mortality rate, and trees with 100 percent dieback have a 100 percent annual mortality rate (Nowak 1986). Trees with less than 50 percent dieback have a user-defined mortality rate based on the tree size class and diameter.
Trees are assigned to species-size classes: small trees have an average height at maturity of less than or equal to 12. . Each size class has a unique set of seven diameter ranges to which base mortality rates are assigned based on measured tree mortality by diameter class (Fig. 14) (Nowak 1986). The same distribution of mortality by diameter class is used for all tree-size classes, but the diameter range of the classes differed by size class. The actual mortality rate for each diameter class is adjusted so that the overall average mortality rate for the base population equaled the mortality rates assigned by the user. That is, the relative curve of mortality stays the same among diameter classes, but the actual values change based on the user-defined overall average rate. If desired, users can set custom mortality rates for individual genera, condition classes, or strata.

A B
Forecasting i-Tree Forecast classifies the population into tree species and diameter cohorts (e.g., all 5-cm red maples are one cohort). Based on estimated growth rates for the species and study area, the computer model simulates annual tree population totals by "growing" and "killing off " part of each cohort based on the user-defined mortality rates. The original stem diameter of newly established trees is input by the user. New cohorts are added based on user-defined annual planting rates. In simulating the annual addition of new trees to the model, the species composition of new trees is assumed to be proportional to the current species composition. New trees are assigned a crown light exposure (CLE) of 5 for leaf area calculations and the average CLE from the existing population for growth rate calculations. Total population numbers and characteristics (e.g., d.b.h., leaf area, tree cover) are determined for each year of the projection by summing the values from the individual cohorts.
Projected ecosystem services are estimated for carbon storage and air pollution removal. Carbon storage is based on the carbon equations and processes from i-Tree Eco. Pollution removal is based on county removal rates (g/m 2 of tree cover) from i-Tree Canopy, but with variable pollution removal rates based on the LAI. These variable rates are determined by running the i-Tree Eco model for every U.S. county using LAI values ranging from 0 to 18, in 0.5 increments. The projected LAI for each year is calculated as:

LAI = LA p / TC p
Where LAI = leaf area index per unit tree cover, LA p = projected leaf area (m 2 ) for the population, and TC p = projected tree cover (m 2 ) for the population study area.
The annual LAI values are then used to estimate the annual pollution removal by multiplying the appropriate removal rate based on LAI (g/m 2 tree cover) by area (m 2 ) of projected tree cover.

Extreme Events
Users can choose to simulate potential changes due to insect or disease outbreaks or storm events. Users can select an individual insect or disease, what year the outbreak will occur, how long the outbreak will last and what mortality would be expected from the outbreak. Based on the selections, only species susceptible to the outbreak would be killed at the assigned rate. Also, susceptible species would not be planted in the projections, unless the option to plant these species is selected by the user. For storm events, the user can select the storm type (e.g., class 1 hurricane), year of the event, and whether to user default or custom mortality rates for the event. Based on the selection, all trees would be killed during the event based on the assigned mortality rate.

Planned Improvements
Update mortality rates based on field data from monitoring of plot data.
Develop a tree establishment rate to sustain desired level of tree cover.
Update crown width equations with new equations from Westfall et al. (2020).

I-TREE HARVEST
i-Tree Harvest allows land managers to estimate the amount of carbon stored in harvested wood products; the estimating method was originally known as the PRESTO Wood Calculator. This tool calculates carbon values based on methods detailed in Smith et al. (2006) and does not use i-Tree methods.

Planned Improvements
None.

I-TREE HYDRO
i-Tree Hydro is designed to assess hourly changes in water quantity and quality due to changes in tree and other land cover types within a watershed or nonwatershed area. i-Tree Hydro calculates hourly interception, evapotranspiration, runoff, and other hydrologic values based on a semi-distributed, mechanistic rainfall-runoff computer model. Interception is simulated using an improved Rutter methodology (Valente et al. 1997) and evapotranspiration is simulated using improved Penman-Monteith equations (e.g., Shuttleworth 1993). More detailed methods for those and other parts of the model water budget are given in Wang et al. (2008), Yang et al. (2011) and . Hydrological parameters of the model can be calibrated to produce the best fit between predicted and observed streamflow. Water quality impacts are also simulated based on national and localized event mean concentration (EMC) pollutant coefficients.
Hourly weather data are derived from the weather station located closest to the study area and are preprocessed as described in Hirabayashi and Endreny (2016). Tree, grass, and impervious land cover percentages for a watershed are derived from i-Tree Canopy, i-Tree Eco, satellite data, or other means. The model can be calibrated for watershed areas using hourly stream flow data with model-independent parameter estimation analysis calibration routines (Wang et al. 2008). Calibration coefficients are calculated for peak flow, base flow, and balance flow (peak and base). A coefficient of +1 indicates a perfect fit to the model, 0 indicates the model predicts the same as the mean value, and a negative indicates that using the mean is a better predictor than the model (Moriasi et al. 2007). Differences between measured and estimated flow can be substantial due to mismatching of stream flow and weather data, as the weather stations are often a distance away from the watershed area. For example, it may be raining at the weather station and not in the watershed or vice versa. Default parameters are provided as a starting point for calibration or for use in projects where calibration is not desired or possible.
Leaf area index (LAI) is estimated as 4.7 for trees, 2.2 for shrubs, and 1.6 for herbaceous cover. The percentage impervious cover connected to the stream (known as directly connected impervious area or DCIA) varies with percentage impervious cover estimated for land cover determinations, with percentage connected increasing as percentage impervious cover increases (Sutherland 2000). The percentage of directly connected impervious cover represents the portion of impervious cover that drains directly to the modeled stream or any of its tributaries. The phrase "drains directly" describes a situation where precipitation that falls on a portion of the watershed's impervious cover is conveyed, overland or through a storm sewer network, directly into the stream or its tributaries.

Water Quality Effects
Event mean concentration (EMC) data are used for estimating pollutant loading into watersheds. The EMC is a statistical parameter representing the flow-proportional average concentration of a given parameter during a storm event and is defined as the total constituent mass divided by the total runoff volume (M/V). Estimates of EMC are usually obtained from a flow-weighted composite of concentration samples taken during a storm.
Mathematically (Charbeneau andBarretti 1998, Sansalone andBuchberger 1997): Where and are the time-variable concentration and flow measured during the runoff event, and M and V are pollutant mass (typically in milligrams) and runoff volume (typically in liters).
EMC results are from a flow-weighted average, not simply a time average of the concentration. Data from EMCs are used for estimating pollutant loading into watersheds.
The pollution load calculation from the EMC method is:

L = EMC x Q = EMC x d r x A
Where L is the pollutant load in mass per unit time (such as mg/h), EMC is event mean concentration (such as mg/l, or mg/m 3 , etc.), Q is surface runoff per time period associated with EMC (such as l/h, m 3 /day, etc.), d r is runoff depth per time period associated with EMC and study area (such as mm/h, m/h, m/day), and A is the study area (m 2 ), which is the catchment area in i-Tree Hydro.
The EMC is multiplied by the runoff volume to estimate the pollution load to the receiving water. Under most circumstances, the EMC provides the most useful means for quantifying the level of pollution resulting from a runoff event (U.S. EPA 2002).
Data from the three sources are used to compute estimates of EMC means and medians for 10 pollutants (Table 9): Nationwide Urban Runoff Program (U.S. EPA 1983), U.S. Geological Survey [USGS] urban storm water runoff database (Driver et al. 1985), and National Pollutant Discharge Elimination System (Smullen et al. 1999). These estimates are based on nationwide data and do not account for regional variation in soil types, climate, and other factors.
The new i-Tree Hydro program in the research suite (i-Tree Hydro+) also estimates air temperature changes due to trees and impervious surfaces based on methods detailed in .

Planned Improvements
Add mean and median pollutant coefficients specific to each hydrologic unit code (HUC) 8 13 and National Land Cover Database (NLCD) class (Homer et al. 2015) for sediment, total nitrogen, and total phosphorus (White et al. 2015).
Include green infrastructure impacts.
Add spatially-explicit land cover inputs and processing.
Add analyses of nutrient "hot spots" of nitrogen and phosphorus to identify areas to mitigate these chemicals (Stephan and Endreny 2016).

I-TREE LANDSCAPE
i-Tree Landscape integrates national landscape and environmental data to aid in forest management and planning. Users can explore tree and impervious cover, land cover, and basic demographic information. Users can also assess the benefits and values of trees (carbon storage, air pollution removal, reduced storm water runoff) in their area; understand local risks to people and forests (insects and diseases, wildfire potential, wildland-urban interface, forest basal area and composition, ultraviolet radiation exposure, air pollution exposure and areas of poor air quality, surface temperatures, projected climate change, brownfields, hardiness zones, flood and riparian zones, projected urban development, impaired waterways, tree species shifts due to climate change); and map areas to prioritize tree planting or protection efforts that improve forest and human health. i-Tree Landscape provides the details on the sources for the various forest, population, and risk data.
The program displays maps and associated data, as well as allows for prioritization among data sets to select the best areas for management among user-selected management units across the country. Management units include: State County 13 HUC 8 refers to maps at the sub-basin level, analogous to medium-sized river basins. There are about 2200 nationwide. Based on 2011 tree and impervious cover data from the National Land Cover Database (NLCD) (U.S. Geological Survey 2014) or locally supplied high resolution tree and impervious cover maps (https://landscape.itreetools.org/hires), along with other local data, the following ecosystem services for trees or environmental data in each area of analysis are assessed:

Air Pollution Removal
Air pollution removal and value estimates are based on procedures detailed in . This process uses local tree cover, leaf area index, percentage evergreen trees, weather, pollution, and population data to estimate pollution removal (g/m 2 tree cover) and values ($/m 2 tree cover) in urban and rural areas for each county. These values are applied to the area of tree cover to determine estimated removal and associated values of carbon monoxide (CO), nitrogen dioxide (NO 2 ), ozone (O 3 ), particulate matter less than 2.5 microns (PM 2.5 ), particulate matter between 2.5 and 10 microns (PM 10 ), and sulfur dioxide (SO 2 ). Value estimates are based on local health impacts estimated using U.S. EPA BenMAP (U.S. EPA 2012) model for each U.S. county for all pollutants except for CO and PM 10 , which use externality values ($/t) to estimate pollutant removal value. Estimated health impacts from changes in concentrations of NO 2 , O 3 , PM 2.5 and SO 2 due to trees are also provided based on BenMAP methodology (U.S. EPA 2012).
Estimates of pollution removal vary by county. 15 Average county removal rates are used, but have a maximum and minimum value. Average differences from the mean vary from a low of 30 percent for NO 2 to a high of 106 percent for PM 2.5 . The maximum and minimum values are likely unreasonable values as they assume a maximum or minimum removal rate for every hour of the year. No maximum or minimum values are estimated for CO.

Carbon Storage and Annual Sequestration
These values are calculated from two separate sources depending upon location (NLCD land cover class).
Nonforest carbon: For nonforest NLCD classes, total carbon storage and net annual sequestration are estimated using values estimated from urban forests (Nowak et al. 2013a). Net annual sequestration is an estimate of carbon accumulation from tree growth minus estimated carbon lost through decomposition due to tree mortality. Carbon storage is estimated based on the national average storage value of 7.69 kg C/m 2 tree cover (standard error [SE] = 1.36 kg C/m 2 ). Net sequestration is based on state-specific estimates that vary based on length of growing season and average = 0.226 kg C/m 2 tree cover/yr [SE = 0.045 kg C /m 2 tree cover/yr]). State-specific values vary from 0.430 kg C/ m 2 tree cover/yr (Hawaii) to 0.135 kg C/m 2 tree cover/yr (Wyoming) (Nowak et al. 2013a).
These values are applied to the tree cover estimates (m 2 ) from the tree cover map to estimate total carbon (kg).
Forest carbon: For forested regions, total carbon storage and net annual sequestration are derived from USDA Forest Service Forest Inventory and Analysis (FIA) data for each U.S. county. Net annual sequestration is calculated as carbon accumulated annually between FIA remeasurements based on accumulation from tree growth and new trees, minus carbon lost through tree mortality. The estimation of sequestration in forests is based on field measurements of change including the addition of new trees and loss of existing trees. In nonforest areas, net sequestration is estimated from computer models and is based on tree growth of existing trees and estimated mortality based on tree condition over a 1-year period. The nonforest sequestration estimate does not include the addition of new trees and includes only a partial loss of carbon from mortality due to decomposition (entire carbon from trees is not removed, only part of carbon lost to decomposition is removed).
Total carbon storage and net carbon sequestration per hectare of land is converted to total carbon storage and net sequestration per hectare of tree cover by dividing the carbon per hectare by percentage tree cover of the forested land in the county. Tree cover estimates from NLCD forest classes are used. In U.S. counties where tree cover is less than 10 percent (19 counties), tree cover is set to 10 percent to avoid inflating carbon density values. If no FIA carbon storage data exists for a county, but the county has tree cover within NLCD forest land, carbon storage density from the closest county is used. FIA carbon storage densities average 6.3 kg C/m 2 ; carbon storage density adjusted for tree cover equals 9.8 kg C/m 2 tree cover. The average SE associated with these estimates is 1.3 kg C/m 2 tree cover.
As NLCD forest land estimates (197 million hectares) are less than FIA forest land (264 million hectares), all estimates derived for forest land in i-Tree Landscape will be lower than FIA estimates. For example, FIA-defined forest land stores 16.7 billion metric tons of carbon in the conterminous United States but estimates from NLCD-defined forest land shows 13.3 billion metric tons of carbon. FIA forest land and carbon estimates are about 25 percent more than those derived using NLCD data due to differences in definition and classification of forest land between FIA and NLCD. i-Tree Landscape uses NLCD for the classification of forest area.
Net sequestration per area of tree cover is calculated in the same manner as for carbon storage. For net carbon sequestration, values for some counties are missing. If a county had a missing value, sequestration density values (kg C/m 2 tree cover/yr) from nearby counties in the same state are used. If the entire state had missing values, the county sequestration value is estimated based on converting the national FIA sequestration density value from all known counties to state values based on the ratio of state sequestration densities to national sequestration density for nonforest areas: Forest sequestration density for state = national average forest sequestration density x (state nonforest sequestration density / national average nonforest sequestration density).
This procedure was used for net forest sequestration in many western states (Arizona, California, Idaho, Montana, New Mexico, Nevada, Oregon, Utah, Washington, Wyoming).
The average net sequestration value for forests is 0.14 kg C/m 2 tree cover/yr (average SE = 0.10 kg C/m 2 tree cover/yr) (https://www.itreetools.org/landscape/resources/Carbon_ storage_and_seq_by_county_FIA.xlsx). This value is about 60 percent of the nonforest sequestration value. This difference is likely due to increased growth rates in urban areas (due to more open-grown nature of trees) and differences in means of calculating net sequestration (forest estimates remove all carbon from trees that die, but in urban estimates only a small portion are removed).
The value of carbon storage and sequestration is estimated at $188/metric ton of carbon. This value is derived from the Interagency Working Group (2016).

Hydrologic Effects
Estimates of evaporation, transpiration, precipitation interception, and avoided runoff for each county in the conterminous United States in 2010 were developed using i-Tree Eco, local leaf area indices, and weather data. Methods are detailed in Hirabayashi (2016) and Hirabayashi and Endreny (2016). The standard errors on these estimates are unknown.
Input data are estimates from a primary source (e.g., tree and impervious cover estimates come directly from NLCD). Ecosystem services and value estimates are derived as secondary estimates from the input data and other sources. The primary data layer used is tree cover, thus limitations in this layer will affect ecosystem service estimates (Table 10).
In addition to ecosystem services, other data were processed to produce environmental layers.  (2016) for several pollutants were obtained from the U.S. EPA Green Book (2017b).

Bird Species Richness
The estimated number of bird species were derived from BirdLife International (2020).
The following modifications to this dataset were conducted by Curt Flathers, USDA Forest Service: The original eastern towhee (Pipilo erythrophthalmus) showed the ranges for both the eastern and spotted towhee (Pipilo maculatus); two range maps were created to reflect each species.
The original least flycatcher (Empidonax minimus) did not show the breeding range extending down the northeast United States along the Appalachian Mountains; ranges were modified to correct this omission.

Brownfields
A brownfield is a property with the presence or potential presence of a hazardous substance, pollutant, or contaminant. Locations of brownfields were obtained from the U.S. EPA (2018c).

Climate Change Data
Projected climate data were obtained from the National Center for Atmospheric Research (NCAR) Community Climate System Model (CCSM) projections (National Center for Atmospheric Research 2016). These data were obtained at a 4.5 km resolution that covered the conterminous United States, but were resampled for display/analysis at 30 m resolution and projected to WGS 1984 Web Mercator Auxiliary Sphere.
Data represented an "ensemble average" of six model runs for representative concentration pathways (RCP) 4.5 and 8.5 for decades starting in 2010 and ending in 2100 (i.e., 2010, 2020, 2030, ... 2100). These data include both the projected values and the differences between current (2010) and future decadal modeled values for annual mean air temperature, annual total precipitation, and mean January and July temperatures.

Drought
Drought information is based on the Cumulative Drought Severity Index (1987-2013) (Peters et al. 2014

Endangered Species
Number of threatened, endangered, and species of concern for both plants and animals were derived from the U.S. Fish and Wildlife Service (2018).

Flood and Riparian Zones
Riparian zones (i.e., water areas plus the land area within 108 m of surface water edges, 2018) and 100-year floodplain (2016) areas were obtained from EnviroAtlas (U.S. EPA 2018b).

Forest Basal Area and Composition
Forest basal area data by tree species (2012) were obtained from the USDA Forest Service's Individual Tree Species Atlas parameter maps (Ellenwood et al. 2015).

Forest Type
National forest type data were obtained from the USDA Forest Service FSGeodata clearinghouse (USDA Forest Service 2004).

Future Plant Hardiness and Heat Zones
Future plant hardiness and heat zones (Matthews et al. 2018(Matthews et al. , 2019 are given for three time periods (2010-2039, 2040-2069, 2070-2099) for two model scenarios (CCSM4 under RC4.5 concentration pathway, GFDL CM3 under RC8.5 concentration pathway). Future heat was quantified as the number of days per year with a maximum daily temperature >30 °C (86 °F). These heat days were also calculated for 1980-2009 and the differences between these base years and the three sets of future years are reported.

Hardiness Zones
The 2012 USDA Plant Hardiness Zone Map is based on the average annual minimum winter temperature, divided into 10-degree Fahrenheit zones (Daly et al. 1990).

Human Populations
Information on human population and housing statistics were derived from the U.S. Census Bureau (2010b).

Impaired Waterways
Impaired waters are waters that are too polluted or otherwise degraded below state water quality standards. Data on impaired waterways were obtained from U.S. EPA (2018a).

Insects and Diseases
Insect and disease range maps from 2017 were obtained from the Forest Health Technology Enterprise Team (USDA Forest Service 2017b).

Land Surface Temperature
Land surface temperatures were estimated for the United States as described below.
Landsat scene selection. Landsat 8 scenes were downloaded from the U.S. Geologic Survey (USGS) (U.S. Geological Survey 2013) for the United States and U.S. territories that were as cloud-free as possible (June, July, August) for the years 2013, 2014, and 2015. Scenes were selected to provide total U.S. coverage for each year. As scenes overlapped, many areas had two or more scenes covering a location.
Converting scene data to surface temperatures. Landsat 8 thermal infrared sensor band values were converted to land surface temperature by: Converting satellite data to at-satellite brightness temperature based on equations detailed in U.S. Geological Survey (2017).
Calculating emissivity values from normalized difference vegetation index (NDVI) using the NDVI thresholds method (Sobrino et al. 2001).
Converting at-satellite brightness temperature and emissivity information for each 30 m pixel to Land Surface temperature based on equations given in section 3.4 in Weng et al. (2004).
Cloud and snow removal. To remove the existing clouds from the images, Fmask (Function of mask) software was used (https://github.com/prs021/fmask). Fmask masks clouds, cloud shadows, and snow for Landsat TM/ETM+ images. Pixels where clouds were removed were converted to "no data. " Land surface temperatures and temperature differences. Land surface temperatures were estimated for the United States based on Landsat 8 data and standard procedures from the literature. As Landsat scenes had various dates across the United States, each Landsat scene was processed independently to minimize the effect of different scene dates on surface temperatures. To standardize the land surface temperature estimates, the Landsat scene's median land surface temperature was subtracted for each pixel land surface temperature value to create a relative surface temperature difference estimate. With this approach, about half of scene pixels will have above-median temperatures and half will have below-median temperatures, but the amount they differ from the mean will vary based on local land surface temperatures.
Within each year's data (e.g., 2015) all scenes were mosaicked and where scenes overlapped, an average of surface temperature differences among the multiple scenes was used to calculate the surface temperature difference at each pixel with multiple values.
To help fill in for "no data" pixels due to clouds or snow, data from the various years (2013)(2014)(2015) were used. The first priority was given to data from 2015. If "no data" pixels existed in 2015, data from 2014 were used to fill in these missing pixels. If "no data" pixels existed in 2015 and 2014, data from 2013 were used to fill in these missing pixels. If "no data" pixels existed in 2015. 2014 and 2013, these pixels were labeled as "no data. " Land surface temperature difference are projected at 90 meter resolution in i-Tree Landscape.

Projected Urban Development
Projected urban development was mapped based on the Integrated Climate and Land-Use Scenarios version 2.1 of the Fourth National Climate Assessment (U.S. EPA 2018d). These data were modified and generalized from 18 original land use codes to five generalized land uses as follows: Water: natural waters, reservoirs and canals.
Urban: urban low density, urban high density, commercial, transportation.
Projected changes from 2010 to 2060 and 2100 were mapped for the following class changes to illustrate and calculate areas that are projected to be developed: Nonwooded to exurban Nonwooded to urban Wooded to exurban Wooded to urban

Sea Level Rise
The area submerged by sea level rise (NOAA 2017) of 1 to 6 feet, at 1-foot increments, are provided.

Storm Surge
Storm surge data (Zachry et al. 2015) for average depth of storm surge and average area submerged by the storm surge are provided, by hurricane category (1-5).

Street Walkability
The U.S. EPA's Walkability Index (U.S. EPA 2015) characterizes every Census 2010 block group in the United States based on its relative walkability. Walkability depends upon characteristics of the built environment that influence the likelihood of walking being used as a mode of travel. The index ranges between 1 and 20 and were categorized as: 1-5.0 (least walkable) 5.1-10.0 (below average walkable) 10.1-15.0 (above average walkable) 15.1-20.0 (most walkable)

Tree Species Shifts Due to Climate Change
To illustrate and calculate projected habitat changes for 314 tree species, we use data from the ForeCASTS Project (North Carolina State University 2018). Current species range maps were contrasted with projected species ranges in 2050 and 2100 based on the Hadley, Scenario A1 climate projection model. Each ~4 km pixel was classified based on the probability of presence for each tree species in each year. The differences in probability between the years was standardized to values between -100 (the species had 100 percent probability in the current year, but 0 probability in the future) and 100 (the species had 0 percent probability in the current year, but 100 probability in the future).

Ultraviolet Radiation Exposure
The average and maximum UV index at local solar noon for all days between 2008 and 2013 were obtained from the Tropospheric Emission Monitoring Internet Service (TEMIS 2016). The UV index is an estimation of the UV levels that are important for the effects on the human skin, where 1 unit equals 25 mW/m 2 . The index ranges from 0 to greater than 11: 0-2: Low UV exposure (damage to skin in >60 minutes). 3-5: Moderate UV exposure (damage to skin in 45 minutes). 6-7: High UV exposure (damage to skin in 30 minutes). 8-10: Very high UV exposure (damage to skin in 15 minutes).

Water Quality Index
A water quality index (Brown and Froemke 2012) is reported that expresses the overall water quality (0-5; 0 = no risk, 5 = highest risk) at a certain location and time based on several water quality parameters: Water clarity

Dissolved oxygen
Oxygen demand Nutrients Bacteria

Wildfire Potential
Wildfire hazard potential data were obtained from the USDA Forest Service (2014) Fire, Fuel, Smoke Science Program and are designed to depict the relative potential for wildfire that would be difficult for suppression resources to contain.

Wildland-Urban Interface
The 1990 and 2010 wildland-urban interface (WUI) data were obtained from the University of Wisconsin-Madison Silvis lab (Martinuzzi et al. 2015, Radeloff et al. 2017. Two types of WUI data for 1990 and 2010 are displayed: percent intermix and interface (along with percent non-interface area). Intermix WUI are areas where housing and vegetation intermingle; interface WUI are areas with housing in the vicinity of contiguous wildland vegetation.
Based on all of these local environmental data and estimates of ecosystem services and values, users can weight these layers to determine area of prioritization for management actions.

Prioritization
To determine the best locations to plant or protect trees, tree and impervious cover data are used in conjunction with U.S. Census data and various layers (e.g., areas with highest population density, lowest tree density, or lowest tree cover per capita). i-Tree Landscape creates an index that prioritizes areas among the selected geographic units. The higher the index value, the higher the priority of the area for tree planting or protection. The index is developed with user-determined importance values, or "weights", assigned to up to three layers. Each layer is weighted between 0 to 100, such that the sum of the layers must equal 100.
As geographic areas differ in size, all index inputs are either in percentages or standardized per unit area or person. Each layer is standardized on a scale of 0 to 1 with 1 representing the geographic area with the highest priority (e.g., areas with highest population density, lowest tree density, or lowest tree cover per capita).
Although users can select their own weights and prioritization schemes, three common scenarios are given: Population: (default) an index weighted toward areas of relatively high population density, low tree cover per capita, and high available planting space.
Minorities: an index weighted toward areas of relatively high minority population density, low tree cover per capita, and high available planting space.
Poverty: an index weighted toward areas of relatively high proportion of population below the poverty line, low tree cover per capita, and high available planting space.
Information on methods and where the numerous layers of data (e.g., census data, wildfire data) were obtained is available at the i-Tree Landscape reference web page (i-Tree Team, n.d.).

Planned Future Improvements
Add links to local FIA and i-Tree field data.
Add more high resolution cover maps.

MYTREE
MyTree is a mobile smart phone application that allows users to easily quantify the benefits and values of individual trees. MyTree calculations are based on the methods in i-Tree Design.

Planned Improvements
Create ability to pin trees to a global map.
Create option to calculate past and future (total life-span) benefits of trees.
Add optional tree health variables and linking these data with the Nature Conservancy's Healthy Trees Healthy Cities Initiative (Nature Conservancy 2019).

I-TREE PLANTING
i-Tree Planting is a tool to estimate the long-term environmental benefits from a tree planting project of numerous trees and species. Its methods are based on methods from i-Tree Design and the i-Tree Forecast tool.

Planned Improvements
None.

I-TREE PROJECTS
i-Tree Projects allows users to visualize plot and community data, as well as access and download field data and results for further analysis. The beta version of i-Tree Projects launches with only one i-Tree Eco project from Baltimore, MD, collected as part of the Baltimore Ecosystem Study (https://lternet.edu/site/baltimore-ecosystem-study/). Once more projects are added, users will be able to compare data among communities.

Planned Improvements
Add more cities.
Add ability to compare city results.
Add map of known project locations.

I-TREE SPECIES
i-Tree Species is designed to help users select the most appropriate tree species based on the species potential environmental services and geographic area. Users select and rank the importance (0-10) of each environmental service desired from trees. The program then calculates the best tree species based on the user-provided weighting of environmental benefits of tree species at maturity.
Species are selected based on three types of information: Hardiness zone-as determined by state and city.
Mature height-user-specified minimum and maximum heights. i-Tree Species methods are detailed in Nowak (2008).

Planned Improvements
Update program with more common species.
Estimate species effects over the estimated life span of the species.
Estimate benefits estimates for conditions from numerous U.S. cities.
Add local species limitations to help limit species list (e.g., soil tolerances).

I-TREE STREETS
i-Tree Streets is a tool designed to assess the benefits and values of street tree populations. This tool uses data from 16 reference cities to interpolate values for other U.S. cities. Due to this reference city approach, which is different from other i-Tree tools, i-Tree Streets' outputs have been integrated within i-Tree Eco and street trees are now analyzed using i-Tree Eco. The Streets tool is no longer supported, but is available as a legacy tool.
The primary reason for integrating Streets within Eco is that the Streets program is limited to 16 cities and all subsequent results are based on results from one of the 16 cities. Thus, local city environmental data are not being used, but rather data from the selected reference city. Conversely, i-Tree Eco results are based on local environmental data. For example, if you are analyzing tree data from San Antonio, TX, your reference environmental data in Streets would be coming from Charlotte, NC; using Eco, your environmental data would be from San Antonio. This integration allows for most of the Streets capabilities to be completed using Eco, but with better, more locally appropriate environmental data.

Planned Improvements
None. Street trees should now be assessed using i-Tree Eco.

I-TREE WOOD MARKETPLACE
i-Tree Marketplace connects urban wood harvests to users of removed trees for upcycling and waste reduction. Information about logs are captured in the field, uploaded to the Cloud, and displayed to potential users. Potential users can peruse these logs and contact their owners to arrange for their disposition. i-Tree Marketplace is designed to be a simple application accessed via most modern Web browsers. This program is currently a beta prototype.

Planned Improvements
Expand pilot test.
Track logs from removal to recycled use.
Set a means for potential monetary exchange.

ACKNOWLEDGMENTS
I thank the many people who have worked on i-Tree methods, code, and support (see appendix 1), the Davey Tree, and the numerous i-Tree partners for their significant support, financial and otherwise, that helped make i- Special thanks also go to the following groups that helped fund the development of various tools: American Forests-helped fund i-Tree Canopy.

California Department of Forestry and Fire Protection and California Urban
Forest Council-helped fund i-Tree Planting.
Northeastern Area Association of State Foresters-provided substantial funding and input in the development of the original version of Landscape.
USDA Climate Hubs-helped fund land surface temperatures in i-Tree Landscape.
A special thank you to all the people who have contributed to the success of i-Tree (appendix 1). Many people have been instrumental in securing funding for i-Tree and overall program support through the years. These people include: Michael Rains, Pat Covey, Karl Warnke, Jim Hubbard, Jan Davis, Lauren Marshall, Dennis May, former Forest Service Chief Tom Tidwell, Phil Rodbell, Robin Morgan, Peggy Castillo, Thomas Schmidt, Elizabeth Larry, Steve Koehn, Lon Yeary, Robert Bridges, Richard Birdsey, Dudley Hartel, and Eric Kuehler. A special thank you also to the many users who had provided advice and input to improve i-Tree.