Satellite-driven estimation of terrestrial carbon flux over Far East Asia with 1-km grid resolution
Research Highlights
► Terrestrial carbon fluxes with 1-km grid resolution are estimated by using BEAMS model and satellite observations. ► Not only GPP but also NEP, net radiation, and latent heat were validated at six flux-tower sites. ► Results of the model validations showed reasonable seasonal and annual patterns. ► 1-km grid carbon fluxes estimated by our approach represented reasonable spatial and temporal patterns in Far East Asia.
Introduction
It is difficult to estimate a detailed spatial variation of the land carbon sink. This is mainly because land cover is generally more heterogeneous than the sea surface and there is much uncertainty about land cover change emissions (Denman et al., 2007 in IPCC AR4, Chapter 7), leading to an incomplete understanding of terrestrial ecosystem processes. In fact, the Intergovernmental Panel on Climate Change (IPCC, 2007) showed that the uncertainty for the land carbon sink is much larger than that of the oceanic carbon sink. Some model inter-comparison projects have also reported that spatial and temporal variations of the land carbon sink contribute significantly to the differences among existing studies and future scenarios (e.g., Cramer et al., 2001, Cramer et al., 1999, Friedlingstein et al., 2006, Murakami et al., 2009).
It is important to estimate terrestrial carbon fluxes with detailed spatial resolution. There are complex and diverse local-scale phenomena and events that continually occur on the surface of the land (e.g., climate and meteorological changes, deforestation, and urbanization). These local-scale events have significant effects on climate change and global carbon dynamics (Pongratz, Raddatz, et al., 2009, Saigusa et al., 2010), and consequently have a major impact on the global carbon cycle. Furthermore, with increased human activity, the effect of local events on the global carbon cycle is likely to increase steadily with each passing year (Pongratz, Reick, et al., 2009, Strassmann et al., 2008).
Unfortunately, we currently have a poor understanding of such local-scale changes in terrestrial carbon sinks. Previous studies have indicated that local-scale land surface events are a major controllable factor in terrestrial carbon dynamics. For example, in the summer of 2003, a rainy season front hovered over the vicinity of Japan for slightly longer than usual (Japan Meteorological Agency, 2005). The extended time that the stationary rainy front lingered greatly limited measured gross primary production (GPP) and ecosystem respiration in some parts of Japan because the amount of sunlight reaching the surface and the air temperature both decreased, leading to a large decrease in annual land carbon uptake in the entire region (Saigusa et al., 2010, Saigusa et al., 2008). Another example is the fact that humans continue to selectively cut timber from the earth's forests. An existing study analyzing terrestrial carbon fluxes in the Amazon region suggests that recent Amazonian deforestation, specifically in moist tropical forest areas, may decrease net primary production (NPP) (Potter et al., 1998). Finally, IPCC's Fourth Assessment Report: Climate Change 2007 (AR4) (IPCC, 2007) predicts that differences in air temperature and precipitation among continents and latitudinal belts might be gradually increasing due to increased concentrations of atmospheric carbon dioxide. Local-scale climate changes affect global terrestrial carbon dynamics via all ecosystem activities, but we have not established an approach for making detailed, high-resolution assessments of the effect of local-scale climate and meteorological changes on carbon exchanges.
In order to discover the precise role played by the carbon cycle in local-scale land surface condition changes, we need to derive a detailed spatial variation of terrestrial carbon fluxes using a model with the highest spatial resolution possible. The higher the spatial resolution of the model, the more we can understand local-scale land surface characteristics. By making use of downscaling approaches (high-resolution simulation, nesting, and statistical methods) with meteorological model simulations, an existing model has been estimated with a grid resolution of approximately 5 to 20 km (IPCC, 2007). In concert with such high-resolution analysis of climate change, it is important to assess the major factors affecting carbon dynamics with as high a grid resolution as possible, including experiments to estimate the terrestrial carbon cycle at higher resolution. The spatial resolutions of such models are not yet high enough to allow detailed local analyses, but many studies have assessed the effect of local factors on global land carbon dynamics. For example, for global NPP, grid resolutions have ranged from about 250 m to 2.5° (e.g., Nemani et al., 2003, Quaife et al., 2008, Raich et al., 1991, Running et al., 2004, Zaks et al., 2007, Zhao et al., 2006). For net ecosystem production (NEP), the spatial grid resolutions have ranged from 0.5° to 2.5° (Cramer et al., 2001, Sitch et al., 2003). Some studies have produced fairly high resolutions using just model simulation (Ito, 2008), whereas others have used a combination of satellite data and models (Heinsch et al., 2003, Heinsch et al., 2006, Sasai et al., 2007). Heinsch et al., 2003, Heinsch et al., 2006 attempted estimates of global GPP and NPP with 1-km grid resolution by combining Moderate resolution Imaging Spectroradiometer (MODIS) data with a simple, original algorithm, but did not arrive at an estimate of NEP because remote sensing cannot directly observe below-ground information. We need to apply information gathered by remote sensing into a biosphere model to estimate observation-based NEP. Ito (2008) demonstrated spatial variations in NPP and NEP with a 1-km grid resolution over Far East Asia, excluding built-up areas, using a prognostic-type biosphere model that required only climate and land cover maps as forcing data.
Satellite observation is one of the most efficient approaches to realistic, high-resolution carbon flux estimation at several spatial scales for the following reasons: 1) Since satellite observation is based on measurements, we can deal with realistic land carbon dynamics. 2) Some earth observation satellite sensors make continuous periodic observations (e.g., NOAA/AVHRR, Terra/MODIS and Aqua/MODIS, and TRMM), giving us the ability to cope with an ever-changing land surface environment using continuous time-series observation data. 3) Recent satellite monitoring permits extensive, short duration observations with relatively high resolution, yielding satellite data that has sufficient spatial resolution to examine certain small-scale events. (e.g., the MODIS product can generate a spatial grid resolution in the range from 1 km to 250 m). Generally, since terrestrial carbon fluxes cannot be directly observed by satellite sensors, they are estimated via diagnostic-type models and remote-sensing algorithms requiring satellite data products as inputs. Some parameters used in the models and algorithms cannot be observed by satellite sensors, and so we cannot readily estimate terrestrial carbon fluxes from satellite information only. An auxiliary way to compensate for such insufficient information is to input climate, vegetation, and soil data into the model. Many existing models estimate carbon fluxes from satellite data such as those obtained by NOAA/AVHRR, Terra/MODIS, and Aqua/MODIS (e.g., Goetz et al., 2000, Heinsch et al., 2006, Nemani et al., 2003, Potter et al., 1993, Prince and Goward, 1995, Ruimy et al., 1994, Sasai et al., 2005). Hence, by combining as much time-series satellite data as possible into a model, we can estimate detailed temporal carbon fluxes with more realistic land surface conditions. This combination approach could also be applied to global analysis with possible higher spatial resolution.
In order to analyze the spatial variation of NEP at a high resolution on a global and regional scale, we attempt to develop a satellite-driven approach for estimating terrestrial carbon fluxes that include urbanization's deforestation effects on land carbon dynamics. This suggests a need to propose an applicable method for evaluating the effect of local-scale events on carbon exchange in any area and on any spatial scale. Therefore, we attempt to develop a satellite-driven approach with a 1-km grid resolution. The approach basically requires the MODIS Land Products provided by NASA's Earth Observation System (NASA-EOS), and as subsidiary data, the Tropical Rainfall Measuring Mission (TRMM) product, the Shuttle Radar Topography Mission (SRTM) product, and others. We used the biosphere model integrating eco-physiological and mechanistic approaches using satellite data (BEAMS), which we expected to provide an efficient approach to calculating terrestrial carbon exchanges from satellite observations in any region of the world (Sasai et al., 2005, Sasai et al., 2007).
Section snippets
Study area
The study area was the northern part of East Asia between 125° and 150° east, ranging between 30° and 50° north, including Japan, Korea, and parts of China and Russia (Fig. 1). The total area is approximately 4,882,000 km2 (3000 × 2400 pixels). Of this area, land covers about 1,592,000 km2, and areas of water, such as the ocean, rivers, and lakes, cover 3,290,000 km2. Geomorphic type varies widely in the study area, and there is relatively less flat land than other landforms. Japan, in particular,
Model
We employed the BEAMS model (Sasai et al., 2005, Sasai et al., 2007), which simulates carbon, energy, and water fluxes on the basis of satellite observations and climate data. The model consists of heat, water, and carbon submodels (Table 1). The heat submodel has four heat processes: net radiation and latent, sensible, and ground heat fluxes. Similarly, the water submodel has four water pools and eight water processes comprising canopy-intercepted precipitation and evaporation, ground surface
Model validation
We validated the BEAMS estimations with flux measurements at six tower sites (Table 2a, Table 2b). In most previous studies, only carbon fluxes and/or biophysical parameters (fPAR, LAI, and LUE) have been compared with ground measurements to validate existing models (e.g., Heinsch et al., 2006, Ito, 2008, Turner et al., 2006). However, not just carbon fluxes but also hydrological and energy fluxes need to be included in model validation because most biosphere models consist of carbon processes
Model experiments
Conducting a carbon cycle simulation with 1-km grid resolution and monthly time increments, we analyzed the spatial and temporal patterns in terrestrial carbon exchanges with the atmosphere. All carbon and water pools in the model were initialized assuming that the carbon and water cycles in 2001 were in equilibrium. To do this, an iterative calculation was carried out with first-year inputs (spin-up), and the model was operated until all carbon and water pools reached steady state conditions.
Input data
We pre-processed some satellite, climate, and soil datasets to use as model inputs in this analysis (Table 3). We used MODIS, TRMM, and SRTM satellite data. fPAR, LAI, LST, albedo, radiation, vapor pressure, and land cover map data were based on the MODIS high-level land products. The original MODIS fPAR, LAI, LST, and albedo data were converted into monthly mean model inputs using the following five image pre-processes:
- 1)
First, we formed a mosaic of the MODIS tiles and used Nearest Neighbor
Results and discussion
We estimated the spatial variations of annual mean GPP and NPP from 2001 to 2006 (Fig. 3). The results showed two typical spatial variation characteristics. First, the two fluxes decreased from south to north. Plant physiology and microbial activity in soil are limited by dryness and low temperature, so the land carbon turnover rate is small. The spatial variation is also strongly influenced by air temperature, and it has the same tendencies as Hirata et al.'s (2008) integration analysis of
Conclusion
We proposed a diagnostic approach for estimating terrestrial carbon fluxes with sufficient spatial resolution to analyze the relationships among local-scale events and terrestrial ecosystem activities. In order to produce our “observed” carbon fluxes, we were particularly diligent in two parts of the estimation process. First, we used as many satellite datasets as possible to minimize errors from model inputs. Second, the model was validated at as many sites as possible. These two efforts give
Acknowledgements
The corresponding author is deeply grateful to Dr. S. Togashi from the National Institute of Advanced Industrial Science and Technology. This study was supported in part by the Global Environment Research Fund from the Japanese Ministry of the Environment, Integrated Study for the Terrestrial Carbon Management of Asia in the 21st Century Based on Scientific Advancements, by a Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) Grant-in-Aid for Young Scientists (B)
References (70)
- et al.
A modeling and spatio-temporal analysis framework for monitoring environmental change using NPP as an ecosystem indicator
Remote Sensing of Environment
(2009) - et al.
MODIS Collection 5 global land cover: Algorithm refinements and characterization of new datasets
Remote Sensing of Environment
(2010) - et al.
Carbon stock estimates for sugi and hinoki forests in Japan
Forest Ecology and Management
(2003) - et al.
Terrestrial vegetation and water balance: Hydrological evaluation of a dynamic global vegetation model
Journal of Hydrology
(2004) - et al.
Satellite-based estimation of surface vapor pressure deficits using MODIS land surface temperature data
Remote Sensing of Environment
(2008) - et al.
Seasonal and interannual variations in carbon dioxide exchange of a temperate larch forest
Agricultural and Forest Meteorology
(2007) - et al.
Spatial distribution of carbon balance in forest ecosystems across East Asia
Agricultural and Forest Meteorology
(2008) The regional carbon budget of East Asia simulated with a terrestrial ecosystem model and validated using AsiaFlux data
Agricultural and Forest Meteorology
(2008)- et al.
Modelling the spatial pattern of net primary productivity in Chinese forests
Ecological Modelling
(1999) - et al.
Validating MODIS land surface reflectance and albedo products: Methods and preliminary results
Remote Sensing of Environment
(2002)
Energy consumption and evapotranspiration at several boreal and temperature forests in the Far East
Agricultural and Forest Meteorology
Integrating remotely sensed data with an ecosystem model to estimate net primary productivity in East Asia
Remote Sensing of Environment
Global products of vegetation leaf area and fraction absorbed PAR from year one of modis data
Remote Sensing of Environment
Monitoring and forecasting ecosystem dynamics using the Terrestrial Observation and Prediction System (TOPS)
Remote Sensing of Environment
Assimilating canopy reflectance data into an ecosystem model with an ensemble Kalman filter
Remote Sensing of Environment
Temporal and spatial variations in the seasonal patterns of CO2 flux in boreal, temperate, and tropical forests in East Asia
Agricultural and Forest Meteorology
Interannual variability of carbon budget components in an AsiaFlux forest site estimated by long-term flux measurements
Agricultural and Forest Meteorology
Comparing terrestrial carbon fluxes from the scale of a flux tower to the global scale
Ecological Modelling
SEIB-DGVM: A new dynamic global vegetation model using a spatially explicit individualbased approach
Ecological Modelling
Modeling and measuring the effects of disturbance history and climate on carbon and water budgets in evergreen needleleaf forests
Agricultural and Forest Meteorology
Evaluation of MODIS NPP and GPP products across multiple biomes
Remote Sensing of Environment
Validation of the land-surface temperature products retrieved from Moderate Resolution Imaging Spectroradiometer data
Remote Sensing of Environment
Developing a continental-scale measure of gross primary production by combining MODIS and AmeriFlux data through support vector machine approach
Remote Sensing of Environment
The IGBP-DIS global 1-km land-cover data set DISCover: A project overview
Photogrammetric Engineering and Remote Sensing
Description of the “TRIFFID” Dynamic Global Vegetation Model, Hadley Centre technical note 24
Global response of terrestrial ecosystem structure and function to CO2 and climate change: Results from six dynamic global vegetation models
Global Change Biology
Comparing global models of terrestrial net primary productivity (NPP): Overview and key results
Global Change Biology
The shuttle radar topography mission
Reviews of Geophysics
Climate–carbon cycle feedback analysis, results from the C4MIP model intercomparison
Journal of Climate
FLUXNET and modelling the global carbon cycle
Global Change Biology
Interannual variability of global terrestrial primary production: Results of a model driven with satellite observations
Journal of Geophysical Research
Quantifying and mapping the human appropriation of net primary production in earth's terrestrial ecosystems
Proceedings of the National Academy of Sciences
Users Guide: GPP and NPP (MOD17A2/A3) Products NASA MODIS Land Algorithm
Evaluation of remote sensing based terrestrial productivity from MODIS using regional Tower Eddy flux network observations
IEEE Transactions on Geoscience and Remote Sensing
Cited by (38)
Evaluation of forest carbon uptake in South Korea using the national flux tower network, remote sensing, and data-driven technology
2021, Agricultural and Forest MeteorologyCitation Excerpt :For these reasons, the number of observations is limited. Therefore, to assess terrestrial carbon cycling over regions, EC measurements can be upscaled to a large area in conjunction with data from a meteorological observation network, remote sensing, and modeling (Ichii et al., 2017; Jung et al., 2009; Maselli et al., 2006, 2008; Papale and Valentini, 2003; Sasai et al., 2011; Xiao et al., 2014; Yang et al., 2007). As the number of EC observations increases, the EC network covers various land covers, climate regions, and species (Baldocchi, 2008).
Improvement of satellite-based estimation of gross primary production through optimization of meteorological parameters and high resolution land cover information at regional scale over East Asia
2019, Agricultural and Forest MeteorologyCitation Excerpt :For the same area of about 1,268,477 km2 (i.e., 30 °N – 50 °N and 125 °E – 150 °E) used in Ito (2008), 1056.38, 1525.37, and 1363.84 (gC/m2/yr) of GPPs were calculated from MOD17A2H, optimized FROM-MCD, and BESS GPPs respectively, from 2002 to 2005. Sasai et al. (2011) reported an annual sum of GPP over six years from 2001 to 2006 in the same area as Ito (2008) of 1655.53 (gC/m2/yr) using the BEAMS process-based model. Similarly, 993.32, 1519.85, and 1292.89 were found from MOD17A2H, optimized FROM-MCD, and BESS GPPs for the same area and period, respectively.
Estimation of global soil respiration by accounting for land-use changes derived from remote sensing data
2017, Journal of Environmental ManagementSpatial representativeness and uncertainty of eddy covariance carbon flux measurements for upscaling net ecosystem productivity to the grid scale
2016, Agricultural and Forest MeteorologyEstimate of biogenic VOC emissions in Japan and their effects on photochemical formation of ambient ozone and secondary organic aerosol
2015, Atmospheric EnvironmentCitation Excerpt :The horizontal distributions of the twelve tree groups in 1 × 1 km meshes all over Japan are available in J-IBIS. The leaf masses of the twelve tree groups in prefectures were obtained from the record for forest resource available in each prefecture, and were allocated to 1 × 1 km meshes based on the monthly LAI data contained in the Moderate Resolution Imaging Spectroradiometer (MODIS) products (Myneni et al., 2002; Sasai et al., 2011). Then, the monthly gridded EFs of twelve tree groups were developed by multiplying the monthly basal emission rates, the gridded horizontal distributions, and the gridded leaf masses of each tree group.
Embodied carbon budget accounting system for calculating carbon footprint of large hydropower project
2015, Journal of Cleaner Production