Which are the factors influencing the integration of mitigation and adaptation in climate change plans in Latin American cities?

As cities are major contributors to GHG emissions and places where people face multiple climate change impacts, their critical role in responding to climate change is becoming increasingly evident. Cities are developing climate change action plans (CCAPs) focusing their efforts on reducing GHG emissions and adapting to climate change impacts. Despite having the highest urban population in the world, there are a few studies on urban CCAPs in Latin America and the Caribbean (LAC) region. This study assessed the level of integration of mitigation and adaptation (IMA) in urban climate change plans across 44 major LAC cities. The level of IMA was measured by the utilization of the IMA index, a comprehensive evaluation framework of indicators. The results showed that more than half of the examined LAC cities have a moderate level of IMA. The study further explored and statistically analyzed 42 institutional, socioeconomic and environmental factors to identify which ones potentially drive or constrain the level of IMA. Five out of 42 factors were found to have a significant impact (p-value < 0.05) on the IMA index. Of the five significant factors, memberships in regional networks FLACMA and UCCI respectively, and donor agencies’ contribution to the development of urban policies had a positive impact on IMA index; while the national climate fund and membership in the global network Urban LEDS had a negative impact. This suggests that cities are most likely to integrate mitigation and adaptation when the development of their CCAPs are supported by donor agencies or collaborating with other cities. The results highlight the important role of donor agencies, international organizations and cities’ networks on providing the necessary capacity to cities for addressing climate change in an integrated manner.


Introduction
Cities produce more than 70% of global anthropogenic GHG emissions and consume around 75% of total energy demand (IPCC 2014a, UN-Habitat 2016). Latin America and the Caribbean (LAC) is highly urbanized: 81% of the population lives in cities, when the global figure is approximately 55% (UN-DESA 2018). In addition, 60%-70% of LAC regional GDP accrues in urban centers (Baŕcena et al 2017). As the most unequal region in the world (ECLAC 2016), there might be more needs on intervention for adapting climate change in the region since poor people, particularly living in slums, are exposed and vulnerable to climate impacts (Reyer et al 2017). LAC cities have started developing local climate change action plans (CCAPs), often supported by international organizations, to limit their GHG emissions and adapt to increasing climate change and variability.
The IPCC 4 and the World Bank have highlighted the importance of the interrelationships between and integration of climate mitigation and adaptation (IBRD-WB 2010, IPCC 2014b. Integrating mitigation and adaptation can result in multiple co-benefits (Harlan andRuddell 2011, Seto et al 2014). However, mitigation and adaptation plans can be counterproductive when disjointed or improperly coordinated (Laukkonen et al 2009). Furthermore, studies on the integration of mitigation and adaptation (IMA) have increased with focuses on land and water management and urban planning (Swart and Raes 2007); local climate strategies (Laukkonen et al 2009); urbanization typology (Solecki et al 2015); and joint institutionalization in city administrations (Gopfert et al 2018).
The IMA in CCAPs was explicitly addressed for the first time at the national level by Klein et al (2005) in the forestry sector in Bolivia and at the local level by McEvoy et al (2006) in urban areas in the UK. Integrating mitigation and adaptation efforts in CCAPs is increasingly recognized as a way to maximize co-benefits and synergies, minimize trade-offs and conflicts and enhance the cost-effectiveness of planning and implementation (Di Gregorio et al 2017, Grafakos et al 2018).
An evaluation framework for estimating the level of IMA in CCAPs (an IMA index) was only recently developed by Grafakos et al (2019). Only a few studies have addressed the factors associated with IMA in climate change policies (Duguma et al 2014, Grafakos et al 2018.
This study aims to assess the level of IMA in CCAPs in major LAC cities and to explore which institutional, socioeconomic and environmental factors are potential influences. To the best of our knowledge, there is very little related research on CCAPs in the LAC region. Building on an existing body of literature on the analysis and assessment of urban climate policies (Araos et al 2016, Aguiar et al 2018, Reckien et al 2018, this is the first study to address potential factors influencing the level of IMA in local climate action plans in general and in LAC cities in particular.

Methods and data
The study statistically tested factors potentially influencing the level of IMA in LAC cities' CCAPs.
We selected 44 cities in LAC as target cities in this study. The 'IMA index' by Grafakos et al (2019) was adopted for assessing the IMA level in each city's CCAP. Institutional, socioeconomic and environmental factors possibly influential to the IMA level were identified in the relevant literature, and 42 factors were selected based on the context of the LAC region and data availability. Finally, we conducted Pearson's correlation analysis and multiple regression analysis between IMA index and these factors. Detailed methods are described below.

Selection of target cities
The criteria for city selection were: (1) a population size of more than one million inhabitants based on data from UN-DESA (2016) and (2) development of climate policies including (a) stand-alone climate plans, (b) sustainable development or environmental plans, or (c) strategic or territorial plans which include action plans on climate change, climate resilience, sustainable energy, or renewable energy. Where a city developed more than one type of plan that contain climate change actions, priority was given in order of (a), (b) and (c). Additionally, sectoral plans were excluded as these did not focus on overall urban climate issues. Metropolitan-level plans were prioritized over city-level plans. Draft plans, plans in the approval process and adopted plans were all included.
Given the above parameters, we initially identified 68 cities with more than one million inhabitants. San Juan in Puerto Rico was excluded from the sample because Puerto Rico is a US territory. Of the 67 remaining cities (see table A1), 44 had developed some type of climate policy or plan (i.e. type a, b, or c). Therefore, 44 cities in 16 countries were selected (see figure 1), accounting for 28% of the total population of the region.

IMA index (dependent variable)
Local climate policy documents were collected from official websites of LAC city governments in July 2018. We conducted a content analysis of these documents to convert qualitative data into quantitative for evaluating the IMA level. This IMA level was represented by the 'IMA index' based on the evaluation framework of Grafakos et al (2019). Content analysis of web-based data in combination with statistical analysis has been used extensively in climate policy studies (Araos et al 2016, Aguiar et al 2018, Klein et al 2018, Reckien et al 2018. Moreover, utilizing policy documents allows for consistent use of data since all local governments publish and renew climate policy related documents regularly. The evaluation framework of Grafakos et al (2019) consists of 22 qualitative indicators related to the three stages of planning of CCAPs: (1) identifying and understanding, (2) envisioning and planning, and (3) implementation and monitoring. The indicators were scored based on a content analysis of CCAPs in policy documents. The assessment and aggregation of these indicators led to the construction of IMA index (see tables 1 and A3). Cities were classified into three groups according to their total score, IMA index: (i) early-stage integrators (up to 10), (ii) moderate integrators (between 10 and 20), and (iii) advanced integrators (above 20).

Institutional, socioeconomic and environmental factors (independent variables)
Factors potentially influencing the development and implementation of CCAPs were reviewed and assessed  as potentially affecting the level of IMA in CCAPs. These were identified in the literature related to either integrated or stand-alone CCAPs. Integrated CCAPs feature both mitigation and adaptation actions in one plan, while stand-alone CCAPs feature either a mitigation or an adaptation plan (Grafakos et al 2018).
Overall, similar to the study of Reckien et al (2015), factors identified in the literature can be categorized into three types: institutional, socioeconomic, and environmental. Among them, institutional factors were the most common in the literature (IPCC 2007, Corfee-Morlot et al 2009, Bulkeley et al 2011, Duguma et al 2014, Fuhr et al 2018, Grafakos et al 2018. Regarding integrated CCAPs, Duguma et al (2014) identified national-level factors such as common policies and strategies, institutional arrangements, financing, and programs and projects. In addition, Grafakos et al (2018) addressed city-level factors that can drive or hinder integrated climate actions such as structural conditions, along with available resources and technical means.
With regard to stand-alone CCAPs, Corfee-Morlot et al (2009), Reckien et al (2015), and Fuhr et al (2018) identified factors at the city level. According to Fuhr et al (2018), institutional and socioeconomic factors such as the capacity of response to climate-related problems, local democratic practices, and enabling policy frameworks can drive the development of local climate policies. Reckien et al (2015) explored drivers of and barriers to the development of stand-alone CCAPs in European cities; however, the IMA was not explored.
Previous studies have suggested a range of factors at different levels of governance. Considering the vertical and horizontal integration that aligns CCAPs with national policies (Corfee-Morlot et al 2009, Hardoy and Lankao 2011), we included both national and city level factors. Several additional factors were newly included as shown in table 2. We collected data for all independent variables from official websites of international organizations and national and local governments (see table A2 for data sources).

Correlation and multiple regression analysis
Pearson's correlation coefficient analysis was used to compute the level of significance of independent variables (institutional, socioeconomic, and environmental factors) related to the dependent variable, (IMA index). Based on the results of the correlation analysis, independent variables with 0.05 or higher probability value were considered statistically insignificant. These independent variables therefore are not potentially influential to the dependent variable, IMA index, and were excluded from the next stage: a multiple regression analysis. A multiple regression analysis was conducted to test a model to determine the mathematical expression of the relationship between the independent variables (potentially influential factors) and the dependent variable (IMA index).
We used the software Atlas.ti for qualitative analysis of urban policy documents as part of content analysis to measure IMA index. SPSS and Microsoft Excel were used to conduct correlation and multiple regression analysis in order to explore the relationship between potentially influential factors and IMA index.

IMA index
Bogota, Colombia's capital, showed the highest level of IMA among the cities under investigation, with a total score of 28, followed by Mendoza in Argentina, Mexico City in Mexico, and Asunción in Paraguay, all with a total score of 25 (see table 3). The average IMA index of the 44 cities was 14.8, indicating a moderatelevel of integration. Detailed results showed that out of 44 cities, 23 (52%) are moderate integrators, while 11 (25%) fall into the early-stage integrators category and the remaining 10 cities (23%) to the advanced integrators category (see figure 1).
Out of 44 cities, 13 cities explicitly referred to interrelationships between mitigation and adaptation in their action plans (27 actions in total, see table A4). Of these 13 cities, 6 were included in the top 10 ranked cities based on IMA index (tables 3 and 4). Of the total 27 actions, 13 adaptation actions (48%) with mitigation cobenefits and 5 mitigation actions (19%) with adaptation co-benefits were identified. The remaining 9 (33%) were identified as synergistic actions that could achieve both mitigation and adaptation objectives. None of the cities stated any conflicts or trade-offs between mitigation and adaptation. This result could be explained by the rather negative connotation that conflicts and trade-offs between mitigation and adaptation actions may carry. It was found that positive interrelationships (synergies and co-benefits) could occur in the urban greening sector (33%), followed by biodiversity (22%), water (19%), built environment, energy, agriculture, and land use (see chart 1).

Multiple regression analysis with significant factors
Five factors identified from Pearson's correlation coefficient analysis significantly correlated (p-value< 0.05) with the IMA index and were considered predictors when testing for modeling. Those are: -National common climate fund -Global network Urban LEDS -Regional network FLACMA -Regional network UCCI, and -Donor agencies' contribution to the development of CCAPs.
The result of multiple regression analysis using the 'enter method' showed that the model explains 47.3% (R square=0.473) of the cases and can be considered as a model of good-fit based on F-value (6.823>1) and significance p (0.000125<0.001). One predictor donor agency contribution to the development of CCAPs To identify other factors, in addition to donor agency contribution to the development of CCAPs, that may contribute to the model, we applied the 'stepwise method'. This method tests the model by excluding predictors at each step. It is not as commonly used as the 'enter method' due to the risk of the Type II error of missing a significant predictor. However, this risk of Type II error was considered insignificant in this test because the unique significant predictor: donor agency contribution to the development of CCAPs, identified with the enter method, was resulted as one of three factors contributing to the model from the stepwise method. Moreover, this study does not aim to identify the causality (Field 2013).
Multiple regression analysis utilizing the stepwise method showed that the prediction of the model was correct in 45.3% (R square=0.453) of the cases and could be considered as a model of good-fit (F-value 11.029>1 and significance p<0.001). Three predictors were identified as significantly contributing to the model (p<0.05): donor agency contribution to the development of CCAPs, membership of regional network FLACMA and of global network Urban LEDS. Donor agency contribution to the development of CCAPs and membership of FLACMA showed positive relationships with IMA index (0.492, p<0.001 and 0.361, p<0.05, respectively) while the remaining predictor membership of Urban LEDS had a negative relationship. Therefore, the possibility of an increase in IMA index (the level of IMA) rises when receiving donor agencies' assistance in developing CCAPs and being a member of FLAMA, but not of Urban LEDS (phase I).

Discussion
Eight out of the 10 highest scored cities (see table 3) developed CCAPs with support from donor agencies; six are capital cities with the largest population in each respective country. Donor agencies may be inclined to Chart 1. Positive interrelationships stated in action plans by sector.     A6). This implies that smaller cities may receive less support for developing their CCAPs and thus be less likely to have IMA in their planning. In addition, all 10 highest scored cities are members of at least one, global or regional city network. Similarly, Reckien et al (2015) identified climate networks (i.e. Covenant of Mayors, C40 and ICLEI) as significant drivers of both mitigation and adaptation plans. Networks are involved in climate change experimentation/innovation, which is essential for governing climate change in cities (Broto and Bulkeley 2013). Thus, cities' primary expectation for joining networks might be technical support as well as financial resources from networks (Fünfgeld 2015). This engagement might have eventually influenced cities to integrate mitigation and adaptation in their CCAPs. Regional networks FLACMA and UCCI were found to be potential driving factors. Both networks were established in the early 1980s with a common purpose: the development of the region. They also have developed strong, steady relationships between member cities and municipalities over a significant period of time. FLACMA, in particular, has recently undergone organizational restructuring in line with SDGs, which may have led to the incorporation of both mitigation and adaptation policy objectives into their policies. In this sense, strong relationships between member cities and the adoption of a common integrated approach to climate change and sustainable development may have positively influenced the level of IMA in their CCAPs. The global network Urban LEDS showed a negative correlation with IMA index. This is because the program aimed to encourage cities to integrate low emissions and green economy strategies into city development plans. The prioritization of mitigation strategies limited the IMA. During the Urban LEDS phase I (2012-2015), four Brazilian cities out of the 44 target cities were included in its cities network: Belo Horizonte, Curitiba, Fortaleza and Rio de Janeiro. These cities showed an average IMA index of 8.5, a relatively low level of IMA. However, in Urban LEDS phase II (2017∼), the program has adopted the concept of adaptation co-benefits of low emissions development strategies. Therefore, it may provide more support for IMA in the future.
With regard to the driving factor donor agencies' contribution to developing CCAPs, the Inter-American Development Bank (IDB) has been implementing the sustainable urban development program Ciudades Emergentes y Sostenibles (CES) 8 in the region since 2011. Program's approach to the development and execution of action plans includes diagnostic analysis and planning policies addressing mitigation and adaptation simultaneously. Nine 9 out of 44 target cities have developed sustainable development action plans including climate actions under the CES program. The average IMA index of those nine cities is 20.78, an advanced integrator score.
In addition to CES, Mexico implemented the program Plan de Acción Climática Municipal (PACMUN) 10 with support from ICLEI and funded by DFID 11 to promote a policy framework on mitigation and adaptation actions at the local level. Four Mexican cities in our target cities, Aguascalientes, Cuernavaca, Puebla, and Toluca de Lerdo, have participated in this program.
As mentioned at the beginning of this section, eight out of the top 10 ranked cities according to IMA index have developed local CCAPs with support from international organizations. Thus, the implementation of a city-level program adopting a framework with integrated components of mitigation and adaptation may effectively support Latin American cities in enhancing the level of IMA. The remaining two cities from the top 10, Mexico City and Buenos Aires developed CCAPs without external support. In the introduction section of these plans, they clearly outlined an integrated approach to drawing up action plans in response to climate impact analysis. Mexico City has made continuous efforts to design and implement integrated CCAPs joining multiple city networks 12 that promote an integrated approach to climatic challenges. Buenos Aires, likewise, not only has multiple memberships in city networks 13 but also has financial capacity for climate actions. The city showed the third highest GDP per capita among 67 cities with over one million inhabitants in the region (after Panama City in Panama and San Jose in Costa Rica).
A national common climate fund was identified as a significant constraint on the IMA level. Brazil and Mexico established national climate funds in 2009 (regulated in 2010) and 2013 (regulated in 2015) respectively. Even though the Brazilian national climate fund aims at promoting both mitigation and adaptation, it includes more sub-programs on mitigation than adaptation. Under this climate fund, there are two city-focused sub-programs, and these also put more emphasis on mitigation than adaptation (see table A7). Moreover, only 15% of the fund was allocated for adaptation in 2011 (Ludenã and Netto 2011 established the socio-environmental initiative for reducing urban vulnerability, which is based on the national environment fund and climate fund. Thus, a revision of their national climate fund to create a balance between mitigation and adaptation is necessary to help cities achieve integrated CCAPs. Additionally, although Mexico's national climate fund supports both integrated and stand-alone mitigation and adaptation actions, the fund's establishment came after several cities of our sample developed CCAPs. Our study, which focuses on CCAPs in the LAC region, contradicts Duguma et al (2014), who in examining a global sample of countries, found that a national common climate fund was a significant driver of IMA in climate policies. Reckien et al (2015) found that socioeconomic and environmental factors such as population size and density, GDP per capita, unemployment rate, proximity to coast, and average summer and winter temperatures were potentially influential for the development of CCAPs in Europe. Fuhr et al (2018) found that environmentally-concerned civil society and green industries had a significant positive association with the development of CCAPs. In contrast, Duguma et al (2014) identified national income-level as insignificant when it came to the potential synergy between mitigation and adaptation. In our study, all of the socioeconomic and environmental factors proved to be insignificant in relation to the level of IMA. First, IMA requires the preexisting of CCAPs. Second, this might be due to the low explanatory power 14 of the tested factors. As the integration of policy objectives is usually more concerned with institutional and policy arrangements, our results also show that institutional factors are significantly associated with the level of IMA.
Although our approach addresses for the first time the factors that potentially relate to the level of IMA, it has also some limitations. Most of the data used were collected through online searches. Policy documents used for drawing indicators of IMA index were mainly from official websites of local governments. Therefore, cities that have not shared CCAPs documents online inevitably were not considered. As documents were collected from May to July 2018, policy documents published or revised after that period were not considered.
There were challenges regarding the collection of data relevant to the selected factors for the target cities. The ECLAC 15 has been working to disseminate environmental statistics 16 in the region (Quiroga 2018). However, the database is still limited to national level and therefore does not provide city-level data. Data for CO 2 emissions per capita were gathered from different sources (see table A2) since none of the existing data sources provided information on CO 2 emissions per capita for all the sample cities. Thus, the year of reported CO 2 emissions per capita and methods used for measuring them may differ depending on the data source.
In addition, challenges of IMA in urban CCAPs faced by policymakers and local stakeholders were also out of the scope of this study. These could be studied by other methods such as surveys, in-depth interviews, and case studies.
Despite the above limitations, utilization of secondary data produced by governments and international organizations may improve the reliability of the data. Moreover, correlation analysis before multiple regression analysis may contribute to reducing multicollinearity by decreasing the number of variables, excluding insignificant indicators.
To our knowledge, only two studies in the literature addressed the influential factors of the IMA: Duguma et al (2014), focusing on national level and Grafakos et al (2018), with an extensive selection of factors at city level. However, these studies were not region-specific, and the relationship between possible influential factors and the level of IMA was not studied. Reckien et al (2015) addressed both driving and constraining factors for the development of standalone climate plans of a large number of European cities. In this regard, this study is the first one that addresses potential driving and constraining factors associated with the level of IMA in CCAPs. In addition, it is the first study to assess the level of IMA in CCAPs in the LAC region.

Conclusion
Our study, into the potential driving and constraining factors of the level of IMA in CCAPs in LAC cities, found that the significant factors were all institutional factors. Among them, potential driving factors were: (1) membership in regional networks FLACMA and (2) UCCI; and (3) contributions of donor agencies to developing CCAPs. In contrast, factors that potentially constrained the level of IMA were: (1) national common climate fund; and (2) membership of global network Urban LEDS. The results of multiple regression analysis suggest that the level of IMA may increase when a city receives donor agencies' assistance in developing CCAPs or having a membership in FLAMA and may decrease when having a membership in Urban LEDS (phase I). The contribution of donor agencies to the development of CCAPs was identified as the strongest relationship with IMA index, which means that this factor seems most likely to contribute to the level of IMA in CCAPs in the LAC region.
Further research could investigate the causal relationships between influential factors and IMA level, 14 R-squared of all the tested factors and the regression model were under 0.5, and 'Standard error of the estimate' of the regression model was over 5. 15 Economic Commission for Latin America and the Caribbean. which correlation and multiple regression analysis do not determine. Additionally, further study on the relation between the existence of a national climate fund and the level of IMA is needed. The current negative relationship could change in the future for several reasons: the Brazilian government has recently established a new initiative for strengthening urban resilience utilizing the national environment fund and climate fund; and Mexico very recently established an integrated climate fund. Last, case studies could be conducted based on in-depth interviews with policy makers and stakeholders of CCAPs with high-level of IMA to gain a better understanding of the challenges and opportunities of integrating mitigation and adaptation in urban CCAPs.
68 cities were identified with more than one million inhabitants based on UN-DESA (2016), and one city, San Juan in Puerto Rico, was excluded from target cities of this study since Puerto Rico is a USA territory. 67 cities are listed in the table below.
Out of 67 cities, 44 cities were identified with climate-related action plans, and these target cities can be classified by type of climate plans: 32 integrated plans, 9 mitigation plans, and 3 adaptation plans.        Water saving in public buildings and collecting rainwater • Primary objective: adaptation -Reduction in water usage to secure water supply • Mitigation -indirect reduction in CO 2 emissions by using less energy when processing water Suppression of water leakage and rehabilitation of water pipes • Primary objective: adaptation -Reduction in water leakage • Mitigation -Indirect contribution to reducing CO 2 emissions by using less energy in the pumping stations Agriculture Mexico City (Mexico) Production control for the standards of food harmlessness • Primary objective: adaptation -Improvement of local food production Diversification of energy sources for energy supply • Primary objective: mitigation -Reduction in GHG emissions • Adaptation -Improvement of energy system flexibility for the adaptation to hydrology, temperature, wind and other climatic factors Table A5. Results of correlation analysis.

**
Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). b National-level governance structure: Cannot be computed because at least one of the variables is constant.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.