Abstract
The relative magnitude of precipitation and temperature changes obtained from 21st-century climate change projections from general circulation models (GCMs) is compared to the changes in a selection of socio-economic indicators from countries in Central America. The objectives of the study are: (1) to determine the relative influence of climatic and socio-economic variables for different 21st-century scenarios; and (2) to compare the relative situation of each of the countries of Central America, considering climate and socio-economic variables during present and future scenarios. Each socio-economic variable along with the projected changes in precipitation and temperature are used to produce red-green-blue (RGB) composite maps during the historical, mid-century, and end-of-21st-century horizon scenarios. The most consistent result is that the current north–south socio-economic contrast between the countries (in which the southern countries of Panama and Costa Rica present better living conditions than the rest of the Central American countries), is not diminished in the future; and for some combination of scenarios this contrast is exacerbated by future socio-economic differences and climate change impacts. Moreover, Panama and Costa Rica are the only countries that present improved living conditions at the end of the century when considering increases in gross domestic product (GDP) and the effects of climate change. It is worrisome that the north–south differences in the living standards will keep growing in the region, and attention should be given to socio-economic and physical aspects that may play a role in increasing these differences.
Keywords
I Introduction
A considerable amount of scientific research on Earth’s climate based on observations, proxy indicators and numerical modeling has demonstrated that anthropogenic changes in the composition of the atmosphere have resulted in significant changes in global climate (IPCC, 2007). Other authors have also suggested that human-caused global warming may be partly responsible for increases in heavy precipitation (e.g. Allan, 2011). For example, Min et al. (2011) established that recent increases in the intensity of heavy precipitation events over a large part of the Northern Hemisphere land area are critical for reliable projections of future changes. These effects are evident in regional hydrometeorological parameters, suggesting a potential for increasing risk of negative alterations in human and environmental systems. Based on the existence of a large number of endemic species in the area, Karmalkar et al. (2011) argued that Central America is a biodiversity hot-spot. Unfortunately, the region has experienced an exceptional loss of habitat in recent years. They add that the biodiversity hot-spot, spanning most of Central America, is home to lowland dry and montane/cloud forests that host all subtropical and tropical ecosystems from central Mexico to the Panama Canal. Karmalkar et al. (2011) explain that mountain ecosystems and species, where climate zonation is constrained by topography, are particularly susceptible to changing climate. It should be mentioned that although ecosystems have adapted to changing conditions in the past, current environmental changes are occurring at a much faster rate, and pose a serious threat to biodiversity. Moreover, they add that changes in the magnitude and the seasonal cycle of temperature, precipitation, and humidity levels will affect Central American ecosystem dynamics and water availability; or, more importantly, the shortage of water in the future negatively impacts hydropower generation, human consumption, and agricultural activities in the region, also affecting protected areas schemes. The impact on these actual protected areas could have important economic effects on local economies, due to the dependence on import activities like eco-tourism and their associated economic chains (Moreno-Diaz et al., 2011).
In addition, present and future socio-economic deficiencies are factors that can contribute to exacerbate the impacts of climate variability and change and vice versa. One commonly mentioned example of climate hazard that can cause significant damage in Central America is the presence of hurricanes. Kunkel et al. (2008) observed that Atlantic tropical cyclone (hurricane) activity, as measured by both frequency and the Power Dissipation Index (which combines storm intensity, duration, and frequency) has increased. The increases are very significant since about 1970, and are substantial since the 1950s and 1960s in association with the warming of Atlantic sea surface temperatures (SSTs). Gutowski et al. (2008) add that it is likely that hurricane/typhoon wind speeds and core rainfall rates will increase in response to human-caused warming. Analyses of model simulations suggest that for each 1°C increase in tropical SSTs, hurricane surface wind speeds will increase by 1–8% and core rainfall rates by 6–18%. The resilience of a country or region that has been hit by a hurricane can be significantly diminished for many years after the event. Such was the case when the impacts of the 1998 Hurricane Mitch in Central America were estimated in 2009 at six billion dollars (AECID, 2009), suggesting that this extreme event caused long-term effects in the economies of the region.
This diminished capacity to recover is particularly true for the countries in Central America, a region characterized by economically vulnerable societies and affected with large hydroclimatic variability and contrasts (Alfaro et al., 2010). In fact, Central America is a region that may have increased vulnerability to extreme events in the future (e.g. Gutowski et al., 2008; IPCC, 2007; Min et al., 2011). We adopt a three-dimensional approach in this study, understanding vulnerability as the combination of different social and environmental factors. Following Rodriguez-Herrero and Bozada-Robles (2010), we adopt their definition of vulnerability as the probability for a community to suffer human and material damages when it is exposed to the more frequent natural, technological, or anthropogenic hazard. That probability is related also to the fragility of elements like infrastructure, housing, productive activities, degree of organization, warning systems, and institutional political development, among others, in which the magnitude of the damages is associated with the degree of vulnerability. Therefore, being vulnerable is to be susceptible to damages and to have difficulty in recovering from these damages, in addition to poor adaptation capacity. In this way, the term resilience is associated with the capacity of coping with the damages situation.
Regional climate projections from the fourth assessment report (AR4) from the Intergovernmental Panel on Climate Change (IPCC) obtained from a compilation of general circulation models (GCMs) have been assessed for a number of regions in the world (Giorgi and Francisco, 2000; Karmalkar et al., 2011). One of these regions is the Central America and Mexico region (10–30°N latitude and 83–116°W longitude). In these simulations, the warming in Central America from the multimodel ensemble under the SRES A1B scenario is greater than the global warming. Moreover, 19 (out of 21) GCMs agree on the direction of the precipitation change predicting a decrease in precipitation (Karmalkar et al., 2011). Giorgi (2006) developed an index based on projected changes in the mean and variance of precipitation and temperature. They found that the Central America region is a climate change hot-spot. No other region in the tropics showed changes as large as Central America (Giorgi, 2006; Karmalkar et al., 2011). Rauscher et al. (2011) argued that this broad future dry pattern observed in Central America is consistent with an amplification of the subtropical high-pressure cells and an equatorward contraction of convective regions (Sachs and Myhrvold, 2011), but Rauscher et al. (2011) add that an important factor that contributes for this drying signal is that SSTs over the Tropical North Atlantic do not warm as much as the surrounding ocean in future scenarios.
Chapter 13 of IPCC (2007) contains a summary of some of the projected impacts of climate change in Latin America and their association with socio-economic vulnerabilities. For example, it is mentioned in this report that the poorest communities are among the most vulnerable to extreme events and that some of these vulnerabilities are caused by their location in the path of hurricanes (about 8.4 million people in Central America; FAO, 2004), on unstable lands, in precarious settlements, on low-lying areas, and in places prone to flooding from rivers (IPCC, 2007). In addition, Latin America in general and Central America in particular have been subjected to climate-related impacts caused by increased occurrence of El Niño-La Niña events that have contributed to heightened vulnerability of human systems to natural disasters such as floods, droughts, and landslides (IPCC, 2007; Trenberth and Stepaniak, 2001). In addition to weather and climate, the main drivers of increased (local) vulnerability are demographic pressure, unregulated urban growth, poverty, rural migration, low investment in infrastructure and services, and problems with intersectoral coordination (IPCC, 2007). A detailed account of all these vulnerabilities and other non-climatic stresses is outside the scope of this study, but it is assumed here that the use of socio-economic indexes can summarize some of these aspects in a broad context. Also, climate change is represented here using the projected changes in precipitation and temperature data from the GCMs, but climate change can manifest itself through other (not considered) impacts in Central America, such as damage caused by increases in the intensity or number of hurricanes, coastal damage due to mean sea level rising, impacts in human health due to changes in environmental conditions, landslides, local flooding and drought, and others. Additionally, recent studies highlight the fact that local vulnerability in Central America is strongly related to land use (Amador-Astua and Bonilla-Vargas, 2009). These authors concluded that it is normal for local governments of the region to produce poor land-use plans and this produces a lack of understanding about risk in the communities. Amador-Astua and Bonilla-Vargas (2009) state that inadequate land-use conditions combined with poor living conditions caused most of the human dimension impacts associated with 1998 hurricane Mitch. In that sense, it is necessary to develop extensive educational plans in order to make the communities understand the local risk and vulnerability, and to alleviate the disaster occurrence and impact in rural and urban areas.
In this article we contrast socio-economic variables with environmental variables for different projected horizons of time in the 21st century. The objective is to compare the relative importance of society’s present and projected vulnerabilities and climate impacts. Instead of an assessment of the impacts of climate change in society, the approach is comparative: we would like to know what constitutes (now and in the future) the dominant hazard and the largest differentiating factor for the people of Central America’s countries: socio-economical challenges or climate change. The proposed analysis will also allow us to determine differences between the countries of the region with respect to their present and future vulnerabilities. Four societal variables will be used: the Human Development Index (HDI), gross domestic product per capita (GDP), population totals (POP), and Costa Rica’s ‘Indice de Desarrollo Social’ (IDS) or Social Development Index (SDI). The SDI was selected as it contains socio-economic information at a relatively high resolution (county level), although the index is only available for Costa Rica. Two environmental/meteorological variables will be used corresponding to the changes in annual precipitation and temperature throughout the 21st century, obtained from an ensemble of 30 simulations of future climate projections from general circulation models (GCMs; see section II). The analysis consists of constructing red-green-blue (RGB) composites of each socio-economical variable (red channel) with the changes in precipitation (green channel) and temperature (blue channel) for three different time horizons: the historical horizon, the mid-century horizon, and the end-of-century horizon. By producing these composites, the corresponding hue of the maps will be indicative of the relative importance of each variable, while the saturation will be indicative of adverse or favorable conditions caused by socio-economical and climate projections.
II Data
The 2007 HDI was obtained from the United Nations Development Programme website (UNDP, 2011). The HDI is an index that takes into consideration three major dimensions that quantify the state of a country: health, education, and living standards. This is measured using the following indicators: life expectancy at birth (health dimension); mean years of schooling and expected years of schooling (education dimension); and gross national income per capita (living standards dimension). Larger HDI numbers correspond to better socio-economic conditions. Details on how these indicators are combined to produce the HDI can be found in UNDP (2010). GDP and POP estimations and projections for 2005, 2050, and 2100 for all countries in Central America were obtained from the Economic Commission for Latin America and the Caribbean (ECLAC, 2010). The 2007 SDI was obtained from Costa Rica’s Ministry of National Planning and Economic Policy (MIDEPLAN, 2007). The SDI is an index that is normalized for all the counties and districts (cantones and distritos in Spanish) of Costa Rica. A value of zero (one) correspond to the less (more) developed county or district. According to MIDEPLAN (2007) the SDI for any political division is constructed taking into account the population with access to some basic social rights from four dimensions: (1) access to participate in the economic activities and to enjoy an adequate job that generates enough income to cover at least certain basic needs; (2) access to an active social participation in the national and local elections; (3) access to health services and the achievement of good health standards/networks; and (4) access and use of the education system.
In this study the climate change projections will be obtained from general circulation models (GCMs). A GCM is a model that simulates the global climate numerically. It consists of a mathematical representation of the physical laws and processes that govern Earth’s climate (Amador and Alfaro, 2009). The models are forced with historical and future projections of climate forcings (e.g. solar, volcanic, greenhouse gas emissions). GCMs are limited by our understanding of what drives, forms, and affects Earth’s climate and how it responds to a series of external forcings. They are also limited by the capacity and speed of current computer systems. There are many laboratories and climate centers in the world that run different GCMs with somewhat different physics and assumptions. In addition, an individual model can have multiple runs depending on the initial conditions of each run that result in somewhat different weather patterns. A total of 30 global climate simulations (Table 1) corresponding to monthly precipitation and temperature runs for the climate of the 20th century (known as 20c3m runs) and climate projections for the 21st century for the A2 greenhouse gas emission scenario were obtained from the US Lawrence Livermore National Laboratory Program for Climate Model Diagnosis and Intercomparison (PCMDI, 2010) and from the IPCC (2010). In some cases two runs were available (Table 1). Global climate change data from their original resolution were interpolated to the resolution of the coarser model (2° latitude × 5° longitude) by the nearest grid-point method, but considering separate interpolations for the ocean and land grid-points according to the individual land-sea masks of the models. The data were visually inspected at selected grid-points. The data were also changed to the same units and same file format for the rest of the analysis. More details on the characteristics of these data can be found in Hidalgo and Alfaro (forthcoming). We performed an analysis on how the models reproduce statistics of the 20th-century climate in the Eastern Tropical Pacific, a region very close to Central America. In general the models showed good results for temperature, but did not perform very well for precipitation for a series of metrics used to compare the statistics from the 20c3 m runs and the 1950–1999 NCEP/NCAR reanalysis data (Kalnay et al., 1996). However, consistently with Brekke et al. (2008) and Pierce et al. (2009) it was found that the ranking of the models was very sensitive to the metric used and that the use of ensembles of many models helped to reduce the noise in the data. For this reason, it was considered that the multi-model ensemble of the 30 simulations would reduce the uncertainty of the projections and provide a better indication of future precipitation and temperature.
Climate simulations used in this study
III Methodology
The analysis is based on the RGB composites of each of the socio-economical variables with the climate variables (precipitation and temperature changes). Each socio-economical variable was normalized from zero to one considering the range of variability of the three horizons. The HDI, GDP, and SDI were inverted before the normalization to allow zero (one) to always represent the most favorable (adverse) condition. Conversely an increase in POP was considered an adverse condition. Note that the direction of the favorable or adverse conditions for the socio-economical variables is a convention used in this article that is free to be questioned. The ensemble of all the GCM simulations were computed for three scenarios: ‘historical’ (1950–1999), mid-century (2000–2049), and end-of-century (2050–2099). The global maps consisting of the differences in annual precipitation and temperature (ΔP and ΔT) between each of the scenarios and the historical scenario were computed based on the 30-run ensembles. For each country, the closest GCM grid-point to the country’s capital city was used to select a corresponding ΔP and ΔT from the grids of the GCMs. We used those points for two reasons: first, because the sizes of the countries in Central America are normally smaller than the size of one or two GCM gridpoints; and, second, because the economies and development of these countries are normally centralized around their capital cities. Note that one limitation of this study is that the GCM data were used without downscaling. Downscaling is a process of transferring the data from coarse models (such as GCMs) to fine-scale grids using a regional climate model or statistical procedures (Amador and Alfaro, 2009; Hidalgo et al., 2008; Maurer and Hidalgo, 2008). Since only the ΔP and ΔT were of interest (and not the variability of the GCM data) the nearest gridpoint approach used here to select the impact of climate change on the countries or counties was considered accurate enough for our purposes.
The absolute values of ΔP and ΔT were normalized from zero to one considering the range of variation of the three horizons, in such a way that zero (one) corresponds to the most favorable (adverse) condition. In terms of the climate variable, it should be mentioned that the words favorable and adverse were used here to describe little and large climate change impact, respectively, without regard that in some particular case large ΔP and ΔT could actually be associated with positive impacts in any particular human or environmental sector. Each country (or county) of the resulting maps was colored by composing the RGB color with the three normalized indexes: (1) the socio-economic index (red channel); (2) the ΔP index (green channel); and (3) the ΔT index (blue channel). The composition then suggests that the [0,0,0] color (black) is the overall most favorable condition (corresponding to favorable socio-economic conditions and no climate change) and [1,1,1] color (white) is a dire situation in which unfavorable socio-economic conditions (e.g. a degradation or depletion in the social standards according to the index) are coupled with a large climate change signal in precipitation and temperature. The primary colors correspond then to the dominance of (1) the socio-economic variable (red), (2) the ΔP (green) and (3) ΔT (blue). Composed colors signify as follows: (1) yellow suggests influence of the socio-economic variable and ΔP; (2) magenta suggests influence of the socio-economic variable and ΔT; and (3) cyan suggests influence of ΔP and ΔT. Note that the absolute values of ΔP and ΔT were used when producing the index that generate the colors; for that reason the sign of the climate change signal was not considered in terms of the impacts in the countries (however, the sign of the climate change variables can be found in the text included in some of the figures and in a table presented later). Note also that the normalized socio-economic variable, normalized ΔP and normalized ΔT were given similar weight in terms of their importance to produce adverse conditions in the region, but this is precisely the aim of the study – to present a comparative analysis of their relative contributions in the present and in the future.
In another part of the analysis, the Euclidean distances between the [0,0,0] (most favorable condition) and the three digits that represent the color of each country were computed in order to have a single digit to represent the socio-economic state of each country. Larger (shorter) distances are related to more adverse (favorable) conditions.
IV Results
The spaghetti plots of the 21st century projected precipitation and temperature anomalies from the GCM are shown in Figures 1 and 2, respectively. The spread of the anomalies represent the variability from all the runs in Table 1. In Table 2 the ΔP and ΔT for different horizons are computed. The base period for the anomalies is the 1950–1999 period. As can be seen, the ensemble of precipitation projections at the end-of-century tends to be somewhat drier than historical for all the countries, except for Costa Rica which showed little change. These results agree in general terms with the estimations obtained by IPCC (2007), Rauscher et al. (2011), and Ruosteenoja et al. (2003) for the IPCC SRES scenarios. Studying the seasonality in more detail, Karmalkar et al. (2011) found a dry bias in the wet season and a wet bias in the dry season suggesting that their model was unable to capture the full range of precipitation variability. Projected warming under the A2 scenario is higher in the wet season than that in the dry season with the Yucatan Peninsula experiencing the highest warming, but a large reduction in precipitation in the wet season is projected for the region, whereas parts of Central America that receive a considerable amount of moisture in the form of orographic precipitation show significant decreases in precipitation in the dry season. In Table 2 it is shown that in percentage terms the drying of the region tends to be relatively small for Costa Rica, Panama, El Salvador, and Honduras, but it seems to be important for Guatemala, Belize, and Nicaragua, in particular at the end-of-century scenario. It should be noted that at least one individual model projects a much more severe drying for all the countries, although it seems to be an outlier compared to the majority of the models (Figure 1). In terms of temperature, there is a clearer anthropogenic signal in all the models for all the countries (Figure 2). This signal is a consistent warming of around 1°C for the mid-century scenario and of more than 2°C at the end-of-century scenario (Table 2). It should be noted that Guatemala and Nicaragua showed the largest warming at the end-of-century scenario of all the countries. There is also an outlier model in the case of temperature that did not show much future warming for the region. The spread of the models for the two variables tends to be smaller or larger for certain GCM gridpoints (used for each country). This suggests different uncertainty in the future projections for each of the countries.

Spaghetti plot of the projections of annual precipitation anomalies from 30 GCM runs. The plot for Belize is the same as the one for Guatemala. For each country, the nearest GCM gridpoint to the capital city was selected. The anomalies were constructed with respect to the 1950-1999 period.

Spaghetti plot of the projections of annual temperature anomalies from 30 GCM runs. The plot for Belize is the same as the one for Guatemala. For each country, the nearest GCM gridpoint to the capital city was selected. The anomalies were constructed with respect to the 1950-1999 period.
ΔP and ΔT for different horizons. The precipitation percentage changes shown in parentheses are based on the change with respect to the mean for the 1950–1999 period.
In Figure 3, the RGB composite for the HDI (inverted) socio-economic variable is presented. Contrary to the case of the GDP and POP, there are no future regional projections for the HDI available, and therefore all the maps of the figure were constructed using the 2007 HDI values. For this reason, Figure 3 should be interpreted as the relative influence of climate change given the 2007 HDI conditions. Historical HDI trends across the region in the last decade are normally small, as well as those associated with SDI for Costa Rica (not shown). As can be seen in Figure 3, the historical and the mid-century maps are very similar to each other, suggesting that climate change impacts in relative terms are small during the first 50 years of the century compared to the last 50 years. In the historical and mid-century horizons the northern part of the Central America region tends to be red and lighter in color than the southern countries, suggesting that the HDI is an important factor in comparison to climate change in this region. Conversely, for these scenarios Costa Rica and Panama tend to show black and dark blue colors, suggesting favorable conditions with only a little influence of temperature. The end-of-century horizon map tends to be significantly lighter than the other maps, suggesting a significant influence of climate change in the region. It should be noted that there is an influence of temperature changes in the southern part of the region. In addition the difference between north and south conditions (seen in the difference in saturation) is maintained. This difference is probably associated with the HDI influence. Guatemala is the country with the lightest color, suggesting the most adverse conditions at the end-of-century. Belize shows some influence of precipitation changes and the rest of the countries show some influence of the HDI.

RGB composite of the inverse of the 2007 Human Development Index (HDI) and changes in precipitation and temperature of the ensemble of 30 GCM runs for two different horizons. The color of each country is determined by the color composed of the normalized values of the HDI (inverted), P and T. The values are relative to the data for all countries and horizons. Light (dark) colors are associated with more adverse (favorable) conditions.
The north–south polarization is also evident in terms of the Euclidean distances (Figure 4). Guatemala and Nicaragua have the most adverse conditions in all three horizons, but at the end-of-century climate change may produce comparable adverse conditions in Belize. Costa Rica and Panama have the most favorable conditions of the region, but if the HDI does not improve by the end-of-century climate change would make these countries have the same level of stress as the current values for El Salvador and Honduras.

Euclidean distances with respect to the point [0,0,0] of the vectors composed of the HDI (inverted), precipitation changes, and temperature changes for the different horizons. Large (small) distances are associated with more adverse (favorable) conditions. (CRI: Costa Rica. PAN: Panama. BLZ: Belize. SLV: El Salvador. HND: Honduras. GTM: Guatemala. NIC: Nicaragua.)
As we mention above, in spite of the fact that HDI trends are small, it is unlikely that the HDI will keep fixed at the historical levels in the future. For this reason the analysis was performed with the GDP and POP; although these are more limited indexes, there are projections of them available associated with future country’s social development (ECLAC, 2010). As was previously mentioned, we interpreted that an increase (decrease) in the POP is associated with a more (less) stressfully social condition, and the opposite is true for GDP. The analysis is shown in Figures 5–8. As can be seen in Figure 5, the socio-economic variable dominates the first two horizons and both maps are very similar to each other. This is because the indexes of the climate variables are low (close to zero) for both maps, while the GDP is close to one (GDP is extremely low for the historical scenario compared to its value at the end-of-century according to ECLAC, 2010). This suggests a non-linear increase in the GDP, causing very large values in the second half of the century. A remarkable conclusion of this analysis is that, according to this map, current GDP differences between the countries are small in comparison to the end-of-century GDP differences, suggesting more north–south polarization at the end-of-century caused by economic and climatic influences. These results can be seen in Figure 6 also, where the historical polarization caused by differences in the GDP is very small compared to the end-of-century horizon scenario. In fact at the end-of-century only Costa Rica and Panama were able to improve their overall conditions (represented in a decrease in their Euclidean distances) compared to the historical scenario, suggesting better living standards regardless of climate change influences. Guatemala, Belize, and Nicaragua present the most stressful conditions according to the figure.

RGB composite of the inverse of the Gross Domestic Product (GDP) per capita and changes in precipitation and temperature of the ensemble of 30 GCM runs for two different horizons. The color of each country is determined by the color composed of the normalized values of the GDP (inverted), P and T. The values are relative to the data for all countries and horizons. Light (dark) colors are associated with more adverse (favorable) conditions.

Euclidean distances with respect to the point [0,0,0] of the vectors composed of the Gross Domestic Product (GDP) per capita (inverted), precipitation changes, and temperature changes for the different horizons. Large (small) distances are associated with more adverse (favorable) conditions. (CRI: Costa Rica, PAN: Panama, BLZ: Belize, SLV: El Salvador, HND: Honduras, GTM: Guatemala, NIC: Nicaragua).

RGB composite of the population (POP) and changes in precipitation and temperature of the ensamble of 30 GCM runs for two different horizons. The color of each country is determined by the color composed of the normalized values of the POP, P and T. The values are relative to the data for all countries and horizons. Light (dark) colors are associated with more adverse (favorable) conditions.

Euclidean distances with respect to the point [0,0,0] of the vectors composed of the population (POP), precipitation changes, and temperature changes for the different horizons. Large (small) distances are associated with more adverse (favorable) conditions. (CRI: Costa Rica, PAN: Panama, BLZ: Belize, SLV: El Salvador, HND: Honduras, GTM: Guatemala, NIC: Nicaragua).
In terms of POP (Figure 7) the current differences between the countries are still small compared to the other horizons. By mid-century there is some influence of POP and climate that makes differences between the countries more evident. Although still polarized, at the end-of-century horizon POP shows less north–south polarization compared with the other variables, but Guatemala shows dire conditions at the end-of-century (Figures 7 and 8).
Having a single index representing each of the countries of Central America can hide contrasts at smaller scales. That is why we performed the analysis using the SDI from Costa Rica at the county level (Figure 9). As for Figure 3, the analysis was performed using fixed 2007 SDI data at the county level and GCM data that represent four major climate divisions according with the convention used by the National Meteorological Institute of Costa Rica. During the historical period two counties showed the worst SDI conditions. Those counties are located in the countryside close to the borders with Panama and Nicaragua. Conversely, the Central Valley that contains the major metropolitan area (denoted by region 3 in Figure 9) has the best socio-economic conditions of the country. As can be seen in Figure 9, socio-economic influences prevail in the mid-century horizon in most of the South Pacific and Caribbean regions, while significant influences of precipitation changes are present in the North Pacific and Central Valley region. At the end-of-century horizon, the socio-economic impact is still an important factor in the South Pacific region, precipitation changes are dominant in the North Pacific (climatologically the driest region of the country) and in the Central Valley, while a mix of temperature changes and SDI influences the counties of the Caribbean region. Almost half of Costa Rica’s population lives in the Central Valley. Water demands in this region are consequently high and the population there is thus especially vulnerable to changes in precipitation (Adamson-Badilla and Masis, 2010).

RGB composite of the inverse of the 2007 Social Development Index (SDI) and changes in precipitation and temperature of the ensemble of 30 GCM runs for two different horizons. The color of each county (cantón in Spanish) is determined by the color composed of the normalized values of the SDI (inverted), P and T. The values are relative to the data for all counties and horizons. Light (dark) colors are associated with more adverse (favorable) conditions.
V Discussion and conclusions
The analysis of the socio-economic variables and climate change showed that there is a north–south polarization in the condition of Central America countries that will not be diminished in the future (see also ECLAC, 2010). The main characteristic of this polarization is the difference in living conditions in Panama and Costa Rica compared to the rest of the Central American countries. This polarization is also evident in a cluster analysis of the historical data of the HDI (Figure 10a). In this case, Costa Rica and Panama group together, and Nicaragua and Guatemala group together at similar distances. El Salvador is more similar to Nicaragua and Guatemala than Honduras. In terms of the GDP, Panama and Costa Rica group together, while Honduras and Nicaragua and El Salvador and Guatemala form less-defined groups. In terms of POP, Panama and Costa Rica form a well-defined group, and Honduras, Nicaragua, and El Salvador show small distances.

Cluster tree using the historical values of socio-economic variables. Belize was excluded from the HDI plot because of missing data.
There is a feedback between the impact of natural disasters (caused, for example, by climate change) and the resilience of the countries to future extreme hydroclimatic or geological events. That is, the vulnerability of the country could increase for years after a large hydroclimatic or geological event (Maskrey, 1997). This feedback is not considered in this analysis, but it was evident in the case of the impact of hurricane Mitch in Central America. Nicaragua and Honduras, ranked by the World Bank in the bottom third of the poorest countries, have had difficulties in coping with impacts of natural hazards – for example, Mitch in 1998 and the 1972 earthquake in Managua, Nicaragua. This study has focused on precipitation and temperature changes as environmental variables, but other variables such as sea level change for coastal regions could have an important role for climate change impacts in Central America (ECLAC, 2011). In addition, non-climate environmental hazards such as earthquakes and tsunamis (e.g. Chacon-Barrantes and Protti, 2011) could have an important impact, with effects enduring for decades, in the economies of the Central American countries.
The signal of climate change and the variation in some socio-economic variables in Central America present much less relative impact in the first half than in the second half of the century. Unfortunately, the future impact of climate change depends greatly on what is being done in the present, as there is a lag between greenhouse gas emissions and climate impacts. Therefore, the prospect of observing relatively small impacts during the first half of the century may prevent decision-makers from taking more aggressive actions in terms of mitigation and adaptation to climate change in the long term, which can result in even larger impacts in the second half of the century. Those issues should be taken into account by the Central American Integration System (SICA in Spanish) in the determination of regional climate change policy.
It is interesting to note that according to the GDP and climate change data the only countries that improve their overall condition at the end-of-century scenario compared to current conditions and regardless of climate change are Costa Rica and Panama. The analysis using the SDI from Costa Rica at the county level also revealed that the contrasts are present in this analysis for a single country. Even though the GDP in the rest of the countries will increase, the impact of climate change is severe enough to degrade living standards at the end-of-century especially in Guatemala, Nicaragua, and Belize. When POP is combined with climate change, the living conditions deteriorate in all the countries, but in particular in Belize, a country that showed the best conditions during the historical period and the second worst conditions (after Guatemala) at the end-of-century. Guatemala has historically the largest birthrate in Central America and it will reach its maximum population level in 2080 versus 2055 for Costa Rica (ECLAC, 2010). It is worrisome that north–south differences in the living standards will keep growing in the region, and attention should be given to socio-economic and physical aspects that may play a role in increasing these differences.
Footnotes
Acknowledgements
The authors are obliged to: André Stahl from UCR, who processed much of the raw GCM data; Mary Tyree from Scripps Institution of Oceanography, who provided the land-sea masks of the models; María Fernanda Padilla and Natalie Mora, for their help with the database; Paula Pérez, who helped to elaborate Figure 9; and Carla Vega, who helped with
. Thanks also for the logistic support of the School of Physics of UCR.
This work was partially financed by projects 808-A9-180, 805-A9-224, 805-A9-532, 808-B0-092, 808-B0-654, 805-A8-606, 805-A7-002, and 808-A9-070 from the Center for Geophysical Research (CIGEFI) and the Marine Science and Limnology Research Center (CIMAR) of the University of Costa Rica (UCR) and the project IAI-CRN2050. The authors were also funded through an Award from Florida Ice and Farm Company (Amador, Alfaro and Hidalgo). HH is also funded through a grant from the Panamerican Institute of Geography and History (GEOF.02.2011).
