Scales of Change: The Climatic Impacts of Tropical Deforestation in Chiapas, Mexico
Karen L. O'Brien
Center for International Climate and Environmental Research (CICERO)
Oslo, Norway
Scales of Change: The Climatic Impacts of Tropical Deforestation in Chiapas, Mexico
Karen L. O'Brien
Center for International Climate and Environmental Research (CICERO)
Oslo, Norway
There is a growing recognition that Earth's vegetation plays an important role in the climate system. The relationship between tropical forests and climate is one of the most challenging aspects of atmosphere-biosphere interactions, and the rapid rate of forest loss adds a sense of urgency to efforts to understand the implications of deforestation on climate. Although there is some empirical evidence to suggest that deforestation leads to changes in temperature or rainfall, the vast majority of the work on this subject has been conducted through simulations with mathematical models. General circulation models coupled with atmosphere-biosphere models provide increasing evidence that deforestation can significantly influence the climate at a number of scales.
In scaling up to higher levels of analysis, emergent properties may appear as a result of synergistic interactions taking place at higher levels of system integration, such as the regional or global scale.
Although modeling studies profess to consider the impacts of deforestation at the local, regional, and global scales, the underlying framework of the models remains unchanged at each of these scales. In fact, the same physical equations are assumed to govern the relationships at all scales, and it is simply the extent of spatial representation or analysis that is altered. The only empirical connections between real tropical forests and the models occur in the parameterizations included in atmosphere-biosphere models, which are often scaled up from observed measurements. However, the relationships among variables at larger scales may differ significantly from those occurring at smaller scales. In scaling up to higher levels of analysis, emergent properties may appear as a result of synergistic interactions taking place at higher levels of system integration, such as the regional or global scale. Further more, tropical deforestation usually results in a mosaic pattern of land cover, and there is evidence that the atmospheric response to a heterogeneous land surface is nonlinear (e. g., Pielke and Avissar, 1990). The issue of scale in tropical deforestation simulations has clearly not been adequately resolved.
Tropical deforestation usually results in a mosaic pattern of land cover, and there is evidence that the atmospheric response to a heterogeneous land surface is nonlinear.
One way of identifying such scaling influences is to compare modeling results with historical climate records collected in areas that have experienced significant deforestation. O'Brien presented the results of an empirical study of deforestation and climate change for an area located in the Selva Lacandona of Chiapas, Mexico (Figure 1.20), which forms part of North America's largest remaining tropical rain forest. In this 19,000 km2 of mountainous terrain, a network of climate stations was established during the 1960s, prior to the onset of large-scale deforestation. Daily climate records from 18 of these stations (Figure 1.21) were analyzed, along with land cover changes evaluated using satellite imagery and field observations. A variety of deforestation patterns can be identified around the stations, varying from very little remaining forest cover to almost complete forest cover. Climatic change and deforestation were then considered together to determine whether a relationship between deforestation and climate change is evident at the local scale.
Comparisons among a tight network of highly correlated climate stations can be made based on the premise that variations in average weather show similar tendencies over fairly large regions. Synoptic-scale controls such as El Niño/Southern Oscillation (ENSO) events should exert a similar influence over all of the stations, regardless of the amount of remaining forest cover. Likewise, if global climatic change is evident, then it should appear consistently in the records of stations in close proximity to one another. If highly correlated stations' records exhibit differential trends, then local land cover changes can be considered a contributing factor (Karl and Williams, 1987).
There is a growing recognition that the simulated impacts of deforestation on the climate are regionally specific, in large part due to the different scales of the deforested areas.
In empirical and modeling studies of deforestation and climate change there is a growing recognition that the simulated impacts of deforestation on the climate are regionally specific, in large part due to the different scales of the deforested areas (Zhang et al., 1996). While most modeling efforts have focused on the Amazon basin, more recent experiments covering Southeast Asia and tropical Africa show that the direction and magnitude of the changes may be very different outside of the Amazon region.
O'Brien's local scale analysis shows a strong tendency for maximum daily temperatures to decrease at climate stations exhibiting high deforestation, particularly to the northeast of the station. No changes in precipitation were observed. The temperature decrease is consistent with the results of modeling studies from regions outside of Amazonia, as is the lack of precipitation changes. The observed temperature decreases are driven by a decline in the number of extreme temperature events. This is not consistent with the modeling studies, largely because general circulation models do not adequately represent climate variability and extreme events. The results support the growing recognition that scale is a significant factor in determining the climatic impacts of land cover changes (Avissar, 1995; Raupach and Finnigan, 1995). One possibility is that landscape-level processes influence the climate in deforested areas, introducing local circulations that may not be evident at the microscale or at regional or global scales. There is mounting evidence that global, regional, local and even microscale processes represent different realms of analysis, and that simple extrapolations across scales provide an inadequate means of addressing the impacts of deforestation on the climate.
O'Brien's local scale analysis shows a strong tendency for maximum daily temperatures to decrease at climate stations exhibiting high deforestation, particularly to the northeast of the station.
One means of identifying the local impacts of tropical deforestation is to examine climate records from a tropical forest region and relate them to changes in the surrounding land cover. The Selva Lacandona of Chiapas, Mexico serves as an excellent site for examining such impacts. A network of climate stations was established between 1957 and 1970, when the forest was still a sparsely inhabited frontier (see Figure 1.22). Both mechanized logging and colonization got underway during the 1960s, initiating deforestation in dispersed areas. The colonization process accelerated during the 1970s and 1980s, increasing the population and transforming the land cover in many parts of the Selva Lacandona. By the early 1990s, much of the land within the region had been claimed or titled, and forest surrounding many of the once-remote climate stations had been dramatically transformed. Nevertheless, some areas have experienced relatively little change, particularly those located close to protected areas. If local-scale land cover changes have an impact on the climate, then changes may be evident at some of the stations within the Selva Lacandona, and possibly absent at others.
To evaluate climate trends in the Selva Lacandona, local and regional data sets were compiled and analyzed based on daily readings at 18 climate stations including maximum and minimum temperature and precipitation. First, the mean daily temperature was calculated by averaging the maximum and minimum temperatures. Next, the daily temperature range was calculated by differencing the daily minimum and maximum temperatures. The daily data were also used to calculate the annual average temperatures, temperature range and total precipitation, as well as frequency of temperature and rainfall extremes and the number of rainy days per year. Once assembled, the complete data sets were screened for biases that might appear as artificial climate changes. Finally, the data sets were subject to statistical analysis to determine whether a trend or discontinuity could be identified.
Deforestation in the Selva Lacandona reflects a dynamic process that is neither spatially nor temporally homogeneous, resulting in a mosaic pattern of land cover. Fields of secondary forest in various stages of regrowth are common. To interpret deforestation patterns, an historical analysis of the social driving forces was undertaken (O'Brien, 1998). A political ecology framework was used to understand the driving forces of deforestation, as well as the countervailing pressures for conservation. This analysis, as well as interviews with station personnel and archival photographs taken during the construction period, confirm that most of the areas were forested when the stations were established.
To provide quantitative estimates of deforestation, Landsat images were analyzed and interpreted. Multispectral Scanner System (MSS) and the Thematic Mapper (TM) data from several periods were acquired to create a time series of images for the Selva Lacandona. For each area extracted from the satellite images, the amount of forest change was calculated, along with the percentage of forest cover remaining in the most recent image. Results indicate that there is no clear-cut distinction between "forested" and "deforested" stations. Instead, deforestation appears as a continuum among stations (see Figure 1.23). The distinct patterns that emerge reflect the physical geography of the area, particularly the configuration of mountains and valleys. The location of roads and rivers also influences the patterns of deforestation.
Deforestation in the Selva Lacandona reflects a dynamic process that is neither spatially nor temporally homogeneous, resulting in a mosaic pattern of land cover.
In order to evaluate multiple interpretations of "local," areas of different sizes surrounding the climate stations were extracted for the analysis of deforestation. The regions were centered around the pixel where the climate station was located, and they included circles with radii of 0.5 km, 1 km, 3 km, 5 km, 10 km, and 15 km. Figure 1.27 is an example of this, showing deforestation around Chajul climate station in 1979, depicted at six spatial scales.
There is no clear-cut distinction between "forested" and "deforested" stations. Instead, deforestation appears as a continuum among stations.
Maximum temperatures show the most interesting results in the Selva Lacandona. Ten stations reveal strong decreasing trends (see Figure 1.24). The trends are highly significant at seven of the stations, and striking but non-significant at the remaining three. It appears that 1973, 1983 and 1991 were relatively warm years at most of the stations, whereas 1976 and 1990 tended to be cooler. However, the temperature peaks in later years rarely attain the height of the earlier peaks. There is clear evidence that daytime temperatures are becoming cooler at some stations.
However, the cooling trends are not evident at all stations. For example, no changes are found in the climate records of five stations (see Figure 1.25). Similar to the stations discussed above, the years 1973, 1983 and 1991 stand out as particularly warm years, while 1976 and 1990 are relatively cooler. A small number of stations show increases in maximum temperatures (see Figure 1.26), and of these, only one trend can be considered significant and this one may reflect an abrupt change introduced by the construction of a nearby highway in the late 1970s.
A decreasing trend in daytime temperatures is predominant in the Selva Lacandona. To investigate the nature of this trend, daily maximum temperature extremes were examined. The number of hot days, defined as those days with maximum temperatures greater than one standard deviation above the mean of the time series, was calculated for each year. The year-to-year fluctuation of extreme days follows the mean maximum temperatures extremely well, but a decreasing trend dominates. Results suggest that the marked downward trend in maximum temperatures is driven by a decrease in the number of extremely warm days. There is a tendency for the number of extreme days per year to decrease, from 50-75 in the early 1970s to about 25 by the early 1990s. In contrast, there are no significant decreases in the number of hot days at three stations with no maximum temperature trends. The number of extreme hot days appears to be quite variable, with peaks and dips that correspond closely to the annual maximum temperature average.
Minimum Temperatures
Just as maximum temperatures represent the daytime energy balance and circulation, minimum temperatures portray nighttime processes, typically in the early hours of the day. In marked contrast to the decreasing trends evident in maximum temperature records, minimum temperatures in the Selva Lacandona show an increasing trend at five stations. In three of these five stations, the trends are significant, whereas the other two are notable but non-significant. In contrast, six stations show no significant changes in minimum temperatures. The relationship among annual averages at the six stations appears to be generally consistent, with warmer than average years between 1978 and 1983, followed by cooler period between 1984 and 1989. An analysis of the daily minimum temperature extremes, calculated as those days with minimum temperatures greater than one standard deviation below the mean, revealed significant trends at only three stations.
There is a tendency for minimum temperatures to increase at some of the stations, but the majority of stations show no change.
In summary, there is a tendency for minimum temperatures to increase at some of the stations, but the majority of stations show no change. Four out of the five stations showing increasing trends also showed decreasing trends in maximum temperature (in one case, a non-significant trend). These stations should thus exhibit dramatic changes in the daily temperature range. Such changes are discussed below.
Daily Temperature Range
One characteristic of tropical climates is that daily temperature variations are larger than seasonal temperature variations. The daily temperature range, calculated as the difference between daily maximum and minimum temperatures, is influenced by proximity to oceans or water bodies, elevation, and cloudiness. Recent evidence shows that land use changes are closely correlated to daily temperature range in temperate areas. Four stations in the study area exhibited highly significant, quite dramatic decreasing trends. All of these stations exhibited a decline in maximum temperatures, and most of them show an increase in minimum temperatures as well. Four stations did not reveal changes in temperature range. Three out of the 18 stations show an increasing daily temperature range, but it is only significant at one.
In order to evaluate multiple interpretations of "local," areas of different sizes surrounding the climate stations were extracted for the analysis of deforestation.
Precipitation
Precipitation is perhaps one of the most interesting variables, as changes in rainfall are frequently associated with deforestation. In this study, annual precipitation totals are highly variable with no clear trends. Between 1971 and 1977, rainfall appears to fall below the long-term average at most of the stations, whereas 1981 can generally be considered a peak year. The precipitation results are thus contrary to a widespread perception that rainfall has decreased in the Selva Lacandona (Arizpe et al., 1996). The discrepancy may in part be related to the date when colonizers entered the region; if they came in the early 1980s, it would not be surprising that short-term variability was perceived as a decreasing trend in rainfall.
Although total annual rainfall does not appear to have changed at these stations, it is possible that either the intensity of individual rainfall events or the number of rainy days per year has changed, and either of these could explain the commonly held notion that rainfall is decreasing. To consider this possibility, the number of extreme rainfall events was calculated for each station but no trends were observed at any of the stations. An analysis of the number of rainy days per year showed no trend at most of the climate stations in the Selva Lacandona, with the exception of four stations which showed highly significant increases, particularly during the dry season.
The patterns of both seasonal and annual changes in the Selva Lacandona can be seen in Table 1.28. This table shows the stations and variables that exhibit statistically significant (0.05 level) changes in climate, or trends which are interesting but not statistically significant (0.10 level). The results indicate that trends are clearly evident in the climate records of some stations and absent at others. The changes were considered in the context of large-scale processes that influence climate variability, such as ENSO. Global trends as might be expected under an enhanced greenhouse effect were also considered, as were local perturbations due to events such as the 1982 eruption of the volcano El Chichon in Chiapas (O'Brien, 1995). These types of influences should appear systemically throughout the region, and do not serve as explanations for the changes or lack of changes observed at the 18 climate stations.
Annual precipitation totals are highly variable with no clear trends, contrary to a widespread perception that rainfall has decreased in the Selva Lacandona.
The Issue of Scale
It was concluded that land use changes around the stations was the most likely influence on the climate. However, the issue of scale was still elusive. In particular, at what scale does deforestation impact the local climate? Deforestation was calculated for circles at a number of spatial scales around each station, and at four directions for each circle. To identify whether there was a relationship between the spatial and directional components of deforestation and the observed trend, a series of scatter plots was created. The plots incorporate information regarding forest loss, forest remaining, and changes in minimum and maximum temperatures at each station . The plots were then visually examined to see whether a relationship between deforestation and climate change emerged.
At the largest scales (15 km, 10 km and 5 km) and smallest scales (0.5 km), the stations exhibiting climatic changes appear to be randomly scattered. Given that a good amount of forest remains in the Selva Lacandona, particularly along ridges, this may not be surprising. A relatively deforested station appears to be heavily forested when considered in the context of the surrounding 10 or 15 kilometers. The scale of deforestation that influences the climate appears to be much more local, but not as local as the 0.5 km scale. Within this small area, most of the stations show little forest remaining.
The scale that seems to be most significantly related to trends in the climate record is captured by both the 1-km and 3-km circles. For the 1-km full circles, stations that showed forest loss and had less than 70 percent remaining forest cover often showed a decrease in maximum temperatures. Stations that showed afforestation, yet less than 80% remaining forest cover, also reveal such trends. At the 3-km scale, the stations that showed decreases in maximum temperatures or increases in minimum temperatures were generally deforested, with a remaining forest cover below approximately 85 percent.
There were, however, a large number of stations that showed a forest loss, yet no evidence of climatic change. To explain these anomalous cases, the directional components of deforestation were analyzed more closely. Each quadrant was examined separately for each spatial scale to discover whether the anomalies could be explained. The northeast quadrant emerged as the one which best explains the anomalies. This is not surprising, given the important climatic role of the northeast trade winds.
Deforestation in the 1 km and 3 km northeast quadrants thus helps to establish a relationship between deforestation and climate change in the Selva Lacandona. At the 1-km scale, the northeast quadrant shows a large cluster of stations with a relatively large percentage of forest cover remaining, including four stations that did not show significant climate trends. When the 3 km circle is considered, five stations show a relatively large percentage of forest cover remaining in the northeast quadrant, despite a loss of forest during the period covered by the satellite images. The lack of changes at what otherwise appear to be deforested stations can perhaps be explained by this. The fact that two stations have less remaining forest cover at the 3-km scale can be attributed to the local geography. There are mountain ranges to the northeast of these stations, and the areas behind the mountains have been both colonized and deforested.
The amount of deforestation in the full circle surrounding the climate stations seems to be less important than the location of the clearings.
Conclusions
This analysis suggests that the climatic effects of deforestation are influenced by the land cover at the 1-km to 3-km scale, particularly to the northeast of the station. In other words, the amount of deforestation in the full circle surrounding the climate stations seems to be less important than the location of the clearings. There were, however, some exceptional cases that are not explained by this.
This study does not address the mechanisms responsible for the observed changes. However, decreases in maximum temperatures have been observed in other parts of the northern hemisphere, as well as increases in nighttime temperatures (Karl et al., 1993). It has been hypothesized that these observations are related to changes in cloud cover. The results presented here suggest that landscape heterogeneity may be an important factor in determining the climatic impacts of deforestation. Indeed, the Selva Lacandona is a region of varied terrain, as well as varied land use practices. Forest regrowth plays an integral part in the land use dynamics in the area, resulting in a deforestation process that is neither linear nor concentrated.
These results do not contradict conclusions based on global modeling studies. Recent studies have emphasized that the impacts of deforestation are likely to be regionally specific, and that decreases in temperatures have indeed resulted from deforested simulations (Zhang et al., 1996). However, the results presented here do indicate that the issue of local-scale changes is more complex than the models suggest. There appears to be a minimum spatial extent of deforestation required before local climatic impacts can be discerned in the Selva Lacandona region. However, the magnitude and direction of the climatic trends seem to be influenced by the scale, distribution and geographic location of the clearings. Slope, aspect and terrain may also be important determinants of the climatic effects of deforestation. Different biophysical processes appear to dominate at different scales, and in different tropical regions, varying the nature of positive and negative feedbacks.
This research demonstrates that the nature of environmental problems is not one that can be addressed uniquely by models, satellite imagery, or aggregate national or global data sets. Neither can it be addressed by examining only microscale studies, local data or small-scale surveys. Instead, environmental change research demands an integrated approach that recognizes the complexity of scale, as well as the importance of analyzing data at a number of scales. Environmental research has matured to a point where generalizations can no longer go unchallenged, and extrapolations can no longer suffice for an understanding of the impacts between deforestation and climate change across scales.
These results do not contradict conclusions based on global modeling studies, but they do indicate that the issue of local-scale changes is more complex than the models suggest.
References
Arizpe, L., F. Paz, and M. Velazquez. 1993. Cultura y Cambio Global: Percepciones Sociales Sobre la Desforestacion en la Selva Lacandona. Mexico City: Miguel Angel Porrua. Translated in 1996: Culture and Global Change: Social Perceptions of Deforestation in the Lacandona Rain Forest in Mexico. Ann Arbor: The University of Michigan Press.
Avissar, R. 1995. Scaling of Land-Atmosphere Interactions: An Atmospheric Modelling Perspective, Hydrological Processes, 9:679-695.
Karl, T. R., P. D. Jones, R. W. Knight, G. Kukla, N. Plummer, V. Razuvayev, K. P. Gallo, J. Lindseay, R. J. Charlson, and T. C. Peterson. 1993. Asymmetric Trends of Daily Maximum and Minimum Temperature, Bulletin of the American Meteorological Society, 74(6):1007-1023.
Karl, T. R., and C. N. Williams, Jr. 1987. An Approach to Adjusting Climatological Time Series for Discontinuous Inhomogeneities, Journal of Climate and Applied Meteorology, 26:1744-1763.
O'Brien, K. L. 1995. Deforestation and Climate Change in the Selva Lacandona of Chiapas, Mexico. Unpublished Ph.D. Dissertation, The Pennsylvania State University.
O'Brien, K. L. 1998. Sacrificing the Forest: Environmental and Social Struggles in Chiapas. Boulder: Westview Press.
Pielke, R. A., and R. Avissar. 1990. Influence of Landscape Structure on Local and Regional Climate, Landscape Ecology, 4(2/3):133-155.
Raupach, M. R., and J. J. Finnigan. 1995. Scale Issues in Boundary-Layer Meteorology: Surface Energy Balances in Heterogeneous Terrain, Hydrological Processes, 9:589-612.
Zhang, H., A. Henderson-Sellers, and K. McGuffie. 1996. Impacts of Tropical Deforestation. Part I: Process Analysis of Local Climatic Change, Journal of Climate, 9(7):1497-1517.
Zhang, H., K. McGuffie, and A. Henderson-Sellers. 1996. Impacts of Tropical Deforestation. Part II: The Role of Large-Scale Dynamics, Journal of Climate, 9(10):2498-2521.
Environmental change research demands an integrated approach that recognizes the complexity of scale, as well as the importance of analyzing data at a number of scales.
Forward to Next Section//Back to Table of Contents//AGCI Homepage//Comments: agcimail@agci.org