Ecological systems, defined as spatially co-occuring assemblages of vegetation types sharing a common underlying substrate, ecological process, or gradient, were identified for all of Latin America and the Caribbean (LAC) in a recent classification effort by NatureServe. Field planners in the Conservancy's LAC program are encouraged to utilize this classification as it permits standardization in target selection, goal setting, portfolio assembly, and analysis across LAC ecoregions. The classification work by NatureServe produced a list and description of ecological systems, but the on-the-ground occurrences of these ecological systems had not yet been mapped. This dataset maps the distribution of LAC Ecological Sysetms in South America as described in the NatureServe web page: http://www.natureserve.org/getData/LACecologyData.jsp

Términos de Acuerdo. Al utilizar este shapefile(s), usted (el "Usuario") acuerda a ser limitado por estos Términos de Uso. Si usted no se dispone a los términos de limitaciones y condiciones aquí definidos, por favor no utilice este Shapefile. La Corporación Andina de Fomento (CAF) se reserva el derecho en cualquier tiempo y de vez en cuando a modificar o descontinuar, temporal o permanentemente, cualquier parte o la totalidad de este Shapefile. Se exime a CAF de responsabilidad a cualquier Usuario o Tercer Persona por cualquiera de estas modificaciones, suspensiones o descontinuación. El archivo de metadato forma una parte integral de este Shapefile, y debe ser distribuida y guardada con él. Se prohibe el uso de estos datos para fines comerciales. Su uso para fines de investigación, educación o divulgación está permitido, siempre y cuando se cite la fuente y se dé debido reconocimiento al Programa GeoSUR, la Corporación Andina de Fomento y la Secretaría IIRSA por su elaboración. La cita adecuada es la siguientes. "Mapa digital elaborado por el Programa GeoSUR en colaboración con la Corporación Andina de Fomento y la Secretaría IIRSA". Las áreas son una representación geográfica ilustrativa, aún no se han realizado ajustes a los datos para que respondan a criterios de exactitud. Su uso está muy limitado por la escala de los datos de referencia que sirvieron para la representación de las áreas. No se ofrecen garantias de ningún tipo sobre la disponibilidad permanente de los datos, su exactitud o su utilidad para algún fin específico.

The goal of this project was to map the distribution of LAC ecological systems across South America. Five input data layers - general elevation, general landform, general geology, general bioclimate and general land cover - were combined to produce a map of unique gridcodes (filenames: esa_gridcode and ssa_gridcode). The numeric value of each pixel in the resultant grids was a unique combination of the five input data layers. For example, the unique gridcode 1742020 represented: 1000000 (0 - 500 meters), 700000 (floodplains), 40000 (alluvium), 2000 (tropical pluvialseasonal), and 20 (Tree Cover, Broadleaf Deciduous). These unique gridcodes were considered ecological footprints that, with expert knowledge, could be attributed to a particular LAC ecological system. Each Conservation Program Area was masked from the gridcode grid and converted to a polygon shapefile (filenames: ama_ecosystem_2005_s.shp, atl_ecosystem_2005_s.shp, csa_ecosystem_2005_s.shp and nsa_ecosystem_2005_shp, east_ecosystem_2005_s.shp and west_ecosystem_2005_s.shp). By converting the gridcode grid to a polygon shapefile, pixels with the same unique gridcode that were adjacent to each other were joined together to form a polygon. These unique gridcode polygons were then attributed by selecting the classes that best represented a particular LAC ecological system (e.g. 500 - 1000 meters, plateaus, etc.). Often a system was best represented by selecting multiple classes (e.g. 500 - 1000 meters or 1000 - 2000 meters, plateaus or mountains or hills, etc.). The geographic extent of the selection was reviewed and a subset, based on biogeographic boundaries, was selected and attributed to a particular ecological system (e.g. CES410.139; see Appendix I for a detailed description of each field in the table). This method of attributing the unique gridcodes (searching for specific ecological systems based on their expected elevation, landform, geology, bioclimate and landcover characteristics) allowed multiple gridcodes to be selected and attributed at once, significantly reducing the amount of time it took to attribute all 9352 unique gridcodes that were identified across South America. It was necessary to use a limited number of general classes in each of the input data grids in order to keep the final number of unique gridcodes generated low. This resulted in many gridcodes repeating themselves across the study area. But these gridcodes clearly represent different ecological systems in different places and were appropriately coded based on their biogeographic location. In five areas, Southern Chile, Peruvian Yungas, Bolivian Yungas, Equatorial Pacific Forest, and Chaco, existing ecological system data, produced by in-country programs at expert-derived ecological system mapping workshops, was unioned with the modeled unique gridcode data. Each polygon on the map still possessed a 'modeled' unique gridcode value but the final ecological system linework and attribution was based on the expert-derived map information, instead of the gridcode. The CES codes and ecological system names from these five datasets were used to populate the Ecode and Econame fields. Identifying non-vegetated and degraded vegetation was important in this project because it focused on identifying the extent of existing ecological systems, rather than potential ecological systems. Therefore the general GLC class 50 (intensive agriculture, mosaic agriculture/degraded vegetation, forest plantations, permanent snow/ice and urban) and 52 (mosaic agriculture/degraded forests) were selected from the unqiue gridcodes across South America and their ecode values respectively coded as converted and degraded. GLC class 83 (water bodies) was also selected and coded as water. In the five special areas noted above that were mapped using existing ecological system data, any information about the ecological system code or name was preserved in the comment field before the polygons were recoded as converted, degraded or water.

The spatial resolution, or the size of the smallest feature that was mapped, was a single 450 meter pixel (20 hectares). Some small patch ecological systems were identified at this minimum polygon size. The input datasets were produced at a variety of scales though, some coarser and some finer than 450 meters. The finest resolution data was the landform and elevation grids, produced from the 450 meter DEM (originally derived from 90 meter SRTM DEMs). The GLC landcover data was produced at a scale of 1 km and resampled to 450 meters. Geology data was produced from hardcopy maps ranging in scale from 1:500,000 to 1:2,500,000 (with the majority at 1:100,000 or better). Bioclimate data was produced from 30 second arc (1km) Worldclim temperature and precipitation data (though the scale of the bioclimate data is actually much coarser than 1 km for many parts of South America because the weather stations are so far apart). Some people would argue that GIS data is only as accurate as the worst (smallest scale, least accurate) input data layer, which in this project would be the bioclimate data. But bioclimate was the least important dataset in attributing the ecological systems. It was used as a general guide for confirming the location of different ecological systems, but was understood to be continuous data with poorly-defined transitional zones. The three most important datasets in attributing the ecological systems were landcover, landform and elevation. Landcover was used as the primary criteria for identifying most ecological systems, the selection was then modified by landform and elevation, and then, of lesser significance, by geology and bioclimate. So the spatial resolution of the final ecological systems data based on the most significant input data layer (landcover) would be limited to 1 km (if you believe the argument that limiting scale in GIS is defined by the least accurate dataset). But the nominal working scale of the final ecological system linework was 'improved' by sub-dividing the landcover data (1 km) with the more detailed landform and elevation data (450 meters). But how exactly does 450 meters grid resolution translate into a working scale ratio? Some studies suggest that if the smallest unit of measurement on a map is 10 meters the scale would be 1:20,000 "Generally, a line cannot be drawn much narrower than about 1/2 a millimetre. Therefore, on a 1:20,000 scale paper map, the minimum distance which can be represented (resolution) is about 10 metres" (Scale, Accuracy, and Resolution in a GIS. 1999). Using this guideline, the smallest unit of measurement on the ecological systems map is 450 meters, so the scale would be approximately 1:900,000. A simple visual comparison of the ecological systems linework against Landsat imagery suggests the working scale is somewhere from 1:500,000 to 1:1,000,000. The linework was evaluated against visibly distinct features in the Landsat imagery. By progressively zooming in until the linework started to move off of a feature it was possible to get a rough sense of the accuracy of the data at different scales