A standardized, sequential, physical stratification of the earth into a set of unique physical environments with their associated biota is the basis of the geospatial ecosystems methodology. This approach is framed on the understanding that the biodiversity that occurs in any area is largely the result of a biotic response to physical environmental potential and local dynamic processes, and that unique physical settings tend to give rise to unique assemblages of biodiversity. Implementation of this modeling approach are based on the mapping and integration of fundamental ecosystem structural elements, including land surface forms, surficial lithology, bioclimates, topographic moisture potential, and associated land cover as a proxy for vegetation. This methodology was refined and tested during the completion of three continental ecosystems mapping pilot efforts in South America, the conterminous United States, and Africa.
The first implementation of this methodology occurred during a Nature Conservancy and NatureServe collaboration to model ecosystem distributions for South America. It was based on the acquisition or generation of five base layers - bioclimate (WorldClim, 2004 and TNC, 2005), landform (SRTM, 2000 and WWF-Hydrosheds, 2005), landcover (GLC, 2000), elevation (SRTM, 2000), and geology (TNC and GDS, 2005) - followed by the integration of these layers into a set of unique ecological attributes. This resulted in the identification of 659 local-scale terrestrial ecosystems for South America at the finest spatial resolution (450 meters) ever attempted for the continent. This work is described in detail in Sayre et al., 2008 (Chapter 9 in http://www.aag.org/cs/publications/special/nalcs)
Conterminous United States
Subsequently, the ecosystems mapping methodology was used during a USGS effort to model the distribution of terrestrial ecosystems for the conterminous United States. Four major structural components of ecosystems – isobioclimates, land surface forms, topographic moisture potential, and surficial lithology - were modeled and then spatially combined into a set of unique biophysical settings. This work is described in detail in Sayre et al., 2009 (https://pubs.er.usgs.gov/publication/pp1768), and is summarized below.
Bioclimate information for the model was provided by implementing the Rivas-Martínez global bioclimatic classification system that quantifies key bioclimatic indices reflective of vegetation distributions. Implementation of this methodology, using the Daymet temperature and precipitation data, resulted in the generation of four climate layers for the conterminous United States: macroclimates, bioclimates, thermotypes, and ombrotypes. However, since the biophysical stratification approach required a single climate layer it was determined that the best choice to achieve the required climate variations would be a combination of the thermotypes and ombrotypes. Combination of the ombrotypic regions (dry/wet gradients) with the thermotypic (warm/cold) regions resulted in 127 unique combinations.
Land surface forms for the conterminous United States were based on the Missouri Resource Assessment Partnership (MoRAP) methodology that uses slope and local relief. After applying this methodology to the USGS 30-meter National Elevation Dataset (NED) eight land surface classes were generated. However as part of the USGS implementation the original low mountains class was than redefined into two separate classes, low mountains or high mountains/deep canyons. As a final step, an additional class of drainage channels was derived independently to identify wet and dry river channels. This resulted in a final set of 10 land surface forms classes: flat plains, smooth plains, irregular plains, escarpments, low hills, hills, breaks/foothills, low mountains, high mountains/deep canyons, and drainage channels.
During this implementation of the methodology an additional base layer, topographic moisture potential, was added to help contribute substrate moisture regimes into the ecosystems model. The four topographic moisture potential classes – periodically saturated or flooded land, mesic uplands, dry uplands, and very dry uplands - were based on the derivation of ground moisture potential using a combination of computed topographic characteristics (CTI, slope, and aspect) and mapped National Wetland Inventory boundaries.
The surficial lithology classes were derived from a generalization and reclassification of the 28 lithology classes of the USGS map "Surficial Materials in the conterminous United States" into a set of the 18 lithologies that typically control or influence the distribution of vegetation types.
Once the base layers were generated they were spatially combined into a set of 49,168 unique physical environments that characterize the abiotic (physical) potential of the environment at a 30-meter spatial resolution. A minimum pixel count threshold (20,000 pixels) was then applied resulting in a final set of 13,482 classes delineating physically distinct areas as the fundamental structural units ("building blocks") of ecosystems. As a final step this set of classes were aggregated into 419 NatureServe ecosystems using a semi-automated labeling process based on rule-set formulations for the attribution of each ecosystem.
Modeling the terrestrial ecosystems for Africa was the final continental implementation of the ecosystem mapping methodology prior to its global implementation. The Africa approach used three base layers – isobioclimates, land surface forms, and surficial lithology – and source data of differing origin and coarser spatial resolution than the United States effort. This work is described in detail in Sayre et al., 2013 (http://www.aag.org/cs/publications/special/map_african_ecosystems, and summarized below.
African isobioclimates were generated using the spatial algorithms from the United States effort (with minor adaptations) and the Worldclim climatological data. After modeling the ombrotypes and thermotypes for Africa, these two datasets were combined into an isobioclimates map with 157 composite classes.
The Africa implementation of the MoRAP classification used 90-meter elevation source data that was created by void-filling and re-sampling the 30-meter SRTM elevation data provided by the National Geospatial Intelligence Agency. Once generated, the slope and relief classes were subsequently combined to produce the eight MoRAP land surface form classes and then the USGS refinement was applied to identify the additional "high mountains/deep canyons" class. The classes "flat plains" and "smooth plains" were combined into a single class, and "low hills" were merged into "hills". As a result, only seven land surface form classes - smooth plains, irregular plains, escarpments, hills, breaks, low mountains, and high mountains/deep canyons - were identified in the final dataset.
The African surficial lithology dataset is a map of parent materials - a mix of bedrock geology and unconsolidated surficial materials classes. It is a compilation and reclassification of twelve digital geology, soil and lithology databases into nineteen surficial lithology classes delineated based on geology, soil and landform. Due to the varying spatial and classification resolutions of the geologic source data, this surficial lithology map varies in spatial complexity and classification detail across Africa.
The base layers were then combined as part of the biophysical stratification approach, and the resulting composition and distribution of these unique footprints of the physical and biological landscape were reviewed by regional vegetation and landscape ecology experts and attributed (labeled) to an intermediate scale African ecosystem class.