Mapping pre-European settlement vegetation at fine resolutions using a hierarchical Bayesian model and GIS

Hong S. He, Daniel C. Dey, Xiuli Fan, Mevin B. Hooten, John M. Kabrick, Christopher K. Wikle, Zhaofei Fan
2006 Plant Ecology  
In the Midwestern United States, the General Land Office (GLO) survey records provide the only reasonably accurate data source of forest composition and tree species distribution at the time of pre-European settlement (circa late 1800 to early 1850). However, GLO data have two fundamental limitations: coarse spatial resolutions (the square mile section and half mile distance between quarter corner and section corner) and point data format, which are insufficient to describe vegetation that is
more » ... ntinuously distributed over the landscape. Thus, geographic information system and statistical inference methods to map GLO data and reconstruct historical vegetation are needed. In this study, we applied a hierarchical Bayesian approach that combines species and environment relationships and explicit spatial dependence to map GLO data. We showed that the hierarchical Bayesian approach (1) is effective in predicting historical vegetation distribution, (2) is robust at multiple classification levels (species, genus, and functional groups), (3) can be used to derive vegetation patterns at fine resolutions (e.g., in this study, 120 m) when the corresponding environmental data exist, and (4) is applicable to relatively moderate map sizes (e.g., 792 · 763 pixels) due to the limitation of computational capacity. Our predictions of historical vegetation from this study provide a quantitative and spatial basis for restoration of natural floodplain vegetation. An important assumption in this approach is that the current environmental covariates can be used as surrogates for the historical environmental covariates, which are often not available. Our study showed that terrain and soil covariates least affected by past natural and anthropogenic alternations can be used as covariates for GLO vegetation mapping.
doi:10.1007/s11258-006-9216-2 fatcat:i2noptu6ojcmfhkuhrumaf6y4y