Challenges and opportunities in synthesizing historical geospatial data using statistical models

摘要: Abstract We classified land cover types from 1940s historical aerial imagery using Object Based Image Analysis (OBIA) and compared these maps with data on recent cover. Few studies have used kinds of to model drivers change, partly due two statistical challenges: 1) appropriately accounting for spatial autocorrelation 2) modeling percent which is bounded between 0 100 not normally distributed. studied the change in woody at four sites California's North Coast (1948) (2009) high resolution imagery. eCognition Developer aggregated resulting scale a Digital Elevation Model (DEM) order understand topographic change. Generalized Additive Models (GAMs) quasi-binomial probability distribution account boundedness variable. explored relative influences current variables (grouped principal component analysis) reflecting water retention capacity, exposure, within-site context, as well geographical coordinates. estimated models pixel sizes 20, 30, 40, 50, 60, 70, 80, 90, 100 m, both tree neighborhood scales stand scales. found that had consistent positive effect cover, autoregressive term was significant even after controlling Specific emerged important different scales, but no overall pattern across or any we tested. This GAM framework flexible could be more variables, relationships predictor larger Modeling ecology sources can valuable way plan restoration enhance ecological insight into landscape