Eriastrum_harwoodii species distribution models, historical and mid-century projections

Eriastrum_harwoodii species distribution models. The UCSB Biogeography Lab used Maxent to generate predictions of habitat occupancy for 65 species for the CA Energy Commission’s project “Cumulative Biological Impacts Framework for Solar Energy in the CA Desert”, 500-10-021. Species distribution models were produced at 270 m resolution using a subset of 22 environmental variables. Models were evaluated with 10-foldcross validated AUC scores. Results are preliminary and have not yet been reviewed by expert biologists. Continuous Maxent score surfaces for each species modeled can be obtained by contacting Frank Davis (fwd@bren.ucsb.edu). The data provided here are consolidated binary presence/absence layers depicting predicted suitable habitat that were derived using the equal training sensitivity and specificity threshold. This dataset combines binarized outputs from 7 models that differ in terms of climate inputs. The base distribution model (hst8110) is for the period 1981-2010. That model is used with historical climate data to hindcast the species distribution for the period 1951-1980 (hst5180), and to forecast the species distribution for the period 2040-2069 based on "business-as-usual" emission scenarios (AR4 A2, CMIP5 rcp 8.5) and statistically downscaled outputs of 5 different global climate models. The Goddard Fluid Dynamics Laboratory (gfdl_A2_40) and Parallel Climate Model (pcm_A2_406) are from the IPCC's 4th assessment report (AR4) and are included because these models were widely used for the most recent California climate assessment. More current models from the 5th assessment Coupled Model Intercomparison Project (CMIP5) include the Community Climate System Model (ccsm4_rcp8), the Flexible Global Ocean-Atmosphere-Land System model (fgoals_g2__) and the Institut Pierre-Simon Laplace model (ipsl_cm5_). Habitat is modeled as suitable (1) or unsuitable (0) based on each modeled climate. The niche mask simply indicates whether the cell is within the buffered DRECP study region in California (1) or falls outside of California (0).