ConservationVision Development Vulnerability Model

The Virginia Development Vulnerability Model quantifies the predicted risk of conversion from "natural", rural, or other open space lands to urbanized or other built-up land uses.

The most recent edition of the model was released in 2022. For more information, see here.

---

Script overview

General-purpose

• Helper.py: a set of general-purpose functions, imported by all other python scripts
• BatchDownloadZipFiles.py: downloads a set of zip files from an FTP site

Predictor variable processing

• proc3DEP.py: functions for batch-downloading and post-processing elevation files from the 3DEP program
• procCensus.py: functions for processing U.S. Census data, primarily for developing urban cores from block data
• procConsLands.py: processing for conservation lands data, including predictor variables and the protection multiplier
• procNHD.py: processing for NHD-based predictor variables
• procNLCD.py: processing for most predictor variables from National Land Cover Database data
• procNLCDImpDesc.py: develops cost surfaces and distance to road/ramp predictors from NLCD impervious descriptor (with Tiger-Line roads used as ancillary data).
• procTravelTime.py: runs cost distance (travel time) analyses for urban cores

Model input preparation

• finalizeVars.py: finalizes all predictor variables for the Development Vulnerability Model (clip/mask, convert to integer, and output as a TIF file)
• procSamples.py: processes to create a sampling mask, point samples, attribute those samples with values from raster variables. Also creates development and water masks used to make the final Development Vulnerability model

Random forest model

• RFModel_run.R: creates a new random forest model with specified samples. Includes entire process: variable selection, cross-validation, final model with partial plots
• RFModel_predict.R: functions to run post-model processes for the development vulnerability random forest model, including creating prediction raster(s), adjusting prediction raster using protection multiplier, development, and water, and validating predictions using validation points data set. These functions are called at the end of RFModel_run.R.

Post-processing

• finalizeModelProducts.py: copies raster datasets for the selected model (raw and final scores) to geodatabase and TIF rasters, updating their metadata from template files.
• outputs/Report_Tables_Figures.R: generates tables and figures for final technical report. Includes predictor variable table and figures for variable importance, partial plots, and PRC/ROC curves for cross-validation and independent tests.