Dataset and Code: Understanding and predicting forest mortality in the western United States using long-term forest inventory data and modeled hydraulic damage

The files contain the dataset and code associated to the publication <b>"Understanding and predicting forest mortality in the western United States using long-term forest inventory data and modeled hydraulic damage" </b>by M. D. Venturas, H. N. Todd, A. T. Trugman and W.R.L. Anderegg.<br>The folders and files provided are:<br><b>(1) Folder “Matlab_code”:</b> This folder contains the Matlab scripts used for extracting, analyzing and plotting the manuscript data. The scripts are annotated in order to facilitate their interpretation and utilization. Some of the scripts we would like to highlight are:• “GLM_analysis.m” performs the generalized linear model analysis.• “MASTER_SP_PROGRAM” generates the tables with FIA data, generates hourly weather data, extracts climatic predictors, joins tables with the hydraulic model outputs, and generates the tables used for data analysis.• “Figure*.m” these scripts generate the figures presented in the manuscript and supporting information.<br><b>(2) Folder "RFs":</b> This folder contains the R-code ("RF_analysis.R") for performing the random forest regression analysis plus the sub-folders "RF_no_collinear_vars" and “RF_ouput”.• The sub-folder "RF_no_collinear_vars" contains the data files per species (all, aspen, coloradopinyon, douglasfir, lodgepole, oneseedjuniper, ponderosa, singleleafpinyon, utahjuniper) and simulation (NAS, NAB, AS, AB, AEE) only containing non-collinear variables (retaining only one predictor variable among those with a r=&gt;0.8) and the response variable (MortPerct). These files are named “species_no_collinear_vars_simulation.csv”.• The sub-folder “RF_output” contains four files with the main outputs from the random forest analysis. “Best_RF_species.csv” shows per species which was the simulation with the highest variance explained. “RF_comparison_simulations.csv” shows the variance explained for the random forest constructed for each species and simulation. “RF_comparison_simulations.txt” provides the details of the random forest constructed for each species and simulation, including the importance of the factors selected as mortality predictors. “RF_importance_graphs.pdf” contains per species and simulation a bar plot with the relative importance of factors selected as mortality predictors and a scatter plot comparing observed versus predicted mortality for out of bag samples.<br><b>(3) Folder “Stand_data”:</b> This folder contains the United States Forests Inventory and Analysis (FIA) plot data with their corresponding 5 year climatic and hydraulic model physiological diagnostic variables used for statistical analysis. These files were used for generalized linear model (GLM) regression analyses. Each datafile contains the stands for a given species (all, aspen, coloradopinyon, douglasfir, lodgepole, oneseedjuniper, ponderosa, singleleafpinyon, utahjuniper) and simulation (NAS, NAB, AS, AB, AEE). The files are named “species_simulation.csv”.<br><b>(4) Folder “FIA_DATA”:</b> This folder contains the FIA dataset from where the plot data was extracted for the target species following the selection criteria described in the publication’s Materials and Methods.<br><b>(5) Folder “SperryModel_FIA_AllModes_CodeOnly”:</b> This contains the C++ model code.<br><b>(6) Folder “SperryModel_FIA_CoreConfigFiles”:</b> This folder contains the parameter files required for running the Sperry model. <br>