Included on this page are the code and data used in our analysis "Agreement and uncertainty among climate change 
impact models: A synthesis of sagebrush steppe vegetation predictions"

The resources included in this repository are:

Contents
	draft2_zimmer_et_al_2020.pdf - second draft of final manuscript
	draft2_zimmer_et_al_2020_figures.pdf - second draft of final manuscript figures
  	draft2_zimmer_et_al_2020_supplement.pdf - second draft of supplemental figures

   Analysis - Zipped folder with data and code 
	(NOTE: File paths in the code are written for use with the included R Project. To use, open the analysis.Rproject file,
		 and open scripts through the /scripts folder.)

	raw_images - folder with raw images of model results incoporated in our analysis
	georeferenced_rasters - rasters of model results included in our analysis, after georeferencing (done in ArcMap, no script available)
	classified_rasters - rasters of model results included in our analysis, after georeferencing 
		             and unsupervised classification
	recoded_rasters - rasters of model results included in our analysis, after georeferencing, 
		          unsupervised classification, and recoding values to values indicating increases, 
		          decreases, and no change in vegetation
	masked_rasters - rasters of model results included in our analysis, after georeferencing, 
		          unsupervised classification, recoding values, and eliminating pixels not overlapping
			 BLM lands or the Intermountain West
	count_stacks - raster stacks corresponding to counts of models projecting increases, decreases, or no change
			pixel-by-pixel. These stacks are made for each vegetation type, and for all emissions scenario results, 
			low emissions results, and high emissions results
	renwick_supp_shp - supplemental CSV data from the renwick study, converted into shapefile results
	data - folder with additional data used in analysis
			renwick_supp.csv - CSV of supplemental results from Renwick et al 2018 used in analysis
			study_metadata.csv - CSV of important metadata in reference to the studies and individual 
					     models included in our analysis
			gis - folder of gis layers used in analysis. These include BLM land, ecoregions and states
       scripts - folder of R scripts/code used in analysis
		1_make_shp_from_renwick_results.R - takes the data from renwick_supp.csv and makes it into shapefiles which can be 
				analyzed like other results
		2_unsupervised_classification.R - takes the georeferenced_rasters and performs an unsupervised
				classification to identify similar pixel groups
		3_recoding_rasters.R - recodes the values in the classified_rasters to correspond to increases, decreases, or no change
		4_mask_rasters.R - takes the recoded_rasters and eliminates areas not corresponding to Intermountain West BLM lands
		5_make_stack_all_emissions.R
		6_make_stack_low_emissions.R
		7_make_stack_high_emissions.R
			The above three scripts process the masked all/low/high emissions scenario results for each species, resample
			them, then count the number of models indicating increases, decrease, or no change at each pixel. Then
			saves out a stack for each of these
		8_plotting_rgb_count_withlegend_all_emissions.R 
		9_plotting_rgb_count_withlegend_low_emissions.R
		10_plotting_rgb_count_withlegend_high_emissions.R
			The above three scripts take stacks of counts corresponding to the number of models
				indicating increases, decreases, and no change among all/low/high emissions scenarios. Then creates an
				RGB plot of those, and a legend for each species.

		supplement_rgb_legend.R - Makes a supplemental plot to show a simple example of where points with various values plot on 
			the triangle RGB legned included in plots.
		supplement_only_renwick_pixels.R - Evaluates sagebrush projections only at pixels which correspond to the
			Renwick et al study. Included in supplemental material.		

The order of this analysis is:

0. Georeference the "raw_images". This was completed in ArcMap, no script is available. 


---- Steps 1-10 are completed in R. Open the analysis.Rproject file and open the following scripts from the /scripts folder. ---


1. One set of results came from a csv, which was converted to shapefiles for spatial analysis. This is completed by the script 
	"1_make_shp_from_renwick_results.R"

2. Perform unsupervised classification on georeferenced_rasters to identifty similar pixel groups within images, using script
   	"2_unsupervised_classification.R"

3. Recode the values of classified_rasters. Classification gives the pixel groups arbitrary values. Recoding the values
	to make them meaningful is necessary. We recoded pixels corresponding to decreases in  vegetation as -1, pixels 
	corresponding to increases as 1, pixels corresponding to no change as 0, and pixels not addressing vegetation 
	(irrelevant background, legends, etc) as N/A. The recoding script is "3_recoding_rasters.R"

4. Mask the rasters. In this analysis, we were interested only in Intermountain West BLM lands, so pixels not 
	overlapping BLM lands in the Intermountain West were removed by masking. The masking script is "4_mask_rasters.R"

5-7. Using the masked rasters, resample and stack rasters which all address a given vegetation/emissions scenario combination.
	Then count the number of pixels indicating increases, decreases, or no change in that vegetation type at every pixel. 
	Save out the stack of pixel counts.

8-10. Make plots of RGB intensity corresponding to count of models indicating increases, decreases, and no change, using the stack
	of pixel counts. We analyzed all emissions scenarios together, then only low emissions scenarios, then only high. 

11. Manually merge together RGB plots and legends in a program such as Inkscape (no script available).