Data and code associated with: Storms are an important driver of change in tropical forests

Tropical forest dynamics and composition have changed over recent decades, but the proximate drivers of these changes remain unclear. Investigations into these trends have focused on increasing drought stress, CO2, temperature, and fires, whereas convective storms are generally overlooked. We argue that existing literature provides clear support for the importance of storms as drivers of forest change. We reanalyze the largest plot-based study of tropical forest carbon dynamics to show that lightning frequency – an indicator of storm activity – strongly predicts forest carbon storage and residence time, and its inclusion improves model fit and weakens evidence for effects of high temperatures. Convective storm activity has increased 5-25% per decade over the past half century. Extrapolating from historic trends, we estimate that storms likely contribute ca. 50% of the reported increases in biomass mortality across Amazonia, with all realistic combinations of assumptions indicating a possible range of 12-118%. Spatial variation in storm activity shows weak relationships with drought, demonstrating that forests can experience high drought stress, high storm activity, or both. Accordingly, we hypothesize that convective storms are amongst the most important drivers tropical forest change, and as such, they require significant research investment to avoid misguiding science, policy, and management.

File and folder list, includes:

amazon.zip: This folder contains all of the data for producing maps of storm activity and drought stress across the Amazon region and adjacent forests. This includes 4 folders and 3 files (1 csv and 2 tif). The code to run these analyses is in folder "Code_for_Amazon_analyses" along with the metadata for these files.

• climVars: .Rdata files for each month-year combination from 1971-2019. This is an output from scripts/parseClimData.R, which parses the netcdf files and isolates the target variables of CAPE (cape thresholds) and VPD
• figures: figures for this publication, Gora et al., 2025
• mcwd: Evapotranspiration (et_stacks_annual) and precipitation (precip_stacks_annual) rasters stacked annually for calculating Maximum Cumulative Water Deficit. Evapotranspiration (pet_penman) and precipitation (pr) data were downloaded from Chelsa (see below). MCWD was reset annually for each raster cell via reset_mask.tif after reaching the month the focal cell experiences the greatest water deficit. Annual calculated values for MCWD are aggregated across our study period in MCWD_precip_reset_2009_2019.tif
• netcdf: Climate variables downloaded from Chelsa and ERA5 via the NCAR repository (for more details on the data sources and specifics about each variable, see the ReadMe file for this project)
• processedClimVars1990.csv: aggregated csv with one value of each climate variable for each pixel in the study area, averaged across the timeframe of 1990-2019. This is an output from scripts/parseClimData.R
• analysisRastTemplate.tif: raster template that we use as the base gridded structure to visualize the data. This is created using the extent of the Feng et al., 2023 Amazon shapefile and the pixel resolution of the ERA5 data
• processedClimVars1990_2019.tif: aggregated raster with each pixel containing one value for each climate variable, averaged across the timeframe of 1990-2019. This is an output from scripts/parseClimData.R

Code_for_Amazon_analyses.zip: Repository with code for parsing climate data and creating figures of storm activity and drought stress across the Amazon. Contains the complete Github code repository as of 14 May 2025. (see Github reference below).

• scriptsPublication: 5 R files
• dataSources.md - contains references and notes on all data sources
• metadata.md - list of amazon folder contents, descriptions for all climate variables, and methods of calculation
• README.md

storm_contributions: This folder contains the R script for estimating the contributions of increased storm activity to increases in biomass mortality across the Amazon over the past several decades. This folder only includes 1 R code file.

• increased_storm_contributions_submitted.R

Sullivan_etal2020_reanalysis: These files are for the reanalysis of global forest biomass carbon originally published in Sullivan et al. 2020. "Long-term thermal sensitivity of Earth's tropical forests." Science. This folder contains 3 files (1 csv, 1 R file, 1 txt file),

• ENTLN_for_Sullivan-etal.csv
• Gora_Sullivan2020_reanalysis_submitted.R
• ReadMe_reanalysis_Sullivan-etal2020.txt