Dataset and Analysis Code Supporting "Asymmetric Impacts of Supply- and Demand-Driven Flash Droughts on Leaf Senescence"

Abstract

This repository contains the processed datasets and R scripts supporting the manuscript entitled “Asymmetric impacts of supply- and demand-driven flash droughts on leaf senescence.” The study investigates the asymmetric effects of supply-driven flash droughts, mainly associated with precipitation deficits, and demand-driven flash droughts, mainly associated with enhanced atmospheric evaporative demand, on vegetation leaf senescence across the Northern Hemisphere during 2001–2024. The repository contains the processed datasets supporting the analyses and visualizations. Owing to repository file-size constraints, some processed datasets are provided as ZIP archives. Compression does not alter the data content, and the files should be extracted before use. Phenology_North_poly.csv contains annual satellite-derived vegetation phenological metrics for each grid cell across the Northern Hemisphere and was used for phenological quality control and anomaly calculation. all_GS_FD_years.csv contains the grid-year combinations with at least one FD overlapping the growing season and was used to identify non-FD years for calculating the phenological baseline. dEOS_poly.csv contains the resulting LSD anomalies derived from the enhanced vegetation index (EVI) and solar-induced chlorophyll fluorescence (SIF). The term “EOS” used in the file name and scripts denotes the end-of-season phenological metric and corresponds to the leaf senescence date (LSD) used in the manuscript. phenology_poly_pentad.csv records the long-term mean timing of the start, peak, and end of the growing season, expressed as the corresponding 8-day periods for each grid cell. SMrz_drought_events.csv contains the flash drought events identified from root-zone soil-moisture percentiles using 01_FD_detection.R , including the original P1–P4 event timing indices, minimum soil-moisture percentile, onset rate, and cumulative severity. SMrz_drought_events_mean_environment_anomaly.csv contains flash drought events overlapping the growing season and their mean environmental anomalies calculated over the complete, untruncated P1–P4 event periods. These variables were used to characterize the hydroclimatic conditions of the events and as inputs for their subsequent classification into supply- and demand-driven flash droughts. final_FD_events.csv is the final analytical dataset and contains 1,141,551 event records. It retains the original, untruncated P1–P4 indices and was constructed by (1) retaining FD events that overlapped the growing season, (2) retaining phenological observations within the grid-specific median ± 2.5 median absolute deviations and grid cells with an LSD variance no greater than 30, (3) retaining only grid-year combinations with a single FD overlapping the growing season, and (4) excluding non-target land-cover and biome classes, including croplands, and records with predictor values outside their respective 1st–99th percentile ranges. This dataset was used in the subsequent LSD-response, structural equation modeling, random forest, SHAP, and model-improvement analyses. Meanenv_EarlyGS_anomaly.csv and Meanenv_LateGS_anomaly.csv contain environmental anomalies during the early and late growing seasons, respectively, while Additional_data.csv provides the additional background and vegetation variables used to construct the predictor set for the RF, SHAP, and related analyses. The background spatial datasets include MODIS_IGBP0.1deg_type.csv , which contains MODIS International Geosphere–Biosphere Programme (IGBP) land-cover classifications at a spatial resolution of 0.1°; phe_poly_biome.csv , which contains biome classifications based on the World Wildlife Fund terrestrial ecoregions; and Background_MAP.csv and Background_MAT.csv , which contain mean annual precipitation and mean annual temperature, respectively. worldmap.zip contains the world map and national boundary files used for spatial visualization. The datasets supporting the response and modeling analyses include dEOSsig_supply_poly.csv and dEOSsig_demand_poly.csv , which contain the statistical significance of LSD responses to supply- and demand-driven flash droughts, respectively. RF_compare_supply_demand_matched_EVI.csv and RF_compare_supply_demand_matched_SIF.csv contain the random forest model outputs used to evaluate and compare the simulated LSD responses based on EVI and SIF. FD_ET_series.csv contains the time series of evapotranspiration anomalies before and after flash drought onset. The SHAP outputs from the random forest models are provided in shap_mat_EVI_demand.rds , shap_mat_SIF_demand.rds , shap_mat_EVI_supply.rds , and shap_mat_SIF_supply.rds . These files contain the SHAP values used to quantify and visualize the contributions of climatic, biological, and flash-drought-related variables to the modeled LSD responses under the two flash drought types. The accompanying R scripts perform the principal data processing, statistical analyses, and visualization procedures used in the study. 01_FD_detection.R identifies flash drought events from root-zone soil-moisture percentiles, calculates their characteristics, and applies the prescribed duration criterion of 4–12 consecutive 8-day periods. 02_phenology_extraction.R extracts vegetation phenological metrics from satellite-derived EVI and SIF time series. 03_FD_environment_anomaly.R retains FD events that overlap the growing season without modifying their original P1–P4 indices and calculates environmental anomalies over their complete event periods. 04_phenology_anomaly_nonFD_baseline.R calculates phenological anomalies relative to years without growing-season FDs and applies the grid-specific phenological quality-control and single-event grid-year criteria. 05_final_FD_event_filtering.R generates the final analytical dataset from the records processed in the preceding steps and applies the land-cover, biome, and predictor percentile filters. 06_RF_SHAP_analysis.R conducts random forest modeling and SHAP analysis to identify the climatic and biotic controls of leaf senescence responses. 07_SEM_analysis.R performs structural equation modeling to examine the pathways through which the two flash drought types regulate leaf senescence. 08_phenology_response_significance.R evaluates the significance of phenological responses to flash droughts. 09_visualization_analysis.R generates the principal figures and statistical visualizations, and 10_visualization_SHAP_analysis.R produces the SHAP importance and dependence plots. The original satellite, reanalysis, and ancillary datasets are not redistributed in this repository, except for the processed background and boundary data used for visualization. The original root-zone soil-moisture percentile series and processed LAI time series are not redistributed because of their large file sizes; their data sources and processing procedures are documented in the manuscript and accompanying code. The original datasets are publicly available from their respective data providers, as documented in the associated manuscript. The deposited processed datasets and scripts provide the materials needed to evaluate the reported results and reproduce the principal analyses and visualizations.

Topics & conditions

Detected automatically by Psychedelic Commons from the title and abstract, not labels from the source. They may be incomplete.

Keywords