Skip to content
DSML.KZ
Publications

Dataset · Zenodo · August 20, 2026

Data for "Detecting tipping thresholds in U.S. tidal wetlands using machine learning"

Nail Beisekenov, Rusty A. Feagin, Claire Legaspi, Wenzhe Jiao, Raymond G. Najjar, Maria Herrmann

Nail Beisekenov
Open via DOI

Abstract

This dataset supports the manuscript titled "Detecting tipping thresholds in U.S. tidal wetlands using machine learning," submitted to Ecological Indicators. The study investigates ecological tipping points in tidal wetlands by linking long-term satellite-derived gross primary productivity (GPP) dynamics with manually interpreted historical structural change. Using data from 2000 to 2025, we identify GPP-derived indicators and quantitative thresholds associated with historically tipped tidal-wetland pixels across the conterminous United States. The analysis includes 157,417 MODIS 250 m tidal-wetland pixels, of which 3,790 were manually labeled as tipped or not tipped. The archive contains the manually labeled reference dataset, GPP-derived predictor variables, SHAP values, predictor-threshold results, fixed-holdout predictions, regional summaries, NOAA sea-level comparison data, full GPP and STL residual time series, model outputs for all analyzed pixels, and the analysis code used to reproduce the main CatBoost classification workflow. These data support the identification and spatial mapping of historical tidal-wetland tipping and provide a foundation for future analyses of ecological tipping points, early-warning signals, and forecasting of wetland degradation. Detailed file descriptions, variable definitions, units, sampling information, and data conventions are provided in the accompanying README file.

Coastal wetland ecosystem dynamicsWetland Management and ConservationEcosystem dynamics and resiliencetidal wetlandsecological tipping pointsgross primary productivity

More by Nail Beisekenov