Data and code for conditional projections of cladoceran biomass in Poyang Lake

Description This repository contains research data and Python code associated with the manuscript entitled “Applying machine-learning models to conditional projections of cladoceran biomass in Poyang Lake under climate and nutrient-management scenarios”, submitted to Ecological Informatics. 1. Research data The study uses 408 seasonal observations collected from 17 monitoring sites across five main-lake zones of Poyang Lake, China, during 2019–2024. The dataset includes cladoceran and copepod biomass, water-quality parameters, meteorological variables, and hydrological measurements. 2. Machine-learning models The Python code supports data preprocessing, model development, cross-validation, independent testing, and model interpretation. Nine machine-learning algorithms were evaluated using leave-one-year-out cross-validation over 2019–2023, with observations from 2024 reserved for independent evaluation. Gradient boosting was selected for diagnostic analysis, while extremely randomized trees were selected for conditional scenario projections. SHAP analysis was used to assess predictor importance. 3. Future climate and nutrient-management scenarios Future meteorological inputs for 2025–2050 were constructed using 23 global climate models from NASA NEX-GDDP-CMIP6 under four Shared Socioeconomic Pathways (SSPs). The scenario framework incorporates low, stable, and high copepod-biomass sensitivity trajectories and three nutrient-management scenarios, with prescribed nitrogen and phosphorus reductions. The associated Python scripts support future predictor construction, scenario simulations, uncertainty quantification, and sensitivity analyses of nutrient-reduction magnitude and implementation timing. 4. Reproducibility The archived materials are intended to support the verification and reproduction of the study's data preprocessing, machine-learning evaluation, model interpretation, and conditional ecological scenario analyses. The climate projections are based on the publicly available NASA NEX-GDDP-CMIP6 dataset. Users should consult the original data provider for the underlying climate-model products. 5. Associated publication Yi Weixin et al. Applying machine-learning models to conditional projections of cladoceran biomass in Poyang Lake under climate and nutrient-management scenarios. Manuscript under revision at Ecological Informatics.