Long-Term Event-Based Rainfall Partitioning Dataset from an Atlantic Forest Catchment, Brazil (2012-2023)

This dataset contains long-term, event-based measurements of rainfall partitioning in a seasonal Atlantic Forest catchment located in southeastern Brazil. The data were collected to quantify rainfall interception, effective precipitation, and their spatiotemporal variability under natural forest conditions. The dataset covers a continuous monitoring period from September 2012 to February 2023 and includes 744 rainfall events measured across 32 spatially distributed monitoring points. Observations were obtained within a secondary montane Atlantic Forest remnant, characterized by heterogeneous canopy structure and pronounced seasonal rainfall regimes. Dataset contents The repository includes the following data files: Gross (external) precipitation measured using a tipping-bucket rain gauge installed in an open area. Throughfall measurements collected at 32 under-canopy points using funnel-type collectors. Stemflow measurements obtained from instrumented trees representative of each monitoring point. Georeferenced coordinates of all monitoring points, enabling spatial analyses of interception heterogeneity. Rainfall events were defined as precipitation periods separated by at least 12 consecutive hours without rainfall. All events with measurable precipitation were included to ensure full representation of rainfall conditions, from small convective storms to extreme events. Data processing and quality control Quality control procedures included: Physical constraints on effective precipitation coefficients, Event-based consistency checks, Outlier detection using an Isolation Forest algorithm stratified by monitoring point and calendar month. After quality control, 21,057 valid event–point observations were retained. Potential applications This dataset is suitable for: Ecohydrological and hydrological process studies, Event-scale analysis of rainfall interception and effective precipitation, Development and validation of empirical, physically based, and machine-learning models, Studies on spatial heterogeneity and temporal nonstationarity of forest hydrological processes, Model benchmarking and reproducibility in tropical forest environments. Data usage notes The data are provided at the event scale and are intended for scientific and educational use. Users are encouraged to cite the associated Zenodo record and the corresponding scientific publication when using the dataset.