Investigating Carbon?Water Dynamics in Hemiboreal Forests through Hydrological Modeling and Net Ecosystem Exchange Data

Código

XXVI-SBRH0175

Autores

felipe bortolletto civitate, EMILIO GRACILIANO FERREIRA MERCURI

Tema

F - Processos Hidrológicos e Monitoramento Integrado de Recursos Hídricos

Resumo

This study investigates carbon and water dynamics in hemiboreal forests, focusing on the Järvselja region in Estonia and particularly on the forested Kalli catchment. The objective was to quantify water and carbon cycles and evaluate the interaction between water availability and Net Ecosystem Exchange (NEE). Two hydrological models were employed: the conceptual GR4J-Cemaneige model and the machine learning-based Long Short-Term Memory (LSTM) network. Both were applied to the Reola catchment, which provided streamflow data for calibration and validation, and then regionalized to the ungauged Kalli catchment.To interpret LSTM?s internal states, a diagnostic technique (or "probe") based on Support Vector Machines was implemented to assess the model?s ability to represent physical processes, specifically soil moisture. NEE data, obtained via eddy covariance and processed through EddyPro, were analyzed alongside the basin?s water balance (?S = Precipitation ? Evapotranspiration ? Simulated Flow).In Reola, both models performed well (NSE > 0.80; r > 0.92), with LSTM slightly outperforming GR4J in most metrics. In Kalli, LSTM better captures seasonal flow patterns. The soil moisture probe showed strong temporal agreement with ERA5, SMAP, and water level sensor data, despite magnitude discrepancies. NEE analysis revealed seasonal patterns, with the forest functioning as a carbon sink during spring/summer and a source or neutral during autumn/winter.An increasing NEE trend (weaker carbon uptake) was associated with declining water availability, underscoring the coupling between hydrology and ecosystem carbon fluxes. This work demonstrates that integrating physical and machine learning models enhances our understanding of ecohydrological interactions under climate variability.