Seminar: Deep Prakash Sarkar

Recent development of global SAPFLUXNET dataset which contains sapflow data and meteorological variables along with key plant information, opens opportunity to better understand tree water use from ground up approach. So far, most of the available global datasets are either physical model based or partitioned from evapotranspiration (ET) data. In this study, we tried to understand various uncertanities associated with this dataset and suitable modelling approach to generalize sapflow based on meteorological drivers and tree characteristics. To better explain spatial and temporal variability in sapflow, a 2-model approach is employed, where one model predicts temporal dynamics and another model predicts spatial variability then both the model outputs are combined to construct actual sapflow prediction. This approach improved performance and generalization of sapflow. Improvement in temporal dynamics is significant. However, the spatial model is limited by tree microclimate and more tree level information. This unique modelling approach helps us minimize random biases related to each site based on how sensors are installed and its effects on individual tree sapflow in temporal model. This modelling approach can be adopted for upscaling sapflow and better understand plant water use in various ecosystems.