Sap flow modelling based on global radiation and canopy parameters derived from a digital surface model

Abstract

Sap flow represents water transport from roots to leaves through the xylem and is used to describe tree transpiration. This paper proposed and tested a procedure to estimate sap flow by calculating global radiation in a digital model of the tree canopy surface obtained by unmanned aerial vehicle imaging. The sap flow of nine trees was continuously measured in the field. In the digital surface model, individual canopies were automatically delineated, their parameters were determined and the global radiation incident on their surface on specific days was calculated. A polynomial relationship was found between sap flow and the calculated incident solar radiation during the morning hours with a coefficient of determination of 0.98, as well as a linear relationship between the decrease in radiation and sap flow during the afternoon with a correlation coefficient of 0.99. Using the Random Forest machine learning method, a model predicting the sap flow of the trees was created based on the global radiation and canopy parameters determined from the digital surface model of tree canopies. The resulting model was deployed on additional days and compared to field measurements of sap flow, achieving a correlation coefficient of 0.918. In addition, two linear regression models were created for a tree group, achieving coefficients of determination of 0.66 and 0.90.

Keywords

• gis
• random forest
• remote sensing
• solar radiation
• transpiration
• unmanned aircraft vehicle (uav)