A study case on the upscaling of tree transpiration in Water Limited Environments

Direct methods are effective for estimating transpiration in areas where the water balance and vegetation characteristics are highly variable. Usually, transpiration is calculated as a component of catchment evapotranspiration in spatial modelling of water resources. This leads to uncertainty about the actual values of plant transpiration and soil evaporation. Nevertheless, this information is important for conservation and planning in sub-humid and arid areas where vegetation can take up a significant amount of water. In this study, transpiration was directly estimated, based on data from a sap flow and leaf ecophysiology measurement campaign conducted in north-western Spain, in a semi-arid catchment where Quercus Ilex and Quercus Pyrenaica are the dominant species. Estimations obtained with the Penman-Monteith model were compared with sap flow transpiration estimations at the tree, stand and catchment levels. In order to identify tree species throughout the catchment, two methods of classification were performed on a Quickbird satellite image. A procedure based on supervised classification and object-oriented analysis proved to have greater accuracy (0.80 overall accuracy, Kappa = 0.31, true accuracy between 0.75 and 0.84 at p 0.05) than a multi-criteria method using GIS (0.79 overall accuracy, Kappa = 0.23, true accuracy between 0.74 and 0.83 at p 0.05). Sap flow measurements were scaled up using sapwood area and canopy area, while Penman-Monteith estimates were scaled up using Leaf Area Index (LAI) and the fraction of sunlit and shaded parts of the canopy. At the tree level, PenmanMonteith estimations were closer to sap flow estimations when LAI was applied using a “big leaf” model (RMSE = 296.2 cm/h for Q. Pyrenaica and 542.6 cm/h for Q. Ilex, as compared with sap flow values) than using a simple 4-layer representation of the canopy (RMSE = 430.2 cm/h for Q. Pyrenaica and 8543.9 cm/h for Q. Ilex). For a 1 Ha stand with 19 trees, the estimated transpiration was 0.148 m/h with sap flow measurements, and 0.138 m/h with the Penman-Monteith model. Finally, at the catchment level, the sap flow estimation of transpiration was 4.94 m/h and 11.92 m/h with the Penman-Monteith model. Differences arise mainly from the assumptions and generalizations taken for both estimations. It is concluded that simple, direct methods (e.g. sap flow measurements, “big leaf” upscaling procedures) are effective for transpiration estimation, but the generalizations and assumptions associated with them are an obstacle for validation.