Water yield modelling, sensitivity analysis and validation: a study for Portugal

Abstract

The spatially explicit assessment of freshwater is key to introduce the ecosystem service (ES) concept into decision-making processes. Many tools are being developed to model water balance and to analyse the effects of meteorological conditions on water ES behaviours at multiple spatial scales. Nevertheless, there is a lack of studies that analyse the sensitivity of these models and estimate the model’s accuracy. The current research uses the InVEST Water Yield Model (WYM) to assess freshwater ES between 1990 and 2018 at River Basin District (RBD) level in mainland Portugal. The methodology included sensitivity analysis simulations to test different parameters of the model over two different sources of meteorological data. We validated the model with the European Environment Agency (EEA) database on the quantity of Europe’s freshwater resources. To evaluate the models’ sensitivity, Pearson’s correlation coefficients and statistical methods were calculated for each simulation. The model which obtained the best performance in the sensitivity tests was retained for further analysis and calibration and its accuracy was assessed by comparing the mean estimated water yield (WY) with mean observed values for 2018. Results at the national level show a correlation coefficient of 0.803 with statistical significance for 0.01 one-tail. The WY in the RBDs of the North of the country was underestimated by 56.5 mm/ha/year and for the RBDs in the South of Portugal, the WY was overestimated by 58.1 mm/ha/year. This difference was explained through the spatial-temporal assessment of the main climatic variables used as input. This study contributes to a methodology to assess the level of confidence in WYM outputs and can be used to support the trustworthiness of water availability studies at sub-watershed, watershed, or RBD, using open access data and software.