Impact of climate, vegetation, soil and crop management variables on multi-year ISBA-A-gs simulations of evapotranspiration over a Mediterranean crop site
Abstract. Generic land surface models are generally driven by large-scale forcing datasets to describe the climate, the surface characteristics (soil texture, vegetation dynamic) and the cropland management (irrigation). This paper investigates the errors in these forcing variables and their impacts on the evapotranspiration (ET) simulated from the Interactions between Soil, Biosphere, and Atmosphere (ISBA-A-gs) land surface model over a 12 year Mediterranean crop succession. We evaluate the forcing datasets used in the standard implementation of ISBA over France where the model is driven by the SAFRAN high spatial resolution atmospheric reanalysis, the Leaf Area Index (LAI) cycles derived from the Ecoclimap-II land surface parameter database and the soil texture derived from the French soil database. For climate, we focus on the radiations and rainfall variables and we test additional datasets which includes the ERA-Interim low spatial resolution reanalysis, the Global Precipitation Climatology Centre dataset (GPCC) and the MeteoSat Second Generation (MSG) satellite estimate of downwelling shortwave radiations. The methodology consists in comparing the simulation achieved using large-scale forcing datasets with the simulation achieved using local observations for each forcing variable. The relative impacts of the forcing variables on simulated ET are compared with each other and with the model uncertainties triggered by errors in soil parameters. LAI and the lack of irrigation in the simulation generate the largest mean deviations in ET between the large-scale and the local-scale simulations (equivalent to 24 and 19 months of ET over 12 yr). The climate induces smaller mean deviations equivalent to 7–8 months of ET over 12 yr. The soil texture has the lowest impact (equivalent to 3 months of ET). However, the impact of errors in the forcing variables is smaller than the impact triggered by errors in the soil parameters (equivalent to 27 months of ET). The absence of irrigation which represents 18% of cumulative rainfall over 12 years induces a deficit in ET of 14%. It generates much larger variations in incoming water for the model than the differences in rainfall between the reanalysis datasets. ET simulated with the Ecoclimap-II LAI climatology is overestimated by 18% over 12 years. This is related to the overestimation of the mean LAI over the crop cycle which reveals inaccurate representation of Mediterranean crop cycles. Compared to SAFRAN, the use of the ERA-I reanalysis, the GPCC rainfall and the downwelling shortwave radiation derived from the MSG satellite have little influence on the ET simulation performances. The error in yearly ET is mainly driven by the error in yearly rainfall and to a less extent by radiations. The SAFRAN and MSG satellite shortwave radiation estimates show similar negative biases (−9 and −11 W m−2). The ERA-I bias in shortwave radiations is 4 times smaller at daily time scale. Both SAFRAN and ERA-I underestimate longwave downwelling radiations by −12 and −16 W m−2, respectively. The biases in shortwave and longwave radiations show larger inter-annual variation for SAFRAN than for ERA-I. Regarding rainfall, SAFRAN and ERA-I/GPCC are slightly biased at daily and longer time scales (1 and 0.5% of the mean rainfall measurement). The SAFRAN rainfall estimates are more precise due to the use of the in situ daily rainfall measurements of the Avignon site in the reanalysis.