On the impact of the choice of the evapotranspiration equation in a crop model : climate data error propagation and climate change impact projection

International audienceAs it enables the understanding and the quantification of the transfer of water in ecosystems and from ecosystems to the atmosphere, evapotranspiration is a key component to assess climate impact on hydrology and agriculture. In crop models, the estimation of the evapotranspiration rate requires first calculating potential or reference evapotranspiration from climate data. To compute reference evapotranspiration different formulas requiring more or less climate data are used. The choice of the formulation of this key process is very likely to have an impact on calculated crop yield. The FAO recommends using the Penman-Monteith (PM) equation if all the climate data required for this equation are available and using Hargreaves (H) equation when climate data, especially net radiation, are missing. The Priestley-Taylor equation is also widely used in crop models. Which of these equations is the most accurate when all the climate data required are available but contain errors ? Does the choice of the evapotranspiration equation have an impact on crop yield projection in a context of climate change ? Does the use of some equations induce more pessimistic crop yield projection ? We studied the impact of the reference evapotranspiration equations on simulated crop yield using climate data with errors. 4 equations (PM, H and 2 versions of the Priestley-Taylor equation - PT) were tested simulating pearl millet over 12 stations in Senegal. In this case, we found that the use of a PT equation may introduce a percent mean bias error of more than -35% on simulated crop yield while it is limited to 2% when using the H equation. The influence of the evapotranspiration equation on the quantification of climate change impact on crop yield is examined applying the AgMIP C3MP protocol over the 12 stations in Senegal then analyzing ISI-AgMIP GGCM Intercomparison fast-track project outputs over the world. Our preliminary results show that crop yields computed using a PT equation are more sensitive (crop yield reduced by 31% on average) to temperature change than PM and H (-23%) equations and this for each of the 12 stations considered in Senegal