Evapotranspiration modelling in PALMSIM: The case of Indonesia

Abstract
The Global Yield Gap Atlas (GYGA, www.yieldgap.org) is an international project assessing potential yields, water limited yields and yield gaps. One upcoming extension of the Global Yield Gap Atlas is oil palm in Indonesia. Oil palm yields are simulated on a monthly basis using the crop simulation model PALMSIM v2.0, which contains a water balance component with rainfall and evapotranspiration (ET) as main fluxes. Accurate estimation of these fluxes is vital for the determination of the water-limited yields. PALMSIM v2.0 estimates crop potential ET using the IRHO method, but the accuracy of this method can be questioned as it lacks a scientific foundation and the relation between the no. of raindays per month and irradiation is weak. Accuracy of the crop potential ET estimation also depends on the quality of the weather data required to estimate ET. The GYGA is often forced to use NASA-Power gridded weather, when observed weather data is not available. It is therefore important to know how well NASA-Power simulates observed weather data in Indonesia. The objectives of this research are (1) to assess the performance of NASA-Power to approximate weather variables in Indonesia, and (2) to find the best ET method for PALMSIM v2.0 when NASA-Power weather data is used. Observed monthly weather data from 20 BMKG weather stations near the major oil palm cultivated areas were collected. The performance of NASAPower to estimate weather variables related to ET estimation were evaluated for all locations. Six ET methods with different data requirements (IRHO, Hargreaves-Samani, Priestley-Taylor, Makkink, Romanenko and the FAO-56 Penman-Monteith) were evaluated using observed and NASA-Power weather data. The FAO-56 Penman-Monteith method using observed weather data was hereby used as a reference. Relatively low R-squares between NASA-Power and observed weather data were found for all weather variables, except relative humidity when compared to literature. Relative humidity was reasonably estimated by NASAPower, because of the stable humidity in Indonesia throughout the year. The low R-squares and the biases found for irradiation and temperature can be explained by the positioning of the weather stations within a NASA-Power gridd cell in combination with the complex geography of Indonesia with many mountainous and coastal areas. The FAO-56 Penman-Monteith using only NASA-Power weather data performed better than the other uncalibrated methods using observed weather data, which was mainly attributed to large biases of the methods. Recalibartion improved the ET methods, but only the radiation based methods (Makkink and Priestley-Taylor) were able to capture the ET estimations of the FAO-56 Penman-Monteith method. The usage of NASA-Power irradiation data resulted in a huge accuracy decrease of these methods, which is stressing the importance of observed irradiation data to accurately estimate ET. The re-calibrated Makkink equation performed best when only irradiation data was available. It was shown that the FAO-56 Penman-Monteith method performed best in Indonesia when NASA-Power weather data is used, which was mainly attributed to the relatively well estimated humidity and relatively low wind speed values found in Indonesia. The outcomes of the study might have been different when the research was done in another climatic zone.