An Improved evapotranspiration model for an apple orchard in northwestern China
Accurately estimating evapotranspiration (ET) is essential for orchard managers to design irrigation schedules and conserve water resources in semi-arid environments where water is often the limiting factor for successful production. Improving water use efficiency helps irrigation scheduling and thus benefits water resource management and the sustainability of the local economy. This study examined three existing ET models that were developed based on the Shuttleworth- Wallace model (SW) for estimating ET from sparsely covered crops in the arid Shiyang River basin (mean annual precipitation = 164 mm) in northwest China. We improved the existing clumping model (C model), a modified version of the SW model that simulates soil evaporation under the tree canopy and from bare soils outside of the canopy using a fixed bare soil/canopy area ratio. Our new ET model (the seasonal clumping model, or Cj model) considered the hourly dynamics of the bare soil surface area both under and outside of the tree canopy in an irrigated apple (Malus domestica Borkh. cv. Golden Delicious) orchard. The Cj model provided an improved estimate of soil evaporation by simulating soil surface areas based on hourly changes in canopy shade patterns and the canopy gap fraction. We validated the SW, C, and Cj models with ET fluxes measured by multiple methods, including sap flow of apple trees, and ET estimated by the micro-lysimeter and soil water balance methods for the 2008, 2009, and 2010 growing seasons. Soil water content, canopy characteristics (e.g., leaf area index), and leaf stomatal conductance were also measured periodically to parameterize the model. The growing season total ET rates estimated by the sap flow and micro-lysimeter method were 667, 674, and 583 mm in 2008, 2009, and 2010, respectively. The relative simulation errors of soil evaporation for the Cj, C, and SW models were 5%, 10%, and 30%, respectively. The absolute error for transpiration modeled by Cj (0.58 mm d-1) was significantly lower than for the C model (0.65 mm d-1) on a biweekly time scale. The growing season ET simulated by the Cj model was 628, 624, and 572 mm during 2008, 2009, and 2010, respectively, and the soil evaporation was 24% to 32% of the total ET. Over all, the Cj model was an improvement over the other two existing models for estimating apple orchard ET with a sparse tree cover. Our study also suggested that a supplement of 400 to 500 mm of irrigation water was essential to grow productive apple trees in the study region. The new model was successful in simulating the soil evaporation process and estimating the additional amount of water required to supplement natural rainfall by irrigation. The Cj model developed from this study was suitable for mature orchards or sparse forests with a bare soil surface, and it could efficiently estimate the water demand for irrigation scheduling.