Water Transfer Processes And Its Simulation In Rainfed Maize Fields Of Eastern Loess Plateau

Plastic mulch can significantly improve soil water and temperature in croplands,then affect hydrological processes in croplands.Therefore,analysis of water transfer and establishment of models for evapotranspiration modeling and partitioning in rainfed croplands of Loess Plateau is of importance for water use efficiency boosting.In this study,a field experiment was conducted in non-mulched（CK）and mulched（PM）maize fields of eastern Loess Plateau to monitoring maize water consumption at different scales by using photosynthesis system,sap flow and eddy covariance systems.Based on the monitoring results,the effects of plastic mulch on the water and heat fluxes transfer were investigated,evapotranspiration models based on machine learning were proposed,and dual crop coefficient approach was calibrated for evapotranspiration modeling and partitioning.The main results are outlined as follow:（1）Compared with CK,soil water content was enhanced by 2.8%and 4.2%,maximum green leaf area index was enhanced by 0.05 and 0.44 m² m^-2,growing season was shorten by 9 and 10 d,plant height was enhanced by 24 and 38 cm,underground drymatter was improved by 2.3and 3.7 Mg ha^-1under PM for 2015 and 2016,respectively.（2）Leaf transpiration,sap flux and evapotranspiration all presented typical diurnal and seasonal variations.They all peaked around noon,and then decreased gradually.PM increased leaf transpiration and sap flux.However,PM decreased evapotranspiration for the whole growing season and altered the components of evapotranspiration that evaporation as well as the ratio of evaporation to evapotranspiration were decreased while the ratio of transpiration to evapotranspiration was increased,thus more water was took for maize growth.（3）PM increased energy imbalance in maize fields.Energy fluxes presented typical diurnal and seasonal variations during the entire growing seasons.Net radiation（R_n）was dominated by latent heat flux（λET）during maize growing seasons while sensible heat flux（H）was the predominant component of R_n during non-growing seasons.Compared with the energy fluxes before rainfall,λET increased significantly while H and soil heat flux（G）decreased after rainfall in development stage,however in late stage,the changes in soil water content resulted by rainfall didn’t increaseλET significantly for the both treatments,which indicated seasonal distribution of rainfall had significant influence on energy budget of maize fields.R_n was the primary factor affectingλET,followed by average air temperature and vapor pressure deficit,wind speed and relative humidity had relatively minor influence onλET.（4）The proposed evapotranspiration models,which were based on artificial neural network optimized by genetic algorithm（GANN）,extreme learning machine（ELM）,generalized regression neural network（GRNN）and support vector machine（SVM）,had good capacity to model maize evapotranspiration.For the whole seasons,ELM had the best performance,and can be recommended to model maize evapotranspiration in the study area.（5）The basal crop coefficient and evaporation coefficient of dual crop coefficient approach were calibrated based on leaf area index.The calibrated dual crop coefficient approach performed well for simulating and partitioning maize evapotranspiration.The simulated results had good agreements with the measured ones,with coefficient of determination,root mean square error Nash-Sutcliffe efficiency coefficient of 0.824⁰.870,0.381⁰.561 mm d^-1,0.817⁰.871 and 0.332⁰.449 mm d^-1,respectively.