Methodology Development For Evapotranspiration Retrieval From Remote Sensing Data

Evapotranspiration(ET),including the evaporation of soil water,vegetation transpiration and the evaporation of trapped water in canopy,is a key parameter for the process of surface energy balance and hydrological cycle and a significant index to monitor the growth status and yield of crops.Compared with the traditional method of single point observation on meteorological station,remote sensing technology hold a great potential on estimating evapotranspiration at regional scale with irreplaceable superiority.For that,many scholars have a lot of study on retrieval of evapotranspiration from remote sensing data with many feasible methods have been developed.In this paper,for(Land Surface Temperature/Fractional Vegetation Cover)LST/FVC method,retrieval and monitoring of global ET are developed by this model.Specifically,three questions are under considered about this model in the follow:(1)The correlation mechanism between soil moisture and evapotranspiration has not been fully considered;(2)Difficulty in obtaining spatiotemporal continuous evapotranspiration because of clouds;(3)Rely on ground based auxiliary data.For that,the following three aspects of work are implemented:Firstly,the correlation mechanism between soil moisture and evapotranspiration in the LST/FVC model is discussed,namely,whether soil evaporation and vegetation transpiration and soil moisture change synchronously("synchronously"hypothesis)or asynchronously("asynchronously"hypothesis).For that,synchronous and asynchronous hypothesis feature space model were driven by using the same data.North American prairie evapotranspiration was estimated under different aridity conditions.The main conclusion was that whether the different soil moisture changes within surface and root-zone layers with respect to soil evaporation and vegetation transpiration should be considered primarily depends on the aridity conditions of the study areas.The asynchronous-assumed trapezoid is more suitable in semi-arid and sub-humid areas.However,for the humid regions,the asynchronous-assumed trapezoid reveals no superiority to the synchronous-assumed trapezoid.In addition,considering different energy closure techniques in ET correction,there is no consistent conclusion which performance is greater between hypothetical scheme under drought conditions.Specifically,it was found that cases of surface soil with zero water availability that were defined by the asynchronous-assumed trapezoid method rarely occur,even in arid areas,which indicated that the critical boundary that determines whether the root-zone layer begins to be water-stressed may need to be redefined.Secondly,for the question of difficulty in obtaining spatiotemporal continuous evapotranspiration because of clouds,the regional-scale scheme of all-sky evapotranspiration from remote sensing is explored preliminarily.In this study ET over clear-sky pixels was estimated from a pixel-to-pixel land surface temperature/vegetation index feature space method.For cloudy pixels,the Penman-Monteith equation was used to calculate ET.The MODIS data and CLDAS(CMA Land Data Assimilation System)grid meteorological data were used to obtained instantaneous evapotranspiration under all-sky conditions over Heihe River Basin of China.The measured ET data from four sites in the study area was used to verify the estimated ET.It shown that the method can obtain highly accurate regional-scale ET with a RMSE around70 W/m~2.Finally,for the question of relying on meteorological data.In this paper,a new method based on the moderate-resolution imaging spectroradiometer(MODIS)is proposed to derive the relative humidity on all-sky conditions.Two key assumptions in the approach under cloudy conditions are that the actual water vapor is linearly related to the total precipitable water vapor(PWV),and air temperature is also linearly related to land surface temperature(LST).The verification results of 30 sites with different underlying surfaces and aridity conditions in the AmeriFlux network shown that this method has high accuracy.For different aridity conditions,RH retrieval reveals considerable accuracy with a root mean square error(RMSE)approximately 15.3%over an arid and semi-arid region,whereas a comparable RMSE of 17.0%can be obtained over a humid region.In particular,air temperature and relative humidity with5km resolution can be obtained simultaneously by this method,producing a data basis for the development of evapotranspiration retrieval model driven by full remote sensing data.In summary,the solution of the three key problems makes it theoretically possible to obtain global spatio-temporal continuous evapotranspiration exclusively from remote sensing data,laying a reliable bias for retrieval and monitoring of global all-weather evapotranspiration by LST/FVC method.