The Data Assimilation Of Farmland Moisture Based On SiB2 And Kalman Filter In The Low Hilly Region Of Red Soil

The land surface is an important part of the earth's biosphere and plays a crucial role in the climate system,among which,farmland soil moisture is an important index to study the relationship between crop growth and soil gas interaction.Establishing an accurate land surface process model to assess land surface conditions,especially the crop root field soil moisture changes in land atmosphere interactions,is of positive significance to promote regional scale weather forecast accuracy.Amass of simulation errors will accumulate over time when the land surface model is in operation,because soil moisture have large changes in different space and time,so it is necessary to design a corresponding data assimilation scheme improving the model accuracy.In the study of soil moisture balance in the farmland watershed in low hilly region of red soil,some indispensable comparisons and developments based on improved SiB2 land surface process model and a variety of Kalman filtering algorithms of data assimilation scheme have great theory value because the SiB2 is a simple and efficient land surface process model and the Kalman filter is one of the most widely used data assimilation method.The main results are as follows:(1)With a lot of meticulous measurements about the seeding plants and soil of the peanut and sweet potato fields in the hilly experimental region in Jiangxi province,all important kinds of elements for SiB2 land surface process model running were acquired,including morphological and physiological parameters of peanuts and sweet potatoes in different growing stages as well as soil characteristic and optical parameters in farmland.Importing these parameters to SiB2,finally,the land surface process model based SiB2 in the hilly area of red soil was successfully established through improving downward long wave radiation calculation method and the Force-Recovery method about surface temperature calculation of SiB2,which can be used to simulate hourly dynamic changes of soil moisture in the root domain in four growing stages of peanuts(seedling,flowering,stem-growing and plump-earing stage)and sweet potatoes(seedling,sprouting,podding and plump-maturing stage).(2)Three kinds of land surface process model data assimilation scheme based on SiB2 were successfully established,such as KF(Kalman Filtering),EnKF(Ensemble Kalman Filtering)and UKF(Unscented Kalman Filtering),and a kind of improved scheme for EnKF was also developed,namely DEnKF(Double Ensemble Kalman Filtering),to respectively assimilate hourly dynamic changes of soil moisture in the root domain of peanuts and sweet potatoes in four growing stages and analyze their statistical errors.(3)Compared with the performance of the four kinds of data assimilation schemes,the accuracy and linear fitting results of KF are the worst,which is not suitable for the development of land surface data assimilation scheme.The accuracy and linear fitting results of EnKF are better,but they require a large number of sets and tremendous demand to calculate.The accuracy of UKF is about the same as that of EnKF,but the linear fitting results are slightly worse.UKF is easy to program and their calculated demand is fewer.The accuracy and linear fitting results of DEnKF are the best,even though their calculation needs are also larger than that of UKF.Satisfactory precision and linear fitting results can be obtained by using a small number of sets in DEnKF,suitable for the experimental zone of soil moisture dynamic simulation data assimilation.