The Research Of Soil Moisture Assimilation Based On Cloud Parameter Model And Esemble Kalman Filter

Natural disasters pose great challenges to our life, drought, as one of the most notorious, in particular makes threats and impacts national economy, society development, and daily life. Acquiring accurate information on soil moisture has great significance in revealing the real state of drought. Unfortunately, existing methods or techniques have some defects in despite of their advantages:①remote sensing has the specialty in large-scale and timeliness monitoring, but as its models are formed on the basis of surface observations (horizontal), they could not depict soil moisture in vertical,②ecause of land model, data assimilation could simulate soil moisture in depth with some certain layers, but the parameterization is not perfect, and land models are complicated and not thematic, having low accuracy in soil moisture estimation,③field observation could.directly acquire the most objective information on soil moisture, but the data amount is quite limited because of the sparse distribution of observation stations and discontinuous observation time, the lack of data would be the most obstacle in research, and further, field measurements are not representative because of their single-point-observation method. Aiming at the above predicament, with the ultimate goal for drought monitoring in a way of multi-mode and multi-source data integration, it's of great practice significance for making research on the assimilation of cloud parameter remote sensing drought monitoring model and soil moisture field observations.This dissertation covers:①methods for soil moisture acquisition were comprehensively summarized and compared, followed by an analysis for their problems and deficiencies,②soil moisture characters like its concept, quantitative expression, factors for vapor loss, and role in the water and energy cycle were analyzed and summarized,③for the assimilation of cloud parameter remote sensing drought monitoring model and soil moisture field observations, research was made on the parameter calibration of Ensemble Kalman Filter, impacts coming from ensemble number, model error, and observation error were studied, and on this basis the framework of assimilation system was projected and details of the assimilation algorithm was depicted,④single-point measurements could not match the requirement for regional observation field in data assimilation, geostatistics was introduced for solving this spatial scale confliction, on the base of ordinary kriging interpolation method, parameter calibration research was made on its variogram, influence brought by the simulation model, range, and nuddget were studied,⑤a new method was proposed for soil moisture estimation with meteorological parameters by the way of BP neural network, principal component analysis was used to extract representative parameters, a study on the construction of the network was made, including the sampling method with the combination of random and hierarchical, the influence of training algorithm, way of using parameters, structure of hidden layer was studied,⑥Gansu province was taken as the demonstrated area, based on 2008 MODIS remote sensing data, agricultural ten-day report (AB Report), and ground meteorological observations (A Report), experiments were made on soil moisture ordinary kriging spatial interpolation, BP neural network soil moisture estimation, and data assimilation for cloud parameter remote sensing drought model and field observations based on Esemble Kalman Filter, experimental programs were made respectively, methods and results were analyzed and verified.This dissertation made innovation in three aspects:(1) Thoughts and methods of geostatistics were introduced in solving the problem of soil moisture interpolation in spatial, took the soil moisture as the regionalized variable, by parameter research of its variogram, good estimation for soil moisture in a regional scale was achieved on the base of ordinary kriging method, ruleless and dot distribution measurements were well extended to a continuous regional space;(2) An estimation algorithm based on character parameter and BP neural network was proposed, good estimation of soil moisture was acquired, and this made a helpful complementarity for soil moisture field observation;(3) The assimilation system based on Ensemble Kalman Filter was constructed, the assimilation with cloud parameter model and soil moisture field observation was achieved, accurate for soil moisture estimation was effectively improved.According to the research of this dissertation, conclusions could be derived as follows:(1) Assimilation could make a balance and complementary between remote sensing model and field observations. On the one hand, the continous transition character of field observations was hand on, and the extremum center induced by the local control of field observation was well improved. On the other hand, the terrain feathers were well represented because of the remote sensing model. The assimilation of remote sensing model and field observations to get high accuracy soil moisture information has great potential in the application of drought monitoring.(2) Regional soil moisture field is a necessary import for assimilation, from the perspective of geostatistics soil moisture could be taken as a typical regionalized variable. The experiment results showed that ordinary kriging method was reliable in soil moisture spatial interpolation. (3) The lack of field observation makes great obstacles for data assimilation, in order to do researches on the key roal which soil moisture was taken in the land and atmospheric interactions, its action and reaction with these meteorological parameters could be derived. The experiment results showed that BP neural network could well simulate such kind of interactions, and made good estimation of soil moisture, which would be a reliable substitute in times with the absence of field observation.