Information Retrieval Of Four Dimensional Soil Moisture Content And Salt Concentration In Salted Farmland Based On Remote Sensing Data

The efficient utilization of soil moisture in farmland is very important for solving water resources shortage and food security in our country, especially in the arid and semi-arid area. However, the existing study on soil moisture and salt transportation is almost emphasizing on the controlled soil experiment and simulation. But in real situation the soil moisture and salt condition will change in accordance with vegetation growth, meteorlogical condition, soil environment and human being activities, as a result, the four-dimensional soil information variation is the important reference for the water-saving irrigation and food production in salt farmland. Concerning the low effiency of tranditional observation techniques, which is time consuming and high cost, forecasting the temporal-spatial variation of soil moisture and salt is difficulity. As a consequence, it is crucial to study the four-dimensional migration mechanism of soil moisture and salinity as well as establish a method to extract the four-dimensional quantity of soil moisture and salinity in farmland.The study initially study the efficient process to predict surface soil moisture and salt content by hyperspectral remote sensing signal. The findings indicate the transformation of raw hyperspectral reflectance, such as normalization or derivation of reflectance, is vital for searching the sensitive waveabnds. Furthermore, a simplified model by reduced sensitive wavebands is of high accurate, fairly simple and reliable, and physically convincing. According to the(Principal Component Analysis ranking system)PCAr, we select fourteen sensitive wavebands to soil moisture and salt, which are 440,540,570,1390,1430,1460,1740,1870,1900,1940,2010,2270,2350 and 2410nm. After which, we analyze the influence of soil moisture, salt and texture on spectral signature, the results show soil salt spectrum is easier to shake when soil moisture or texture variation happens, while soil moisture is not. Then we build a joint model to predict soil moisture and salt simultaneously, and the accurate of predicted salt increases compared with direct model, which means ignore soil moisture when predicting salt. To be more specific, this model first predict soil moisture and then predict soil salt by including the already predicted soil moisture. The statistical result R2 of the direct model is 0.47, and the rRMSE is 0.38; but they increase to 0.63<R2<0.95、 0.132<rRMSE<0.366 by the joint model.The surface soil moisture and salinity content predicted by remote sensing technology is the supplement of field samples, and can help acquire the tempo-space soil information. One direct study on predicting the three-dimensional space soil moisture or salt is based on the spatial relationship between surface and root soil, which can be studied by the cokriging interpolation. However, the spatial-based cokriging method is not feasible for summarizing time variation of soil information. The hydrodynamic model has the power to simulate synchronous changes. A precise simulation require precise hydraulic parameters, however which are often biased estimated in the field. This study establish a global parameter inversion model by modifying hydraulic parameters from inversed global scale factors. The merits of this inversion model help fixed hydraulic parameters to be more representative and also controls the parameters fixed in a physical meaning frame. Totally we build three kinds of inversion models in which five, ten and thirty scale factors are processed, the models are marked as GW5, GW10 and GW30 respectively. The Akaike information criterion(AICc)is a useful information criteria to judge the models and avert over-fitting risk. The AICc of the simplest GW5 model is the lowest(AICc=-467.367)which shows this model is the most suitable for this comparatively homogeneity area. This study further evaluate the GW5 model in the field samples measured in two years, and establish a bunch of models with the same structural as GW5. This models are marked as S_GW5, which also shows the obvious advantages of inversion models. Among the S_GW5, one model(V(RS))only use the predicted soil moisture by remote sensing technology, which indicates better simulated temporal-spatial soil moisture than forward model. The R2 and RMSE for simulated soil moisture of V(RS) during year 2013 are 0.892 and 0.01, and during year 2014 are 0.738 and 0.034 respectively. The study testify that remote sensing technology can help the simulation of S_GW5 inversion models.