| The arable land is not only the essence of the land but also an important resource for agricultural production, the arable soil fertility will directly affect the sustainable development of agriculture and food security. At present, evaluation of farmland using GIS has been a lot of research but they usually need a lot of ground truth data as a basis with long-consuming human and material resources. Therefore the establishment of the arable soil fertility inversion model based on remote sensing images provides a scientific basis for resource management of regional farmland and sustainable use. The study used the field survey analysis of the arable soil fertility and TM remote sensing data to screen vegetation index with better reflection for arable soil fertility.We chose Tancheng and Dongping in Shandong province, which have similar arable soil fertility. First, under the support of GIS,we divided evaluation unit, screened the participating factors and constructed evaluation system with the measured ground data. Finally we selected the best evaluation model for evaluation of farmland of Tancheng and Dongping county and classified the fertility of arable land. Finally we used the Arcmap and Mapgis professional graphics software for mapping in order to achieve cultivated land fertility distribution visualization of Tancheng and Dongping.In this study, we screened11strong universal vegetation indices which have wide range of applications to participate in the analysis to achieve remote sensing inversion models of fertility of arable land based on vegetation indices. Passed through the correlation analysis between11spectral parameters and farmland of score Tancheng County by using SPSS software, we screened the better vegetation index to reflect fertility of arable land. We used linear, nonlinear and multiple regression analysis to select optimal vegetation index as indicators of farmland inversion to establish farmland-vegetation index model by using the best two vegetation index as the independent variable and farmland score of Tancheng as the dependent variable. In order to verify the accuracy of the inversion model and test the versatility of the model, we must select the other regions as validation of the model, the basis for promotion as a model. Therefore, we chosen Tancheng county as the validate region to chose and validate the model. Meanwhile we chose the correlation coefficient, the decisive coefficient, the root of variance, the accuracy and precision five common indicators to validate the degree of match between the model predictions and actual values of statistical tests and selected the optimal inversion model. Finally, we used the optimal model to invert Dongping County arable land fertility and compared with the result of farmland based on GIS. The results showed that:(1) The11vegetation index that we chose had a high correlation with evaluation results of farmland, the correlation coefficient was0.678**.Among the11vegetation index, there were the most significant positive correlation between enhanced vegetation index (EVI) the normalized difference vegetation index (NDVI) and evaluation results of farmland, and the correlation coefficients were0.822and0.802.(2) All the linear, nine non-linear and multi-linear models had the higher fit. Among this the quadratic model using NDVI and Cubic model using EVI as the independent variable had best fit to the evaluation score compared to the other models, and the decisive factor were0.661and0.692, respectively. So we selected NDVI-Quadratic, NDVI-Cubic, EVI-Quadratic, and EVI-Cubic models to conduct further validation.EVI-Quadratic curve had the highest accuracy of95.84%, RMSE and the minimum deviation precision were5.207and0.049, better than the EVI Cubic curve. The aspects of the NDVI model were poorer than EVI model. Therefore, we finally selected EVI-Quadratic model as the optimal model to invert the farmland situation. Model formula: Score of farmland fertility=76.13+5.516EVI+3.346EVI2.(4) We used optimal model to invert Dongping arable land, through comparative analysis between arable soil fertility maps in Dongping and conventional arable soil fertility evaluation map, the cultivated land fertility levels of the model inversion had good spatial distribution uniformity with actual farmland productivity evaluation level, the area ratio difference of high, moderate and low grades was less than3.3%. The Inversion effect was good and accorded with the actual research area.(5) This study proved that it had good feasibility to estimate farmland productivity estimation based on quantitative remote sensing, and provided an effective tool for monitoring and utilizing farmland resources. |
PDF Full Text Request