基于TM影像植被指数的耕地地力状况反演研究

建立基于TM遥感影像植被指数的耕地地力反演模型,为区域耕地资源管理及可持续利用提供科学依据.选择耕地地力相似的山东郯城县和东平县,利用耕地地力实地调查分析和TM遥感数据,通过相关分析筛选对郯城耕地地力有较好反映的植被指数,通过回归分析建立耕地地力-植被指数模型,以郯城县数据建模,东平县数据进行反演和验证.结果显示,增强型植被指数(EVI)与耕地地力评价结果有最显著的正相关性,相关系数达到0.82.以EVI 为自变量建立的二次方程式模型拟合效果最好,决定性系数达到0.69.采用决定性系数(R²)、均方根误差(RMSE)、精密度和准确度4 个指标对模型的反演结果和原始评价结果之间的符合度进行统计检验,EVI 的二次方程式模型准确度最高,为95.84%,RMSE和精密度最小,分别为5.21 和0.04,为耕地地力最佳反演模型.对比分析东平县耕地地力反演图和常规耕地地力评价图,模型反演的各耕地地力等级与实际耕地地力评价等级具有较好的空间分布一致性.将耕地地力归并为高、中和低3 个等级,高中低等级面积比例差异均在3.3%以内,符合研究区实际,反演效果较好.研究证明了基于定量遥感手段进行耕地地力估测的可行性,为耕地资源的监测利用提供了有效手段.

Arable Land Fertility Inversion Based on Vegetation Index from TM Image

The establishment of the arable land fertility inversion model based on vegetation index from TM remote sensing image provides a scientific basis for resource management and sustainable use of regional farmland. The study used the field survey of the arable land fertility and TM remote sensing data to screen vegetation index which can better reflect the arable land fertility. We chose the counties of Tancheng and Dongping in Shandong Province as study area, where the arable land fertilities are similar. Regression analysis was used to establish the model of arable land fertility-vegetation index with data of Tancheng, and the data of Dongping were used to validate the inversion model. The results showed that the positive correlation between enhanced vegetation index (EVI) and evaluation results of cultivated land is the most significant one, and the correlation coefficient was 0.82. The best fitted model was the Quadratic model with EVI as independent variable whose decisive factor was 0.69. The conformity degree between the result of inversion model and the result of original evaluation were tested by use of four indicators which include the decisive coefficient (R²), root mean square error (RMSE), precision and accuracy. The results showed that the Quadratic model built by EVI was the best inversion model of arable land fertility. The accuracy of it was the highest which is 95.84%, and the RMSE and precision was the lowest which are 5.21 and 0.04 respectively. Through the comparison of the result of inversion model and conventional evaluation of arable land fertility in Dongping, we can see that the fertility levels obtained by the inversion model agree with the actual farmland productivity levels in space. Classifying the arable land fertility levels into three grades of high, medium and low, it was found that the inconsistent areas of the three grades all took less than 3.3% of the area of the grade. The inversion effect was good and accorded with the actual situation. This study proved the feasibility of estimating farmland productivity by quantitative remote sensing, and provided an effective tool for monitoring and utilizing farmland resources.