基于高光谱技术的退耕还林地年限判别

自2000年以来,黄河中游坡度较大的不同区域、同一区域的不同部位在不同年度实施了退耕还林工程,促进了黄河中游土壤质量及生态环境的改善.为了研究退耕工程对土壤及环境的影响机制,需要快速获取退耕年限及土壤特征.该研究以黄河中游大宁县不同年限退耕还林土壤为研究对象,获取不同年限退耕还林土壤理化性质,同时测定不同退耕年限土壤光...

Identification of the years of returning farmland to forest land using hyperspectral technology

Soil quality and ecological environment have been improved in the middle reaches of the Yellow River in China. This improvement can attribute to the national project of returning farmland to forest in different years since 2000, particularly on the great slopes. Therefore, it is necessary to rapidly acquire the years of returning farmland and soil characteristics, in order to evaluate the ecological benefits of the project. Taking the middle reaches of the Yellow River as the study area, this study aims to obtain the soil''s physical and chemical properties, as well as the soil''s spectral curves for the returning cropland to forest in different years using hyperspectral imaging technology. Some spectral preprocessing were utilized, including the Savitzky-Golay Smoothing (SG), Reciprocal Logarithm (RL), the First-Order Differential (FD), Continuum Removal (CR), Principal Component Analysis (PCA), and Spectral Characteristic Parameters (SCP). The classification models were constructed for the years of returning cropland to forest using the K-means clustering (K-means), support vector machine (SVM), and Linear Discriminant Analysis (LDA). Among them, the input factors were set as The Principal Component Of Original Reflectance (R-PCA), Principal Component Of Logarithm Of The Reciprocal (RL-PCA), Principal Component Of First-Order Differential (FD-PCA), Principal Component Of Continuum Removal (CR-PCA), and SCP. The results showed that: 1) The content of Soil Organic Carbon (SOC) increased gradually, and the content of sand particles increased first and then decreased, with the increase of the years of returning cropland. The content of SOC was negatively correlated with the soil original reflectance. 2) There was a similar shape of soil original spectral curve in the different years of returning farmland, indicating the overall increasing trend. The CR preprocessing was significantly improved the absorption of the spectral curve, with the outstanding absorption characteristics at 480, 900, 1100, 1400, 1900, 2200, and 2350 nm. 3) The highest accuracy (87.50%) was achieved in the classification model of LDA with the CR-PCA as input factor, which was the optimal classification model. The second highest accuracy (84.38%) was found in the classification model of SVM with the FD-PCA as the input factor. All the classification models with the CR-PCA as input factor shared the highest accuracy of more than 75%, with the maximum of 87.50 %, indicating that the CR-PCA was the optimal input factor to distinguish the different years of returning farmland in this case, followed by the FD-PCA. As such, the rapid distinction of the years was fully realized for the returning farmland to forest, according to the spectral characteristics and classification through the soil spectral curves. The finding can provide a strong reference for the soil properties and environmental impacts of the returning farmland to forest.