基于SOC710VP高光谱成像仪的冬小麦土壤含水率反演模型研究

【目的】实现小麦农田土壤含水率大面积快速监测。【方法】以冬小麦冠层高光谱数据为基础,计算得到8种植被指数,通过对关键生育时期(拔节期、抽穗期、灌浆期)不同水分处理下冬小麦不同土层(0～20、20～40、40～60 cm)土壤含水率与植被指数拟合状况进行分析和筛选,分别构建了基于植被指数的不同土层土壤含水率反演模型,并对模型进行检验。【结果】①各时期植被指数拟合效果有所差异,拔节期0～20 cm土层以植被指数VOG1拟合效果较好,相关系数为0.88,20～40 cm土层以植被指数mNDVI705拟合效果较好,相关系数为0.75,40～60 cm土层以植被指数VOG3拟合效果较好,相关系数为0.59;抽穗期0～20 cm土层以植被指数mNDVI705拟合效果较好,相关系数为0.70,20～40 cm土层以植被指数mNDVI705拟合效果较好,相关系数为0.72,40～60 cm土层以植被指数mSR705拟合效果较好,相关系数为0.57;灌浆期0～20 cm土层以植被指数mNDVI705拟合效果较好,相关系数为0.88,20～40 cm土层以植被指数SARVI拟合效果较好,相关系数为0.68,40～60 cm土层以植被指数SARVI拟合效果较好,相关系数为0.71;②各土层土壤含水率与植被指数拟合效果有所差异,其中利用VOG1和mNDVI705组合构建的模型反演0～20 cm土层,决定系数R2为0.743,利用mNDVI705和SARVI组合构建的模型反演20～40 cm土层,决定系数R2为0.707,利用VOG3、mSR705和SARVI组合构建的模型反演40～60 cm土层,决定系数R2为0.484;③通过建立植被指数对土壤含水率的反演模型,0～20 cm土层含水率反演效果好于20～40 cm和40～60 cm。【结论】高光谱植被指数反演模型中,以0～20 cm土层的估算模型最佳,植被指数组合为VOG1和mNDVI705。综上可知,该研究方法进行土壤含水率的反演是可行的。

Estimating Soil Moisture Distribution in Winter Wheat Field Using SOC710VP Hyperspectral Imagery

【Objective】Soil water controls crop growth and many soil physical and biochemical processes, and the purpose of this paper is to present how hyperspectral imagery can be used to estimate soil moisture distribution rapidly at large scale.【Method】Based on hyperspectral data of the canopy of winter wheat, we calculated eight vegetation indices and then linked them to soil water content at different depths(0～20, 20～40, 40～60 cm) during key growth stages(jointing stage, heading stage, filling stage) of a winter wheat field.【Result】①The fitting between soil moisture and the vegetation indices varied with growth season. At jointing stage, the indices VOG1,mNDVI705 and VOG3 were superior, whereas at heading and filling stages, mNDVI705 and mSR705, and mNDVI705 and SARVI worked better, respectively. ② The fitting between soil water content and vegetation indices varied with soil depth as well. For the 0～20 cm soil, the model using VOG1 and mNDVI705 gave the best result with the coefficient of determination(R2) being 0.743, while for 20～40 cm soil, the model using mNDVI705 and SARVI was most accurate with R2 being 0.707. For the soil in 40～60 cm, the best vegetation indices for estimating the moisture was VOG3, mSR705 and SARVI, with R2 being 0.484. ③It was found that the fitting of the model for 0～20 cm soil was superior to that for 20～40 cm and 40～60 cm soil.【Conclusion】Using VOG1 and mNDVI705 indices calculated from the hyperspectral imagery can estimate the moisture in 0～20 cm soil reasonably well, and can thus help improve irrigation design and water resource management at regional and catchment scales.