基于棉花功能叶高光谱参数的土壤电导率监测模拟

通过2008-2009年在江苏南京农业大学牌楼试验站的盆栽试验,选择耐盐棉花品种中棉所44和盐敏感性品种苏棉12号为材料,模拟5种不同含盐水平的滨海盐土(0、0.35％、0.60％、0.85％和1.00％),分析了棉花生育期棉田土壤电导率与棉花功能叶光谱反射率和高光谱参数的关系,并建立了棉田土壤电导率(EC)的定量监测模型.结果表明:棉花功能叶光谱反射率在近红外和中红外区域均随土壤盐分水平的升高而升高;以敏感波段1350nm和2307 nm构建的归一化光谱指数NDSI(R1350,R2307)与土壤电导率的决定系数最高,基于此构建了基于NDSI(R1350,R2307)的棉田土壤EC监测模型:EC=-42.899NDSI(R1350,R2307)+27.338;在光谱微分参数中,以TM影像第5个波段的光谱反射率(TM5-SWIR)与棉田土壤EC的决定系数最高,构建了基于TM5-SWIR的棉田土壤EC监测模型:EC=0.0574 TM5-SWIR2-2.5928 TM5-SWIR+ 30.021.以NDSI(R1350,R2307)和TM5-SWIR为自变量的监测模型的预测精度均较高,分别为0.887和0.814,根均方差均较小,分别为1.09和1.29 dS·m-1.利用棉花功能叶NDSI(R1350,R2307)和TM5-SWIR均能较好地监测棉田土壤电导率.

Monitoring and simulation of soil electrical conductivity based on the hyperspectral parameters of cotton (Gossypium hirsutum) functional leaves

Taking the salt-tolerant cotton variety CCRI-44 and salt-sensitive cotton variety Sumian 12 as test materials, a two-year pot experiment was conducted at the Pailou experimental station of Nanjing Agricultural University in 2008 and 2009 to study the relationships of soil electrical conductivity (EC) with the spectral reflectance and hyperspectral indices of cotton functional leaves at different growth stages under five simulated salinity levels (0, 0.35%, 0.60%, 0.85%, and 1.00%) of coastal saline soils, and the quantitative monitoring models on the cotton soil EC were established. With increasing salinity level, the cotton functional leaves had an increased spectral reflectance in near-infrared and middle-infrared regions, and the spectral parameter normalized difference spectrum index (NDSI) based on 1350 nm and 2307 nm, i. e., NDSI (R1350, R2307), correlated well to the soil EC. With the NDSI (R1350, R2307) as independent variable, the soil EC monitoring model was constructed as EC = -42.899 NDSI (R1350, R2307) +27.338. Among the derivative spectral parameters, TM5-SWIR was most correlated to soil EC, and thus, the soil EC monitoring model was constructed as EC = 0.0574TM5-SWIR2-2.5928TM5 -SWIR+30.021. The two models with NDSI (R1350, R2307) and TM5-SWIR as the independent variables respectively all had higher prediction precision, with the determination coefficient being 0. 887 and 0. 814 and the root mean square error being 1.09 and 1.29 dS x m(-1), respectively, suggesting that using the hyperspectral parameters NDSI (R1350, R2307) and TM5-SWIR of cotton functional leaves could effectively monitor the soil EC of saline cotton fields.