黄土高原春小麦地上鲜生物量高光谱遥感估算模型

Hyperspectral remote sensing estimation models for aboveground fresh biomass of spring wheat on Loess Plateau

通过田间小区试验,测定了4个春小麦品种(定西24号、陇春8139、高原602和定西38号)在不同生育期和不同种植密度下冠层光谱反射率及其对应的地上鲜生物量,分析了春小麦地上鲜生物量随生育期的变化以及地上鲜生物量与冠层反射光谱和一阶微分光谱之间的相关关系,采用相关系数较大的特征波段及其组合构建光谱特征参数以其作为变量,建立了春小麦地上生物量的高光谱估算模型,并对模型进行检验。结果表明:以参数F780和D719为变量的对数形式y=3.9498ln F780+7.0596和乘幂形式y=512.99D7191.0174估算水平最高,前者均方根误差(RMSE)为0.2173,相对误差(RE)为10.45%,预测值与实测值相关系数为0.854;后者RMSE为0.2188,RE为9.96%,预测值与实测值相关系数为0.853。因此,上述两个模型可作为陇中黄土高原地区春小麦地上鲜生物量的最佳估算模型。

A field plot experiment was conducted to measure the canopy spectral reflectance and aboveground fresh biomass of four spring wheat varieties （Dingxi 24, Longchun 8139, Gaoyuan 602 and Dingxi 38 ） at their different growth stages and under different planting densities. The variations of the aboveground fresh biomass with growth stages as well as the correlations of the aboveground fresh biomass with canopy reflective spectrum and first derivative spectrum were ana- lyzed, and based on these, hyperspeetral remote sensing estimation models for spring wheat aboveground fresh biomass were established, with the characteristic bands and their combinations strongly correlated with the aboveground fresh biomass as the variables. The tests with experimental data showed that models y =3. 9498 In F780 ＋7. 0596 and y =512. 99 D719^1.0174 had the highest estimation level, with the root mean square error, relative error, and correlation coefficient between estimated and measured values being 0. 2173, 10. 45% and 0. 854, and 0. 2188,9. 96% , and 0. 853, respectively. These two models could be used as the best models for the estimation of spring wheat aboveground fresh biomass on Longzhong Loess Plateau.