基于模拟Landsat-8 OLI数据的小麦秸秆覆盖度估算

田间秸秆作为农业生产过程中的重要物质,其覆盖度的遥感估算具有十分重要的意义。Landsat-8 OLI影像作为Landsat系列影像的最新数据产品,具有更精细的光谱特征,明确其在秸秆覆盖度估算中的表现具有重要的现实意义。该研究使用ASD Field Spec 4 Hi-Res地物光谱仪,以实测田间小麦秸秆光谱反射率为数据源,模拟Landsat-8 OLI、Landsat-5TM、Aster、Hyperion影像波段反射率,构建光谱指数,并建立小麦秸秆覆盖度估算模型,通过对比分析,评估Landsat-8OLI数据的估算能力。结果表明,基于Landsat-8 OLI1和OLI2波段构建的NDIOLI21指数模型估算结果最优,决定系数(coefficient of determination,R2)为0.60,均方根误差(root mean square error,RMSE)为9.56%,平均相对误差(mean relative error,MRE)为9.83%,优于Landsat-5 TM构建的光谱指数,且仅次于Aster构建的木质素-纤维素吸收指数(lignin cellulose absorption,LCA)和短波红外归一化差异秸秆指数(shortwave infrared normalized difference residue index,SINDRI)以及Hyperion构建的纤维素吸收指数(cellulose absorption index,CAI)。因此,波段更多、波段划分更加精细的Landsat-8OLI构建的光谱指数在小麦秸秆覆盖度估算方面达到了一定精度,具有良好的应用前景。

Estimation of wheat residue cover using simulated Landsat-8 OLI datas

Remote sensing of crop residue cover is very important because of its significance in agricultural production. Hyperion, Aster, and Landsat-5 TM sensors have been successfully used to estimate crop residue cover. However, Hyperion has extended its active duty, almost no good SWIR data of Aster that are useful for identification of crop residue have been acquired since May 2008, and Landsat-5 TM was officially decommissioned on June 5, 2013. Therefore, it is significant to evaluate the performance of Landsat-8 OLI for predicting crop residue cover, which is the latest member of Landsat series and has several advantages over Landsat-5 TM such as the fine spectral resolution. The study was conducted in Fengqiu, which was an area of conservation tillage., the spectra with varying wheat residue covers on the soil surface were measured by using an ASD spectroradiometer (FieldSpec 4 Hi-Res). The measured spectra were used to simulate the corresponding bands of Landsat-8 OLI, Landsat-5 TM, Aster, and Hyperion, which were then employed to construct spectral indices. The potential of Landsat-8 OLI in the prediction of crop residue cover was evaluated by comparing the correlation coefficients between the wheat residue cover and the simulated band reflectance and spectral indices. Moreover, we used the spectral indices to build the calibration models, which were then validated using the measured values of wheat residue cover. Correlation coefficients between the wheat residue cover and the band reflectance of Landsat-8 OLI and Landsat-5 TM were higher in the shortwave infrared region (1 200~2 400 nm), which should be caused by the fact that wheat residue presents a broad absorption feature near 2100 nm associated with cellulose-lignin. Correlation coefficients between the wheat residue cover and the spectral indices using the Landsat-8 OLI1 band (NDIOLI21, NDIOLI31, and NDIOLI41) were better than those for all spectral indices derived from Landsat-5 TM. The correlation coefficient for NDIOLI21 was highest among all indices derived from Landsat-8 OLI (r=0.78, P<0.01), whereas it was lower than that acquired from SINDRI (r=0.88, P<0.01), LCA (r=0.85, P<0.01) or CAI (r=0.85, P<0.01). NDIOLI42, NDIOLI52, NDIOLI53, NDIOLI54, and BIOLI were better in estimating the wheat residue cover than the corresponding spectral indices derived from Landsat-5 TM;NDIOLI21, NDIOLI31, NDIOLI41, and NDIOLI51 were better than all of the spectral indices from Landsat-5 TM;NDIOLI21 was the best index in estimating the wheat residue cover among all indices derived from Landsat-8 OLI and Landsat-5 TM (R2=0.60, RMSE=9.56%, MRE=9.83%), while it was slightly lower than SINDRI (R2=0.75, RMSE=7.98%, MRE=7.82%), LCA (R2=0.67, RMSE=8.85%, MRE=8.54%), or CAI (R2=0.72, RMSE=8.35%, MRE=8.16%). These results indicate that Landsat-8 OLI is able to effectively estimate wheat residue cover, which will have good application prospects in the prediction of crop residue cover.