首页 > , Vol. , Issue () : -
精精准定量反演光合植被(PV)和非光合植被(NPV)覆盖度对了解植被碳循环过程起着至关重要的作用,同时非光合植被覆盖度信息获取为土地沙漠化及植被转化机制研究提供重要信息。本文以甘肃省民勤县为研究区, 以Sentinel- 1B IW GRD和Sentinel-2A为数据源, 采用线性指数模型和随机森林回归模型(RFM), 基于控制变量法开展微波与光学遥感数据协同反演NPV覆盖度的方法研究, 参考野外获取样地真实性检验数据, 将均方根误差(RMSE)和相对均方根误差(RMSE(%))作为指标评价反演结果精度。结果表明: (1) Sentinel-1和Sentinel-2协同反演NPV比仅采用sentinel-2光学遥感数据能够明显提高NPV覆盖度估算精度; (2) 由Sentinel-1和Sentinel-2获取植被指数构建的随机森林模型在NPV覆盖度估算上较线性指数模型精度更高, 随机森林模型和线性指数模型估算NPV的RMSE分别为0.0149和0.0153, 估算精度提高了1.4%; (3) VH、VV两种极化方式参与建立随机森林模型可有效提高NPV覆盖度的估算精度, 尤其VH极化对非光合植被信息探测更为敏感, 较VV模型估算精度提高了5.1%。(4) 加入表征土壤信息的比值土壤指数（RSI）有效减少土壤对在NPV覆盖度估算影响，提高NPV覆盖度估算精度。由此可见, 微波和光学遥感数据结合是提高NPV覆盖度估算精度的有效方法，同时土壤作为独立重要指标参与模型计算对提高NPV覆盖度估算具有重要意义。
Abstract: Quantitatively estimated fractional cover of photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and bare soil (BS) play an important role in establishing carbon dynamics models. And it provided the important information for the study of land desertification and vegetation transformation mechanisms by accurately obtained the fractional cover of non-photosynthetic vegetation. Although some progress has been made in obtaining NPV fractional cover (fNPV) by optical remote sensing in previous studies, there are still many interfering factors and difficulties. In order to further improve the accuracy of NPV fractional cover estimation, we will try to combinate microwave and optical remote sensing information to obtain NPV fractional cover. In this paper, taking the Minqin County in Gansu Province as the research area, Sentinel-1B IW GRD and Sentinel-2A were applied on the study as data sources. The experiments employed the control variable method with linear index model and random forest regression (RFR) model to research on NPV fractional cover estimation by microwave and optical remote sensing data. Then, the estimated endmember fractions were later validated with reference to fraction measurements. In addition, the Root Mean Square Error (RMSE) and Relative Root Mean Square Error (RMSE%) were employed as indicators to evaluate the inversion accuracy. The result shows: (1) Cooperative Sentinel-1 and Sentinel-2 remote sensing data to estimate NPV fractional cover could effectively improve the estimated accuracy rather than only by Sentinel-2 data . (2) The random forest regression model is an effective method for the sparse NPV fractional cover estimation, and its estimation accuracy is higher than the linear index model. The validation RMSE of the random forest model and the linear index model estimated fNPV are 0.0149 and 0.0153, respectively. Obviously, the accuracy for fNPV estimation is increased by 1.4% with RFR instead of linear index model. (3) The VH and VV polarization bands of Sentinel-1 data could effectively detect the characteristics of NPV. Especially, VH band is more sensitive to NPV, their estimation accuracy is improved by 5.1% compared with VV band participating. (4) The accuracy of fNPV estimation could be improved when soil index participated in each model, which illustrate the soil characteristic information participated in the models is important for NPV extraction. Above all, it can be seen that the combination of Sentinel-1 and Sentinel-2 remote sensing data could effectively improve the accuracy of NPV fractional cover estimation employed the Random Forest Regression model. Within VV and VH polarization modes are sensitive to NPV vegetation detection, especially, VH polarization mode. Furthermore, the accuracy of NPV extraction could be further improved by adding the soil index, which reflected the soil characteristics. Therefore, the combination of microwave and optical remote sensing data is an effective method to improve the accuracy of fNPV estimation. The participation of polarization information with vegetation structure information and soil parameters with soil characteristic information are of great significance to improve the accuracy of NPV fractional cover estimation.