ABSTRACT Inversion on the basis of radiative transfer model (RTM) is the prevailing approach for obtaining the leaf area index (LAI) from satellite images. Ill-posed inversions are a problem in RTM-based methods, however, and so here the vegetation isoline pattern in red and near-infrared spectral space was used, with consideration of the fraction of vegetation cover (fCover), to develop a look-up table (LUT) for the LAI inversion of rice fields. In PROSAIL (PROpriétésSPECTrales and Scattering by Arbitrarily Inclined Leaves) model simulations, to avoid some unreasonable parameter combinations, the values of the input parameters were set with some a priori knowledge, and 3580 parameter combinations were generated in the LUT. This represents much fewer combinations than for a conventional LUT. Comparison tests demonstrated that the small-size LUT built with prior knowledge did not decrease the accuracy of the inversed LAIs; rather it improved the accuracy by taking into account fCover. The proposed LUT was applied to the images captured by the wide field of view (WFV) cameras loaded on the Gaofen-1 (GF-1) satellite. Evaluation of the inversed LAIs using in situ data showed that the root-mean-square error (RMSE) was 0.37, and that the relative error (RE) was 14%. Comparison with the error of inversed LAIs produced by the LUT without considering fCover revealed that taking into account the fCover when building a LUT based on the vegetation isoline pattern improved the accuracy of the LAI inversion. This study demonstrates that vegetation isoline-based LUT, with consideration of fCover, is a promising technique for producing LAI maps of crops with high spatial resolution. It will also be helpful for the validation of moderate-resolution LAI products and agricultural monitoring.

中文翻译：

---

考虑植被覆盖率的植被等值线模式反演稻田叶面积指数

摘要 基于辐射传输模型 (RTM) 的反演是从卫星图像中获取叶面积指数 (LAI) 的主流方法。然而，不适定反演是基于 RTM 的方法中的一个问题，因此这里使用红色和近红外光谱空间中的植被等值线模式，并考虑植被覆盖率 (fCover)，以开发外观-稻田 LAI 反演的上表 (LUT)。在PROSAIL（PROpriétésSPECTrales and Scattering by Arbitrary Inclined Leaves）模型模拟​​中，为了避免一些不合理的参数组合，输入参数的值设置了一些先验知识，在LUT中生成了3580个参数组合。这代表的组合比传统 LUT 少得多。对比测试表明，使用先验知识构建的小尺寸 LUT 不会降低反向 LAI 的准确性；相反，它通过考虑 fCover 提高了准确性。建议的 LUT 应用于加载在高分一号 (GF-1) 卫星上的宽视场 (WFV) 相机捕获的图像。使用原位数据对反向 LAI 进行评估表明，均方根误差 (RMSE) 为 0.37，相对误差 (RE) 为 14%。与不考虑 fCover 的 LUT 产生的反演 LAI 误差的比较表明，在基于植被等值线模式构建 LUT 时考虑 fCover 提高了 LAI 反演的准确性。这项研究表明，基于植被等值线的 LUT，考虑到 fCover，是一种很有前途的技术，可以生成具有高空间分辨率的作物 LAI 图。它也将有助于中等分辨率 LAI 产品的验证和农业监测。