基于物理模型与查找表策略的南繁水稻叶片叶绿素含量反演研究

Retrieval of leaf chlorophyll content in southern breeding rice leaves based on physical model and look-up table strategy

关键词:  南繁水稻  叶片叶绿素含量  PROSAIL-PRO  查找表  植被指数

Key words:  southern breeding rice  leaf chlorophyll content  PROSAIL-PRO  LUT  vegetation index

摘要:

[目的] 利用海南省中南部5市县采集的南繁水稻光谱、叶绿素等数据，研究物理模型与查找表策略对南繁水稻叶片叶绿素的反演，并对不同查找表进行比较分析。[方法] 文章先根据研究区调查与文献调研，利用辐射传输模型PROSAIL-PRO模拟2万条冠层光谱，分别构建波长400～1 000 nm区间的全波段查找表和双植被指数（Vegetation Index，VI）查找表，并使用实测的南繁水稻冠层光谱与叶绿素含量进行验证。其中，双VI查找表包括TCARI-OSAVI查找表、比值型植被指数（Ratio Vegetation Index，RVI）查找表和红边相对指数（Red Edge Relative Indices，RERI）查找表3种。[结果] （1）传统的全波段查找表反演叶片叶绿素含量的精度较差，R值为0.46，均方根误差（Root Mean Squared Error，RMSE）为8.84 μg/cm²，正规化均方根误差（Normalized Root Mean Squared Error，NRMSE）为 0.32。（2）双VI查找表反演结果较传统查找表更好，TCARI-OSAVI查找表的R值为0.55，RMSE为8.16 μg/cm²，NRMSE为0.29；RVI查找表的R值为0.70，RMSE为8.80 μg/cm²，NRMSE为0.32；RERI查找表的R值0.63，RMSE为10.53 μg/cm²，NRMSE为 0.38；其中TCARI-OSAVI查找表反演效果最好，RVI查找表与RERI查找表的反演都存在高估的现象；与前人关于冬小麦的叶片叶绿素反演研究相比，双VI查找表在南繁水稻的叶片叶绿素反演方面的NRMSE相对更高，尽管如此，与其他水稻叶绿素反演的研究相比，该文所用方法具有更高的可靠性与可迁移性。[结论] 双VI查找表相对于传统查找表具有一定的优势，在南繁水稻叶片叶绿素含量反演方面具备应用的潜能，然而该方法仍有一定的改进空间。

Abstract:

[Purpose] Using the in-situ data collected from 5 cities/counties in south-central Hainan Province，a physical model and look-up table（LUT）strategy are developed to invert the leaf chlorophyll content（LCC）of rice in southern breeding base and a comparative analysis of LUTs among different species is conducted.[Method] According to the study area survey and literature review，20 000 canopy spectra were simulated by the radiative transfer model PROSAIL-PRO. Full-band LUT and dual-Vegetation Index（VI）LUTs were constructed in the wavelength range from 400-1000 nm respectively，and were validated using in-situ canopy spectra and leaf chlorophyll content of southern breeding rice. The dual-VI LUTs included 3 types：TCARI-OSAVI LUT，Ratio Vegetation Index（RVI）LUT，and Red Edge Relative Indicators（RERI）LUT.[Result] The results showed that the LCC inversion using the traditional full-band LUT was relatively poor，with an R value of 0.46 and a Root Mean Squared Error（RMSE）of 8.84 μg/cm²，Normalized Root Mean Square Error（NRMSE）was 0.32. The dual-VI LUTs performed better than the traditional LUT. The R value of the TCARI-OSAVI was 0.55，and the RMSE was 8.16 μg/cm²，NRMSE was 0.29；The R value of the RVI LUT was 0.70，and the RMSE was 8.80 μg/cm²，NRMSE was 0.32；The R value of the RERI LUT was 0.63，and the RMSE was 10.53 μg/cm²，NRMSE was 0.38. Among which the TCARI-OSAVI LUT had the best inversion effect，and the inversions of both RVI LUT and RERI LUT were overestimated. Compared with previous studies on winter wheat LCC retrieval，the NRMSE of dual-VI-LUT-based LCC retrieval of southern breeding rice was relatively higher. However，this method had higher reliability and transferability compared with other studies on rice LCC retrieval.[Conclusion] This study demonstrate the advantages of the dual-VI LUTs over traditional LUTs and their potential in the LCC inversion of southern breeding rice，yet there is still room for improvement.