基于Sentinel-2影像与PROSAIL模型参数标定的玉米冠层LAI反演

叶面积指数(LAI)与植被光合作用、蒸腾作用、生物量的形成等有密切联系，是玉米长势监测、灾害胁迫监测、产量预测等重要参数之一，也是辐射传输模型、作物生长模型等机理模型的一个重要参数。Sentinel-2卫星是“全球环境与安全监测”计划的第二颗卫星，具有较高的时空分辨率，且具有红边波段，其可见光和近红外波段的分辨率为10m，是农业遥感应用的理想数据源。PROSAIL辐射传输模型是遥感反演玉米冠层LAI的有效途径，然而在反演中存在输入参数不确定性大、调参困难、病态反演、速度慢等问题。模型的参数标定能够获取观测反射率及不确定性范围内的参数取值，提供丰富准确的参数信息，降低模型反演过程中的偏差。为探索参数标定在玉米冠层LAI反演中的应用，研究以Sentinel-2A卫星影像为数据源，使用马尔可夫链蒙特卡洛方法(MCMC)对PROSAIL模型进行参数标定，通过加入5%的观测光谱不确定性，获取各参数在不确定性范围内的后验取值概率分布，以优化反演过程中的参数设置，提高LAI反演精度。研究结果表明：（1）PROSAIL模型对可见光和近红外波段较为敏感的输入参数有LAI、叶片叶绿素含量及结构系数，将此三个参数作为查找表反演中的可变参数能够有效地进行LAI的反演，反演精度的决定系数达0.7以上。（2）MCMC方法能够对PROSAIL模型进行参数标定，获取研究区内玉米各参数取值分布信息，参数后验分布与实际情况接近，表明利用MCMC方法进行参数标定可行有效。（3）通过参数标定可以有效提高LAI的反演精度，在降低反演偏差和异常值方面尤为明显，参数标定优化后的反演平均偏差由原先的20%降低至8%，同时估算精度由76%提高至90%。研究结果表明：利用MCMC进行PROSAIL模型参数标定，能够提高PROSAIL模型的LAI反演精度，降低反演偏差，为利用PROSAIL辐射传输模型提高作物冠层参数反演精度提供借鉴。

Retrieving Corn Canopy Leaf Area Index Based on Sentinel-2 Image and PROSAIL Model Parameter Calibration

Leaf area index (LAI) is related to photosynthesis, transpiration and biomass accumulating processes of vegetation closely. It is one of the important parameters of corn growth monitoring, disaster stress monitoring and yield prediction, as well as an important parameter of the radiative transfer model and crop growth model. Sentinel-2satelliteis the second satellite of the Global Monitoring for Environment and Security (GMES) plan. It has high spatial and temporal resolution, and visible and near-infrared bands resolution is 10 m, so sentinel-2 satellite is an ideal data source for agricultural remote sensing applications. The PROSAIL radiative transfer model is an effective way to retrievecorn canopy LAI using remote sensing images. However, there are some problems for LAI retrieval currently, including uncertainty of model inputs, difficulty in parameter adjustment, ill-posed inversion and low speed etc. Model inputs calibration can be used to acquire the exact value of model inputs in the uncertainty range of the observed reflectivity. Rich and accurate parameter information is provided to reduce the errors in parameter retrieval. In this paper, a sentinel-2A satellite image was used as the data source, and Markov Chain Monte Carlo (MCMC) method was used to calibrate model inputs. The spectral reflectance uncertainty of 5% was added to obtain the posterior value probability distribution of each parameter, to optimize the parameter setting in the retrieval process and improve the accuracy of LAI retrieval. The results showed that: (1) The sensitive model inputs of the PROSAIL model were LAI, chlorophyll content of leaves and leaf structure coefficient within visible and near-infrared bands. Taking these three parameters as variables in the look-up table could effectively retrieve LAI, and the determination coefficient of retrieval accuracy reached 0.7. (2) MCMC method could be used to calibrate the PROSAIL model input and acquire each model input distribution in the study area. The posterior parameter distribution was close to the actual situation, indicating the feasibility and effectiveness of using the MCMC method for parameter calibration. (3) Input calibration could effectively improve the LAI retrieving accuracy, especially in reducing retrieval deviation and outliers. After inputs calibration, the average error of LAI retrieval decreased from 20% to 8%, while the estimation accuracy increased from 76% to 90%. These results showed that the model inputs calibration of the PROSAIL model by MCMC could improve the LAI retrieval accuracy and reduce the retrieval deviation. It provided a reference for improving the retrieval accuracy of crop canopy parameters by the PROSAIL radiative transfer model.