乌梁素海湿地芦苇最大羧化速率的高光谱遥感

Maximum carboxylation rates of reed in the Wuliangsuhai wetland using hyperspectral remote sensing

湿地植被生产力和固碳潜力的研究是全球碳循环和全球变化的热点研究问题。湿地植被的光合能力能够指示其生长的健康状态。最大羧化速率是重要的植被光合参数之一,对精确模拟湿地植被光合作用和气体交换模型中的固碳过程具有重要的作用。以内蒙古乌梁素海湖泊湿地为研究区,进行了芦苇叶片光合参数和光谱的测量。芦苇叶片最大羧化速率(V_(cmax))数值是基于Farquhar光合作用模型,从光合测量获取的A-C_i曲线计算并校正到25℃得到的。分别基于bootstrap PLSR模型、单波段和高光谱植被指数(包括简单比值指数SR和归一化差值指数ND),构建湿地芦苇叶片最大羧化速率(V_(cmax))估算模型。基于高光谱遥感图像HJ-1A HSI,采用ND高光谱指数中具有较高V_(cmax)估算精度的入选波段702和756 nm,获取研究区湿地芦苇最大羧化速率空间分布图。研究结果表明,湿地植被光谱特征和高光谱植被指数,可用于估算湿地芦苇V_(cmax),其中最高精度产生于基于bootstrap PLSR模型的建模方法(R^2=0.87,RMSECV=3.90,RPD=2.72),ND高光谱指数的V_(cmax)估算精度高于SR高光谱指数的估算精度;从获取的V_(cmax)空间分布图上提取估算值,其与测量值对比,存在较好的相关性(R^2=0.80,RMSE=4.74)。

Studies on wetland vegetation productivity and their carbon sequestration potential are becoming an important focus of the global carbon cycle and global climate change research. The photosynthetic capacity of wetland vegetation can indicate the health status of its growth. In addition, an accurate estimate of maximum carboxylation rate （ V max） is important for accurately simulating wetland vegetation photosynthesis and carbon sequestration processes with a gas exchange model. Here, the wetland of Wuliangsuhai （Inner Mongolia） was chosen as the study area, and the photosynthetic parameters and spectral reflectance of reed leaves were measured. Based on the Farquhar model of photosynthesis, reed leaf V max values were calculated from A-Ci curves, and subsequently standardized to 25℃. Estimation models of Vcmax for reed leaves in the wetland were constructed with a bootstrap PLSR model and single band and hyperspectral vegetation indices （e.g., simple ratio index （SR） and normalized difference index （ND））. Based on hyperspectral remote sensing images from HJ-IA HSI, the bands of 702 and 756 nm, which had a higher estimation accuracy for Vcmax, were selected from the ND hyperspectra]indices. Subsequently, a spatial distribution map of Vcmax for wetland reed was acquired for the study area. The results showed that the spectral characteristics of wetland vegetation, combined with hyperspectral vegetation indices, could be used effectively to accurately estimate reed Vcmax in the wetland. The highest accuracy was produced from the modeling method based on a bootstrap PLSR model （R2 = 0.87, RMSECV = 3.90, RPD = 2.72）. Furthermore, the accuracy of Vcmax estimations from the ND hyperspectral indices was higher than that from the SR hyperspectral indices. Overall, the estimated values extracted from the spatial distribution map of Vcmax had a good correlation with the measured values （ R2 = 0.80, RMSE = 4.74）