融合遥感与涡度相关技术的毛竹林碳通量季节动态监测及其区域扩展

Moso bamboo forest is one of the main vegetation types in the subtropical area of China, which plays an important role in sequestrating CO2 by forests and maintaining the regional carbon balance. However, the researches on the carbon source/sink potential, the underlying mechanisms and processes, the response to climate change, as well as the spatiotemporal patterns of the Moso bamboo forest are still scarce. Therefore, the upriver region of Taihu lake basin, which is one of the major distribution areas of bamboo, was taken as the typical study area. Based on the continuous observation of carbon fluxes and the main components using eddy covariance technique, the seasonal variations in NEE, GEE and Re on different time scales were represented and the responses of biotic and abiotic factors controlling CO2 exchange were also revealed. Then, after analyzing the correlations between these time-series remotely sensed products and the observed NEE, the predictive models including multiple linear regression, rule-based piecewise regression, and neural network improved by genetic algorithm were developed, respectively. And the optimal model with the maximum coefficient of determination (R2), and the minimal root mean square error (RMSE) was selected to predict the temporal changes of carbon fluxes. At last, based on the Moso bamboo forest information extracted using the object-oriented method, the data-driven satellite-based NEE diagnosed model was used to upscale eddy-flux observations to landscape scales and the regional carbon balance was quantitatively evaluated. This research has a great theoretical and practical significance, which not only offers a framework for quantitative assessment of the carbon budget in terrestrial ecosystems at regional scale, but also provides the scientific basis for developing the carbon sequestration potential of Moso bamboo forest and serving for local energy saving and emission reduction.

毛竹林是我国亚热带地区重要的植被类型之一，在森林生态系统CO2减量及维持区域碳平衡方面有着重要的贡献。然而当前毛竹林生态系统碳源/汇潜力、过程机制、对气候变化的响应及区域尺度下的碳收支时空格局研究还很缺乏。因此，本项目拟以我国重要的竹林分布区——太湖上游地区为典型研究区，基于涡度相关技术获取连续观测的碳通量数据及拆分生成的组分数据，在不同的时间尺度上分析其动态变化规律，并揭示其主要环境控制因素的响应特征；融合时间序列遥感产品与碳通量数据集，对比多元线性回归、规则支持下的分段回归及遗传算法优化的神经网络模型预测效果，建立毛竹林碳通量季节动态监测最佳遥感反演模型；而后在毛竹林专题信息精确提取的基础上进行尺度扩展，定量评价其区域碳收支现状。本研究不仅可以为区域尺度陆地生态系统碳收支格局定量评估提供方法参考，还可以为发挥毛竹林固碳增汇潜力、服务地方节能减排提供数据支撑，具有重要的理论与现实意义。

该项目建立了我国第二套（江苏省第一套）针对典型陆地生态系统毛竹林的碳水通量涡动相关观测系统，初步分析了毛竹林碳通量在不同时间尺度上的变化特征。日尺度上，毛竹林在白天均呈现强烈的碳吸收能力，而在夜间受呼吸作用控制，主要表现为碳源。毛竹林一年中大部分时间为碳汇，春季固碳能力最强，其次为夏季，而秋季易受年度水热条件限制。毛竹林比一般林地具有更高的固碳能力。同时发现，应用MODIS反演的环境参数直接反演NEE不具备可行性，可通过分别估算GPP与Re，再间接估算NEE。2015年该站点年均固碳量实测值为1.36 g C m-2 d-1，利用间接建立的遥感估算模型计算的年均值为1.53 g C m-2 d-1，仅高估了约13.3%，达到精度要求，可用于区域扩展。而后，应用面向对象分类方法获取毛竹林空间分布信息，推出该区域年总固碳量约为0.35 Tg。 本项目共发表SCI论文10篇，超额完成考核指标，部分研究成果被“中国科学报”、“中国青年网”等媒体报道。