Carbon Flux In An Enstuarine Wetland Estimated By Remote Sensing Model And Ground-based Observations

The carbon cycle of terrestrial ecosystem is an important part of global climate change study. Estuarine wetland, as an important type of terrestrial ecosystems, has its unique material cycle and energy budget and plays an important role in the carbon cycle. Estuarine wetland also shows high sensitivity to global climate change and is vulnerable to natural or anthropogenic disturbance. Compared to other terrestrial ecosystems, rare research has been done on carbon cycle of the estuarine wetland, especially on its processes and driving factors. Nowadays, eddy covariance (EC) technology has provided continuous measurement on carbon exchange at ecosystem level. However, the measurement can only represent the carbon flux at scale of tower footprint. On the other hand, remote sensing technology provides the real-time surface information and shows potential on estimating the ecological parameters and carbon flux at regional or global scale. Based on the continuous measurement of three carbon flux towers established in Dongtan of Chongming Island since 2004, the impact factors of carbon cycle (micrometeorological factors and tidal flooding) and the estimation of carbon flux were explored by using remote sensing. Therefore, the aims of this dissertation is to 1) examine the relations between main micrometeorological factors and vegetation indices (VIs) estimated from remotely sensed data, and develop the inversion models of these parameters; 2) detect the spatiotemporal changes of tidal flooding inundation on estuarine wetland and evaluate the vegetation successional stages using remotely sensed time series data; 3) close the carbon budget of estuarine wetlands by coupling tower-based measurements and remotely sensed data; 4) to estimate the net ecosystem carbon exchange (NEE) along the tidal inundation gradient and explore the effect of tidal flooding on the ecosystem carbon cycle. The major findings are summarized as follows:1) We analyzed the correlations between VIs and micrometeorological factors at the three towers in 2005 and found that the surface spectral reflectance and VI reflected the spatiotemporal characteristics of micrometeorological factors very well in our study area, and succeeded in capturing the spatial heterogeneity at three towers (site D, M and S). VIs are potential for the inversion of the micrometeorological factors (e.g. VPD, VWC, soil temperature) and different VIs are chosen for each site due to the difference in vegetation composition and cover, and the distance from the sea. For example, the enhanced vegetation index (EVI) and the surface water index (LSWI) is the main explanatory variables at the offshore site (site S), and LSWI is the main explanatory variable at the site far from sea (site D). Our result suggests that the different VIs should be considered to estimate micrometeorological factors at different sites when remotely sensed data are used.2) A wavelet analysis was performed on time series VIs derived from remotely sensed data to investigate the spatiotemporal dynamic of the tidal flooding inundation in tidal marsh. The results show that MODIS is suitable for monitoring tidal flooding inundation from the dam to the sea front. However, due to the complex background of estuarine wetland, EVI can well detect the vegetation dynamics but cannot exclude the effect of water content, and LSWI can well detect the water content but cannot exclude the effect of vegetation. Therefore, on the basis of the time series analysis of vegetation indices, vegetation-water index (VWI) that merges vegetation and water information was developed to track the vegetation succession in estuarine wetland and the yearly value of VWI is nore more effective. For example, at the primary successional stage, VWR is less than 0.25; at the pioneer stage; VWR is between 0.25 and 0.75; at development stage, VWR is between 0.5 and 2.0.3) Compared to other ecosystems, estuarine wetland shows distinct carbon flux dynamics—the lateral carbon flux incurred by tidal flooding and methane generation under anaerobic conditions of wetland soils. In this study, we estimated the 2005's annual carbon budget of an estuarine wetland on Chongming Island and partitioned the losses of carbon due to lateral tidal dynamics and anaerobic methane production using innovative technique. The average GPP calculated from eddy covariance between March and November was 5.65 g C m^-2 day^-1, whereas that from the LUE model was 1.27 g C m^-2 day^-1. The correlation coefficient between GPP simulated from the LUE model and that calculated from flux tower data was low in the growing season (R²=0.55). We further hypothesized that tidal activities and uncounted methane release were responsible for the discrepancy, which partly can be predicted from measurements of LSWI, evapotranspiration (ET) and tide height (TH). We developed an integrated GPP model by combining the LUE model and an autoregression model. The average GPP from the integrated model increased to 5.69 g C m^-2 day^-1, and R² for the correlation between the simulated and calculated data increased to 0.88, demonstrating the potential of our technique for GPP estimation and quantification of seasonal variation in estuarine ecosystems.4) To find how the spatial pattern of NEE varies along tidal inundation gradient, we coupled MODIS data and the tower-based NEE measurement to develop a NEE estimation model by using a piecewise regression model. The results show that the model gives a fairly good prediction of NEE (R²=0.78, p<0.01). We then applied the model to estimate NEE for each 500x500m cell across the transect covering 500 m×3000 m in 2005～2006. Our empirical model captures the expected spatiotemporal patterns of NEE along the tidal inundation gradient. The variations of NEE near the island-side are mainly caused by seasonal shift and yearly cycle of vegetation, whereas in the ocean-side, NEE is more influenced by tidal activities, and in the middle area, NEE is subjected to both phenological conditions of vegetation and tidal cycles. We estimated that the average NEE varies from -2.4 to -1.8 g C m^-2 d^-1 along the tidal inundation gradient from island-side to ocean-side. In conclude, this study illustrates that our NEE estimation model derived from MODIS and tower-base flux data is effective for estimating NEE in the similar ecosystem, and the estimates are useful for analyzing the spatiotemporal pattern of NEE and the impacts of climate variability and disturbances on estuarine wetland carbon fluxes.5) Coupling the observed data of carbon flux towers, the comprehensive research was conducted to analyze the feasibility of remote sensing data in the evaluation and estimation of the carbon flux in the estuarine wetland, and then the improved models were developed and demonstrated the great potential for correcting unavoidable errors when estimating carbon budget of the estuarine wetland. Our study can be used for reference to study the carbon flux in the similar ecosystem as the estuarine wetland in the context of the global climate change.