Integration of satellite and in situ observations to estimate spatial patterns of carbon dioxide, latent and sensible heat in coastal saltmarsh wetlands using the SCOPE model

Background/Question/Methods ：Quantifying spatial patterns of carbon dioxide (CO₂), latent and sensible heat (LE, H) flux in coastal wetlands is important for a better understanding of the role of coastal biosphere in global carbon and water cycle. Remotely sensed data have been considered as important inputs for spatial land surface models to estimate the surface CO₂, LE and H fluxes. The potential of remote sensing data is, however, not yet fully exploited due to too coarse spatial resolution of most satellite data for land surface models, and the indirect relation between radiative transfer parameters and process model parameters. Advances in coherent models promote effective use of the available remote sensing data to reveal land surface processes through observed radiance spectra. The SCOPE model represents one of the currently leading coherent models that integrates land-atmosphere flux exchanges and radiative transfer. In this study, we developed a methodology to combine the virtues of field data, multispectral satellite observation, and SCOPE modelling, for quantitative estimates of spatial patterns of gross primary productivity (GPP), LE and H fluxes in coastal saltmarsh wetlands. More specifically, we used field measurement sets, which include hyperspectral reflectance of the soil and dominant vegetation and eddy covariance (EC) data, to produce Look Up Tables (LUTs) through SCOPE simulations. The next step is to retrieve the spatial variations of fluxes from multispectral satellite images (e.g., Landsat, Sentinel-2) based on the LUTs. This methodology is illustrated by the application to a coastal saltmarsh wetland of the Yangtze Estuary in China. The simulated time series of GPP, LE, and H fluxes were further calibrated and validated by the EC data.

Results/Conclusions: The results show that the correlations between the modelled and observed values are larger than 0.8 in about 80% of cases, with the estimates of H and GPP more accurate in relative to that of LE. Furthermore, the estimated flux maps from Landsat 7 ETM+ and Sentinel-2 MSI images on 29 August 2016 were compared for mutual authentication. These assessments show that the method we put forward is able to reliably estimate spatial patterns of GPP, LE, and H in general, although the retrievals of vegetation parameters are to some extent ill-posed. This study offers a useful tool to quantify present/future carbon and water availability, especially for small geographic areas such as coastal wetlands, through combining coherent model and present/future optical satellite data with high spatiotemporal resolution.