| Photosynthesis in aquatic and terrestrial ecosystems is the key component of the food chain and the most important driver of the global carbon cycle. Therefore, estimation of photosynthesis at large spatial scales is of great scientific importance and can only practically be achieved by remote sensing data and techniques. In this dissertation, remotely sensed information and techniques, as well as field measurements, are used to improve current approaches of assessing photosynthetic processes. More specifically, three topics are the focus here: (1) investigating the application of spectral vegetation indices as proxies for terrestrial chlorophyll in a mangrove ecosystem, (2) evaluating and improving one of the most common empirical ocean-color algorithms (OC4), and (3) developing an improved approach based on sunlit-to-shaded scaled photochemical reflectance index (sPRI) ratios for detecting drought signals in a deciduous forest at eastern United States. The results indicated that although the green normalized difference vegetation index (GNDVI) is an efficient proxy for terrestrial chlorophyll content, there are opportunities to improve the performance of vegetation indices by optimizing the band weights. In regards to the second topic, we concluded that the parameters of the OC4 algorithm and similar empirical models should be tuned regionally and the addition of sea-surface temperature makes the global ocean-color approaches more valid. Results obtained from the third topic showed that considering shaded and sunlit portions of the canopy (i.e., two-leaf models instead of single big leaf models) and taking into account the divergent stomatal behavior of the species (i.e. isohydric and anisohydric) can improve the capability of sPRI in detecting drought.;In addition to investigating the photosynthetic processes, the other common theme of the three research topics is the evaluation of "off- the-shelf" solutions to remote-sensing problems. Although widely used approaches such as normalized difference vegetation index (NDVI) are easy to apply and are often efficient choices in remote sensing applications, the use of these approaches should be justified and their shortcomings need to be considered in the context of the research application. When developing new remote sensing approaches, special attention should be paid to (1) initial data analysis such as statistical data transformations (e.g. Tukey ladder-of-powers transformation) and (2) rigorous validation design by creating separate training and validation data sets preferably using both field measurements and satellite-based data. Developing a sound approach and applying a rigorous validation methodology go hand in hand. In sum, all approaches have advantages and disadvantages or as George Box puts it, "all models are wrong but some are useful". |
