People

Optically shallow waters (particularly coral reef and seagrass environments) have long been tantalizing targets for satellite ocean color scientists. These environments include habitats that draw tourism, provide shoreline stabilization, and serve as nurseries for commercially and recreationally harvested fishes, among numerous other benefits. Despite their importance, quantitative assessment of these targets using ocean color data is difficult, primarily due to the challenges in separating the benthic signal from that of the water column. Toward that end, a number of analytical models have been developed to deconstruct reflectance spectra into relevant water column and benthic properties (including depth). Using in situ hyperspectral reflectance data, these models can be "solved" for a given spectra through optimization. Recent efforts to apply spectral matching optimization approaches to multispectral satellite data have proven quite fruitful. Unfortunately, however, such applications suffer from the intractable problem of needing to derive too many parameters from too few wavebands. To overcome this, various multi-band approaches either require exogenous inputs (which can be quite rare), or make simplifying assumptions that (1) create uncertainties and (2) reduce the utility of the final results.

Given the hyperspectral capacity of the PACE sensor, the objective of the proposed study is to develop and validate a spectral matching optimization approach for application to the hyperspectral reflectance dataset to be delivered through the upcoming mission. Beyond the traditional products possible from multispectral satellite datasets (i.e., concentrations of major in-water constituents, water depth, and benthic albedo), the products derived using this approach will also include (1) classification of benthic habitats at sub-pixel scale, (2) assessment of condition (i.e., ecosystem health) change over time, and (3) assessment of uncertainties associated with each of these parameters. This work will not start from scratch, but will benefit from previous and ongoing works from the PI and collaborators. The approach will be developed primarily using HICO and PRISM data covering coral and seagrass environments throughout the globe (focus locations offshore Hawaii, Florida, and Queensland, Australia), with validation dataset sources from NASA's CORAL dataset, bathymetric lidar, and previously collected measurements of discrete water samples. In addition to these data, simulated data (through radiative transfer and addition of realistic spectral noise) will also be used to understand algorithm uncertainties. Deliverables will include technical reports and publications as well as algorithms and computer codes available to the NASA Science Data Segment (SDS) team (led by Bryan Franz) to implement and test upon the launch of the PACE mission.

Through development of this approach and eventual implementation to the PACE datastream and processing regiment, numerous benefits are expected to be realized. First, raster bathymetries and benthic classification maps for optically shallow waters will provide critical information to data poor regions throughout the globe, and will also elevate capabilities of heritage and ongoing multispectral satellite ocean color missions. Assessment of benthic condition changes over time will allow for identification of the spatiotemporal extent of impacts from local disturbances (e.g., hurricanes, eutrophication), as well as global environmental responses to climate changes (e.g., coral bleaching). Accurate mapping and monitoring of benthic habitat changes are also one of the key components under the "surface biology and geology" theme recommended by the most recent decadal survey.