Aquatic Science

Project title: Enabling Interdisciplinary Aquatic Science Studies from the EMIT Archive

Coastal and fresh water ecosystems are among the most productive ecosystems. They are highly vulnerable to the changing climate (e.g., hurricanes, wildfires, dust storms) and anthropogenic activities (e.g., agriculture, aquaculture) while supporting human lives through their services. For over two decades, satellite systems with ocean color (OC) capabilities have produced valuable science-quality aquatic products (e.g., chlorophyll-a; Chla, total suspended solids; TSS) for studying carbon and biogeochemical cycling in these optically complex aquatic environments. With future planned hyperspectral missions, like the Surface Biology and Geology (SBG) mission, the aquatic remote sensing community will have access to a broader suite of observations across the visible shortwave infrared regions.

Despite all the recent advances in inversion models, retrieving consistent global in-water products in coastal and fresh waters using existing multispectral and hyperspectral remote sensors still presents major challenges. The Earth surface Mineral dust source InvesTigation (EMIT) imaging spectrometer provides an excellent opportunity for calibration and validation of recently developed novel algorithms to accelerate the uptake of SBG observations well in advance. Further, EMIT standard surface mineral maps, in conjunction with historical in-water products from Landsat and Sentinel-2, will enable innovative interdisciplinary case studies, such as assessing the effects of short- or long term atmospheric dust deposition on high-altitude lakes.

By leveraging our existing NASA supports, we propose to refine and adapt our robust processing workflow for EMIT's hyperspectral observations, enabling the generation of a comprehensive product suite for optically deep inland and nearshore coastal waters. We will formulate, implement, and rigorously cross-validate our proven machine-learning (ML) model, termed Mixture Density Networks (MDN), with equivalent products from near-coincident Landsat and Sentinel-2 measurements. Our proposed workflow encompasses four main components. First, we will devise an atmospheric correction method to retrieve hyperspectral aquatic reflectance (rw) using an MDN model trained with in situ rw and simulated top-of-atmosphere reflectance/radiance spectra. Second, we will utilize the derived rw products to retrieve biogeochemical products (e.g., Chla) and hyperspectral inherent optical properties (IOPs), such as phytoplankton absorption spectra (aph). Third, our EMIT products, accompanied by their pixel-level uncertainties, will be extensively cross-validated with products from near-coincident Landsat-8/-9 and Sentinel-2 images. Lastly, we will demonstrate the use of EMIT surface mineral maps along with historical Landsat and Sentinel-2 aquatic products for assessing the impacts of dust deposition and water-level variability in a few lakes in the Western U.S. and the Swiss Alps.

This research directly addresses the objectives outlined in the EMIT SAT request for proposal, which requests submissions to advance the EMIT science objectives, evaluate, and improve existing EMIT data products, and contribute to the SBG mission. Supplemented by other NASA supports, our project will have several far-reaching consequences for future NASA imaging programs. Our products will demonstrate and validate SBG-like products and facilitate research for their synergistic use with those derived from other multispectral or hyperspectral missions in the 2030s. We will further enable advancements in interdisciplinary limnological/geological/atmospheric science via combined, novel use of EMIT's surface mineral maps and historical aquatic products concerning sensitive lake ecosystems.

Project Team

Nima Pahlevan

Science Systems and Applications, Inc.