)Coastal regions are increasingly experiencing degraded water quality, affecting recreation, commerce, and human health. Water clarity in particular can have cascading impacts, from the function of the ecosystem itself (i.e., sea grass and phytoplankton growth) and either encouraging or deterring recreation. Here, we have applied a technique to expand the coverage of observations of water clarity by using non-Euclidean water distance-based kriging. Through many iterations and test configurations, the interpolation expands the spatial coverage of clarity estimates by multiple orders of magnitude with relatively high accuracy. Hold out data for validation of daily estimates of the diffuse attenuation coefficient, Kd, had an R2=0.55 and a bias of 4%, while for Secchi depth R2=0.71 and bias was 4%. These new data products allow for the potential assimilation with models that utilize Kdas a variable, integration with machine learning by gap filling in time and space, and providing labeled data for model calibration and validation.

Computationally intensive artificial intelligence algorithms trained with many sources of observations have the potential to detect impairments to water quality not previously possible through traditional techniques. State agencies in Maryland and Virginia who manage resources in the Chesapeake Bay on the east coast of the United States State currently sample 800 sites around the Chesapeake Bay in boats every month, missing harmful blooms or runoff between their point source observations in space and time. To support their information needs, we exploit multiple satellite data sets of different spatial, spectral, and temporal resolution within our machine learning architecture, entitled ‘deep learning for environmental and ecological prediction, evaluation, and insight with ensembles of water quality’ (DEEP-VIEW). Through fusion of many sources of local and remotely sensed data, we aim to improve predictions of coastal water quality impacted by runoff from land, changes in clarity, harmful blooms, and other indicators needed for resource management of aquaculture, swimming beaches, and boating.

Previous work on predicting water quality indicators has mainly consisted of using both semi-analytical algorithms (SAAs) and empirical approaches, but recently new data-driven machine learning approaches such as neural-network-based regression models are increasingly being explored for their utility and potential adoption. Although these types of data-driven models may achieve higher accuracy compared to previous methods, they can also be prone to biasing their outputs towards the mean value of the target distribution if model inputs are noisy. This paper investigates using a recently published weighted regression approach to alleviate "mean-centric" bias on these types of water clarity estimators in the Chesapeake Bay. Experiments comparing standard and weighted data-driven regression approaches for Chesapeake Bay Secchi disk depth prediction are performed and results are discussed.

Pursuant to recommendations by the National Academies of Science, Engineering and Medicine's Earth Science Decadal Survey [1], the National Aeronautics and Space Administration (NASA) has announced the development of an Earth System Observatory (ESO), a series of missions designed to observe processes across the Earth's interior, surface and atmosphere. A key component of this system is the Surface Biology and Geology (SBG) investigation. SBG will measure the composition and properties of Earth's land, inland waters, and coastal oceans. The notional architecture consists of multiple spacecraft slated for launch in the 2027–2028 timeframe (Figure 1). Target science questions and geophysical variables span diverse disciplines including terrestrial and aquatic ecology, geology, vulcanology, hydrology and cryospheric sciences (Figure 2). Beyond simply measuring geophysical variables for each discipline, SBG will provide information about the links between the different domains, enabling a more comprehensive understanding of the Earth as a connected system. SBG measurements will also benefit a wide range of societal applications including agriculture, terrestrial and aquatic biodiversity, natural hazards, public health, and management of water and other natural resources [2]. SBG will also coordinate measurements, data products, and analyses with other ESO elements to deliver an integrated Earth System perspective of Earth and its changing climate.

This paper examines in situ water quality data measured during 2020–2021 in the Chesapeake Bay for comparison with optical satellite data. This collection was performed as part of a NASA project aiming to develop new methods for water quality monitoring from satellite remote sensing using artificial intelligence. Our objective is to use in situ data as ground-truth to provide water quality classifications, or labels, to their overlapping (in time and location) satellite imagery. Having such labeled data, can help us achieve our project's longer-term goal: to train artificial intelligence models to recognize features in spectral information for monitoring water quality from satellites. Because routine monitoring by state agencies is conducted at discrete locations, we obtained a flow-through system operated from small boats to measure water quality parameters along transects for comparison with two-dimensional maps collected from space, with an initial focus on low oxygen events, due to their large spatial extent and regular occurrence each summer. We also evaluated similar in situ data collected during 1984-2021 by the Chesapeake Program.

On Jan. 5, 2018, at the request of NASA, the National Oceanic and Atmospheric Administration (NOAA) and the U.S. Geological Survey (USGS), the Committee on the Decadal Survey for Earth Science and Applications from Space (ESAS) of the National Academies of Sciences, Engineering and Medicine (NASEM) Space Studies Board, Division on Engineering and Physical Sciences released the 2017 Decadal Survey, “Thriving on Our Changing Planet: A Decadal Strategy for Earth Observations from Space” [1]. The 700-page document is the second such Earth sciences survey produced by NASEM. The first, “Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond,” was released in 2007. The 2018 study designated a global “Surface Biology and Geology” (SBG) investigation that would include both imaging spectroscopy and thermal infrared observations [1]. This suite of measurements would address a wide range of global science questions. Its themes include: flows of energy, carbon, water, and nutrients sustaining terrestrial and marine ecosystems; the variability of the land surface and the fluxes of water and energy; inventory of the world's volcanoes, and the composition and temperature of volcanic products immediately following eruptions; other natural hazards including wildfires; snow accumulation and melt; water balance from the headwaters to the continent; land and water use effects on evapotranspiration; functional traits and diversity of terrestrial and aquatic ecosystems and vegetation; and more. Figure 1 shows example spectra from these surfaces, illustrating the enormous diversity of scene content that would be observed. Tables 1 and 2 show examples of the core and higher-level products that the SBG mission would produce.

The National Academies Decadal Survey for Earth Science recommended that NASA pursue global imaging spectroscopy and thermal infrared measurements in the coming decade [1]. Both measurements would offer repeat coverage on approximately five-day to biweekly cadence, with comprehensive coverage of the globe's coastal and terrestrial area. This would be an unprecedented volume of data with the potential to transform remote sensing practice. To address this recommendation, NASA has sponsored a concept study by NASA research centers and associated university partners (https://sbg.jpl.nasa.gov). This study is determining a family of architecture options - including launch vehicle, spacecraft, instrument, and suborbital components - that could address the Decadal Survey objectives. The architecture study is driven by science needs and builds on input of the research community. As of this writing, the study is entering a phase in which a large field of system possibilities is pared down to a representative handful for an ultimate decision by NASA.

Reliable information on water quality is not currently available at the space and time scales that are required for aquaculture and other resource management needs. For example, shellfish growing areas may be impacted by harmful algal blooms or runoff from land that increases turbidity, lowers salinity, or introduces contaminants. Shellfish resource managers in the Chesapeake Bay are especially concerned with sources of bacteria from land such as failing onsite waste systems, failing wastewater infrastructure, and concentrated animal feeding operations. There is an urgent need for remote sensing of water quality indicators beyond chlorophyll-a and suspended sediments to augment field sampling programs. Artificial Intelligence trained with simultaneous in situ and satellite observations is explored in preparation for future hyperspectral satellite missions, which offer potential to detect additional water quality indicators not previously possible. This first step identifies and develops a method to harmonize disparate, unlinked aquatic datasets to derive information about where water quality is likely degraded.