A document by Natalie Nelson. Click on the document to view its contents.

A document by Natalie Nelson. Click on the document to view its contents.

Data-driven modeling approaches for crop yield prediction have exponentially increased in the last decade due to the greater availability of spatial data from various sensors. Yet, most yield modeling has focused on major commodities, leaving lesser-cultivated horticultural crops like sweetpotato relatively undertooled, though these crops considerably contribute to the global economy and food supply. The U.S. is the primary exporter of sweetpotato (271 K tonnes), with 21% of U.S.-grown sweetpotatoes being exported. Early yield forecasting at the county scale offers crucial insights for growers, packers, wholesalers, and associated industries, enabling them to anticipate variations in yield to make informed decisions. While roots and tubers have demonstrated a relationship between yields and above-ground plant characteristics, it remains uncertain whether forecasting models that utilize remotely sensed data, including vegetation indices, are suitable for sweetpotato. We developed county-scale in-season sweetpotato yield forecast models using machine learning (ML) algorithms and multitemporal remote sensing environmental data. Four of the most commonly used ML algorithms for predicting crop yield-Random Forest Regression (RFR), Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB)-were applied using stationary (topography and soil characteristics), and temporal (weather, NDVI, and Growing Degree Days) variables as potential predictors. Six predictor sets were tested to identify key predictor variables, optimal aggregation time (16 or 32 days composite) of the temporal variables, and how early in the growing season the models can reliably predict end-of-season yields. U.S. Annual CropScape land cover layers were used to identify sweetpotato fields, over which temporal variables were aggregated, and sweetpotato yields were tabulated from the USDA Agricultural Survey from 2008 to 2022. The Boruta method was used for feature selection across each predictor set before training the ML models. RFR outperformed other ML algorithms and the RFR models' evaluation metrics were the most consistent across the six predictor sets. The RFR model that incorporated early and mid season temporal variables as 16-day composites was selected and proposed for future sweetpotato yield forecasting due to its performance (R 2 = 0.44, RMSE = 3.53 tonnes.ha-1), as well as ability to predict early enough 1 in the season to provide actionable information. In the final model, several stationary variables (elevation, nitrogen, cec, soc, and clay content) were the most predictive of sweetpotato yield. After these stationary variables, NDVI and precipitation from the time around storage root initiation and bulking (July), and minimum temperature around planting (June) followed in importance.

ABSTRACT Regular testing of coastal waters for fecal coliform bacteria by shellfish sanitation programs could provide data to fill large gaps in existing coastal water quality monitoring, but research is needed to understand the opportunities and limitations of using these data for inference of long-term trends. In this study, we analyzed spatiotemporal trends from multidecadal fecal coliform concentration observations collected by a shellfish sanitation program, and assessed the feasibility of using these monitoring data to infer long-term water quality dynamics. We evaluated trends in fecal coliform concentrations for a 20-year period (1999-2021) using data collected from spatially fixed sampling sites (n = 466) in North Carolina (USA). Findings indicated that shellfish sanitation data can be used for long-term water quality inference under relatively stationary management conditions, and that salinity trends can be used to measure the extent of management-driven bias in fecal coliform observations collected in a particular area. 1. INTRODUCTIONHealthy estuarine environments are critical for maintaining ecological stability, coastal economies, and human health standards. In order to maintain and even improve these habitats, metrics of current and past conditions must be evaluated to inform proper management. Water quality measurements can be used to indicate overall estuarine health and can aid in understanding increasing coastal threats such as rising sea levels, increased salinities, and urbanization. Long-term water quality analysis is key for developing target thresholds for future management action as well as assessing the efficacy of past management measures (Cloern et al., 2016). The value of historical observations in advancing understanding of estuarine water quality has been demonstrated by multi-decadal studies of several systems, including the San Francisco Bay area (Beck et al., 2018; Cloern et al., 2016), May River, South Carolina (Souedan et al., 2021), Texas’s coastline (Bugica et al., 2020), and the Chesapeake Bay area (Zhang et al., 2018; Harding et al., 2019). Most notably, long-term water quality monitoring in the Chesapeake Bay has led to the identification of climatic and anthropogenic drivers for certain water quality parameters and subsequent evaluation of the effectiveness of past management and restoration efforts (Kemp et al., 2005; Leight et al., 2011; Zhang et al., 2018; Harding et al., 2019).Datasets used for prior longitudinal water quality studies are commonly a product of governmental agencies developing localized programs, like the Chesapeake Bay Program (Chesapeake Bay Monitoring Program, 2022), in response to increasing population and significant degradation of vital estuarine ecosystems. While national and regional efforts have attempted to provide unbiased, sustained monitoring, these programs currently lack the spatial extent needed to capture coastwide water quality trends. The National Estuarine Research Reserve System (NERRS) is one of the few organizations with dedicated coastal water quality monitoring stations, which are included as part of the NERRS System Wide Monitoring Program (SWMP) that maintains 355 coastal water quality monitoring stations across 29 designated coastal reserves along the USA coastline (National Estuarine Research Reserve System, 2022). Compared to the over 13,500 freshwater monitoring stations maintained by the United States Geological Survey (USGS, 2022), the relatively small number of water quality monitoring stations across coastal and estuarine waters (NOAA Tides & Currents, 2022; US EPA, 2022) are likely not representative of the variations in environmental conditions that we observe across the tens of thousands of miles of shoreline along the United States.Because of the limited number of unbiased monitoring programs, the ability to use water quality data from regulatory operations presents a potentially valuable resource for assessing long-term estuarine conditions. Regulatory programs differ from monitoring programs by collecting water quality samples to meet regulatory requirements and inform short-term decision-making. For example, in North Carolina (NC), there are four NERRS SWMP monitoring stations and eight coastal stations with water quality data available through the USGS (South Atlantic Water Science Center, North Carolina Office, 2022) and fifty stations from the NC Ambient Monitoring System (Water Quality Portal, 2021), but the NC Division of Marine Fisheries (NCDMF) shellfish sanitation program maintains 1,924 water quality monitoring stations. In fact, state shellfish sanitation programs across the USA collect an abundance of water quality observations, and often have for decades. Shellfish mariculture is highly dependent on water quality monitoring due to the direct influence that ambient conditions have on the safety of shellfish meat consumption. The U.S. Food and Drug Administration’s National Shellfish Sanitation Program (NSSP) was developed in 1925 to maintain public safety and human health standards in relation to the consumption of shellfish grown in potentially polluted waters (NSSP, 2019). The implementation of the NSSP has resulted in systematic sampling of water quality for day-to-day fisheries regulation, specifically for Fecal Indicator Bacteria (FIB), a group of bacteria that are commonly used as a proxy measure for harmful pathogen loads in the waterway that could potentially be incorporated into shellfish meat through filter feeding. Thus, fecal coliforms (FC), a type of FIB, and other environmental factors that contribute to FC load and water quality, are regularly measured in shellfish growing waters due to the food safety implications. As a product of this regular testing, fisheries operations have accumulated decades of data with the potential to provide insights on historical trends with wide spatial extents, potentially filling gaps in long-term water quality monitoring capacity.However, because of the limited resources and industry specific priorities, regulatory data can maintain underlying biases as a result of the sampling methodology used to collect the water quality sample. Often, the collection of a sample can be motivated by day-to-day operational decisions, such as weather, the availability of field technicians, and ease of collection. These operational decisions lead to non-random sampling that provides observations that are not always representative of the system’s true dynamics. Engaging regulatory personnel to understand their fisheries management and sampling decisions is necessary to properly analyze the observations collected by shellfish sanitation programs.For example, the NSSP permits states to employ one of two sampling strategies when collecting regulatory water quality data in shellfish growing waters: adverse pollution condition sampling and systematic random sampling. The adverse pollution condition sampling strategy describes sampling in periods when known contamination events (commonly due to point-source pollution events or rainfall events) have degraded the water quality, and data collected under these conditions capture peak contamination. States must collect “a minimum of five samples… annually under adverse pollution conditions from each sample station in the growing area” (NSSP, 2019) to meet NSSP sampling requirements. In contrast, the systematic random sampling strategy describes the collection of data across “a statistically representative cross section of all meteorological, hydrographic, and/or other pollution events” (NSSP, 2019), resulting in the data collection under varied environment and climactic conditions. For state programs that use systematic random sampling, the NSSP requires samples be collected at least 6 times throughout the year (NSSP, 2019). As a result of the requirements for the conditions under which the two systems of sampling can take place, the resulting data may be biased and impact their utility for use in long-term water quality assessments. With our growing reliance on aquaculture and the expanding value of shellfish production driving the development of fisheries management infrastructure (Azra et al., 2021), long-term datasets available through shellfish sanitation programs will become increasingly valuable. Realizing the potential of regulatory datasets to inform long-term water quality trends is a vital next step for assessing the health of our coastal ecosystems, but research is needed to determine the utility of these data for water quality analyses.The goal of this study was to utilize shellfish management data to infer long-term spatiotemporal trends in water quality parameters, including FC and salinity, while accounting for variation in routine sampling conditions and environmental landscapes. Study objectives included (1) analyzing spatiotemporal trends from multidecadal fecal coliform concentration observations collected by a shellfish sanitation program, (2) identifying possible management and environmental drivers of fecal coliform trends, and (3) assessing the feasibility of using these monitoring data to infer long-term water quality dynamics. We focused on North Carolina’s shellfish waters as a representative study system due to the availability of public, digitized multidecadal data, and the region’s rapidly growing population, wide variety of land use characteristics along the coast, presence of the second largest estuarine system in the contiguous USA, and growing shellfish industry. Ultimately, this study demonstrates the application of shellfish management data for long-term water quality trend analysis in estuaries, informs future resource management strategies, and reveals new insights into the functioning of coastal systems.

A document by Megan M. Carr. Click on the document to view its contents.