Crop mapping involves identifying and classifying crop types using spatial data, primarily derived from remote sensing imagery. This study presents the first comprehensive review of large-scale, pixel-wise crop mapping workflows, encompassing both conventional supervised methods and emerging transfer learning approaches. To identify the optimal time-series generation approaches and supervised crop mapping models, we conducted systematic experiments, comparing six widely adopted satellite image-based preprocessing methods, alongside eleven supervised pixel-wise classification models. Additionally, we assessed the synergistic impact of varied training sample sizes and variable combinations. Moreover, we identified optimal transfer learning techniques for different magnitudes of domain shift. The evaluation of optimal methods was conducted across five diverse agricultural sites. Landsat 8 served as the primary satellite data source. Labels come from CDL trusted pixels and field surveys. Our findings reveal three key insights. First, fine-scale interval preprocessing paired with Transformer models consistently delivered optimal performance for both supervised and transferable workflows. RF offered rapid training and competitive performance in conventional supervised learning and direct transfer to similar domains. Second, transfer learning techniques enhanced workflow adaptability, with UDA being effective for homogeneous crop classes while fine-tuning remains robust across diverse scenarios. Finally, workflow choice depends heavily on the availability of labeled samples. With a sufficient sample size, supervised training typically delivers more accurate and generalizable results. Below a certain threshold, transfer learning that matches the level of domain shift is a viable alternative to achieve crop mapping. All code is publicly available to encourage reproducibility practice.

The compound effects of fluvial flooding, tidal dynamics, and sea-level rise (SLR) have the potential to mobilize pollutants at contaminated sites, which are often situated in flood-prone areas. We assessed the compound effects of these flood drivers on benzo-[a]-pyrene (B-[a]-P)-contaminated sediments in the Lower Darby Creek Area (LDCA) Superfund Site in Pennsylvania, USA. B-[a]-P, ubiquitous in the sediments of LDCA, is a known human carcinogen and is an indicator of polycyclic aromatic hydrocarbons in the environment. The LDCA is tidally influenced via the Delaware Bay, is projected to experience sea-level rise, and is situated within an active river floodplain. These conditions lead to potential B-[a]P transport within and out of the LDCA. Using a one-way coupling of the Hydrologic Engineering Center-River Analysis System (HEC-RAS) model and the Water Quality Analysis Simulation Program (WASP), we demonstrate that by 2050 fluvial flooding will continue to be the major driver of contaminant transport in the LDCA system. Fluvial-driven sediment transport defines B-[a]P deposition, which is largely influenced by tributary inputs and the distribution of B-[a]P in floodplain sediments. The complex patterns of B-[a]P redistribution at the LDCA, influenced by multiple drivers of flooding, demonstrate the utility of a coupled modeling approach to inform remediation and community resilience.

Starry stonewort (Nitellopsis obtusa (Desvaux) J. Groves) is an invasive freshwater green macroalga in North America that is widespread in the Laurentian Great Lakes Basin and forms thick monotypic meadows in littoral zones. However, little is known about its development throughout the growing season. The objectives of this study were to document the growth and development of N. obtusa in two Lake Michigan drowned river mouth lakes and relate this growth to temperature, nutrients, and other submerged aquatic vegetation. These lakes are significant because they serve as direct connections from the landscape to the Great Lakes system and offer pathways for further range expansion of N. obtusa. Snorkel surveys were conducted in 2020 (biomass) and 2021 (reproductive structure development) at four sites that differed in management and use, three in Pentwater Lake and one in Muskegon Lake. Low biomass of N. obtusa was recorded in mid-July, with exponential growth through mid-August, and stabilization through late summer into October. Biomass was highest at the marina site, and the lowest relative biomass was recorded at the reference site. Water parameters did not explain the growth of N. obtusa well, but were consistent with prior studies. Male reproductive structures (antheridia) developed later than in the other study of N. obtusa in its native range, and females (oogonia) were not observed. Our study suggests N. obtusa growth patterns can differ among lakes and at sites within lakes; marinas should be targeted for management efforts to limit the spread of N. obtusa within and among lakes.

Study region: Serbian Danube River Basin Study focus: As climate change makes weather patterns more erratic, water supply for agriculture is becoming increasingly uncertain. This is concerning in the Serbian Danube River Basin, where crops are mainly rainfed and the growing season is becoming warmer and drier. Assessing the balance between future agricultural water demand and availability in a changing climate is critical to address agricultural water scarcity. To understand how changing climate will affect water availability during 2041-2070, we used the Soil and Water Assessment Tool+ hydrological model with field-scale crop rotations and irrigated extent data and forced with regional climate model data under two representative concentration pathways (RCP4.5 and RCP8.5). New hydrological insights for the region: Declining precipitation, increasing evaporative demand, and lack of widespread irrigation will intensify green water (i.e., soil moisture from rainfall that rainfed systems rely on) scarcity and crop water stress across the spring-planted, rainfed cropping systems in Serbia during the peak growing season. Irrigated fields, currently rare, are barely offsetting green water scarcity and crop water stress and will need to increase irrigation by 10-20 % just to maintain current levels of green water scarcity and crop water stress. These findings highlight that agricultural producers in Serbia will need to adjust agricultural practices and likely expand irrigation to tackle increased water demand, but this may reduce blue water availability.

Cropping patterns give useful information in agricultural production from the perspective of logistics and pricing. In this study, a machine learning classification model based on Random Forest algorithm and Sentinel-2 imagery was used to provide spatial maps of the crops in the Vojvodina region (Serbia), a predominantly agricultural area with 77% cropland and intensive production of arable crops. Ground truth data on geolocation of the fields was collected during the seven-year period (2016-2022) to train the classification algorithm. The most common crops identified are wheat (together with barley), maize, soybean, sugar beet, sunflower and rapeseed. Crop maps resulting from the classification were analyzed. By multiplying the pixel count by the size of each pixel the percentage of area covered by most common crops within the region was calculated annually and averaged across all years. Multi-annual trends were also examined. For the sake of clarity, the most frequent rotations were analyzed at both the county and municipality level. The spatial distribution analysis highlighted variability in total area coverage by certain crops in the region. While some crops maintained a significant area coverage across all counties, others exhibited localized cultivation. Maize occupies the most areas in all counties (29-39%), however there are cases where wheat is equally present as maize (26-33%) and others where wheat is less cultivated in favor of soybean and sunflower. Counties with the most arable area covered by sunflower (18-21%) are in the eastern part of the Vojvodina region, experiencing slightly drier conditions than counties in the western part of the region where soybean takes up 22-30% of arable land. Lastly, sugar beet and rapeseed together cover 2-9% of arable area in all counties and are consequently less present in rotation patterns. Notable variations in crop transition frequencies and spatial distribution patterns were revealed among all counties in the study region. Certain crop transitions demonstrated consistency across all counties, indicating widespread adoption of specific rotation schemes, others showed variability, reflecting local agronomic practices. The transition between maize and wheat emerged as the most frequent across all counties throughout the observed years, however at varying scales. It is observed that the counties in northern and southern parts of the region practice maizewheat transition most frequently (18-32% of all transitions). Sunflower in transition with maize and wheat is more present in the eastern part of the region (> 20%), while soybean and maize transition is the most prevalent in counties in central and western Vojvodina (> 15%). Another notable observation is that some municipalities within one county practice different crop rotation patterns. A typical example is Ju.znoba.cki county, where western municipalities show more arable areas covered by soybean, hence the more frequent transitions with maize. On the other side, it is observed that the eastern parts of the same county incorporate more wheat in rotating crops. The findings of this study contribute to our understanding of crop rotation practices in Vojvodina region which is important in guiding agricultural management decisions and promoting sustainable agriculture.

We use social science survey methods to evaluate the extent to which residents in Great Lakes Areas of Concern (AOC) experience social gaps in accessing benefits of AOC restorations, comparing post-restoration outcomes in two AOCs with differing socioeconomic profiles in Muskegon County, Michigan: Muskegon Lake and White Lake. We find that survey respondents engage in similar types of recreational activities with similar rates of frequency across sites, but Muskegon residents have more pessimistic perceptions of the environmental and community outcomes of the Muskegon Lake restoration. On closer inspection, significant differences in post-restoration assessments exist within the City of Muskegon, with residents from neighborhoods where environmental (in)justice risk factors are high reporting continuing concerns about environmental quality and social problems. We highlight the relevance of these observations to community redevelopment in Great Lakes AOCs that share Muskegon's historic context and present-day demographic profile, pointing to a potential for equity gaps that require conscientious planning.

Continued alteration of the nitrogen cycle exposes receiving waters to elevated nitrogen concentrations and forces drinking water treatment services to plan for such increases in the future. We developed four 2011-2050 land cover change scenarios and modeled the impact of projected land cover change on influent water quality to support long-term planning for the Minneapolis Water Treatment Distribution Service (MWTDS) using Soil Water and Assessment Tool. Projected land cover changes based on relatively unconstrained economic growth led to substantial increases in total nitrogen (TN) loads and modest increases in total phosphorus (TP) loads in spring. Changes in sediment, TN, and TP under two "constrained" growth scenarios were near zero or declined modestly. Longitudinal analysis suggested that the extant vegetation along the Mississippi River corridor upstream of the MWTDS may be a sediment (and phosphorus) trap. Autoregressive analysis of current (2008-2017) chemical treatment application rates (mass per water volume processed) and extant (2001-2011) land cover change revealed that statistically significant increases in chemical treatment rates were temporally congruent with urbanization and conversion of pasture to cropland. Using the current trend in chemical treatment application rates and their inferred relationship to extant land cover change as a bellwether, the unconstrained growth scenarios suggest that future land cover may present challenges to the production of potable water for MWTDS.

Lake Michigan's drowned river mouths (DRM) are hydrologically unique systems with both riverine and large-lake influences. Serving as focal points for human development, DRMs have experienced a history of industrialization, urbanization, and are now moving towards an era of restoration and revitalization. The goal of this study was to examine water quality in 12 DRMs along Lake Michigan's eastern shoreline. We hypothesized that there is a latitudinal gradient in indicators of water quality in these DRM lakes, which is the result of natural land cover, anthropogenic land use, and underlying geology. We identified a latitudinal gradient in land use/land cover with developed land area more abundant in southern DRMs and forest more abundant in northern land area; this distinction was evident at both the local and whole watershed geographic scales but was more distinct at the local scale. Water quality followed suit with specific conductance, chlorophyll-a, and total phosphorus (TP) concentrations higher in southern DRMs and lower in northern DRMs; whereas, Secchi disk depth showed the opposite trend. Multivariate analysis results were consistent with the relationship of water quality and latitudinal gradient. Thermal stratification and low dissolved oxygen (DO) concentrations in the hypolimnion were more common in deeper (i.e., greater than 4 m) areas of DRMs; low DO in the hypolimnion was more likely in southern DRMs than in northern DRMs. These results provide a foundation for future research initiatives in helping separate anthropogenic vs. natural stress in these systems, which provide critical ecosystem services for their surrounding communities.

Fecal pollution is one of the most prevalent forms of pollution affecting waterbodies worldwide, threatening public health and negatively impacting aquatic environments. Microbial source tracking (MST) applies poly-merase chain reaction (PCR) technology to help identify the source of fecal pollution. In this study, we combine spatial data for two watersheds with general and host-associated MST markers to target human (HF183/ BacR287), bovine (CowM2), and general ruminant (Rum2Bac) sources. Concentrations of MST markers in samples were determined with droplet digital PCR (ddPCR). The three MST markers were detected at all sites (n = 25), but bovine and general ruminant markers were significantly associated with watershed characteristics. MST results, combined with watershed characteristics, suggest that streams draining areas with low-infiltration soil groups and high agricultural land use are at an increased risk for fecal contamination. Microbial source tracking has been applied in numerous studies to aid in identifying the sources of fecal contamination, but these studies usually lack information on the involvement of watershed characteristics. Our study combined watershed characteristics with MST results to provide more comprehensive insight into the factors that influence fecal contamination in order to implement the most effective best management practices.

Nitellopsis obtusa was first documented in the St. Lawrence River in 1974 and likely spread via human-assisted activity to at-least seven states in the U.S.A. This invasive macroalga is a nuisance for native plants, animals, and recreational activities. Because eradication of invasive species is more difficult after establishment, early detection plans are an important tool in preventing and slowing their spread. Macro-scale data analyses have improved our ability to predict changes in freshwater ecosystems and are important to assess and control invasive species. We developed species distribution models (random forest, boosted regression trees, and Maxent) using presence records of N. obtusa coupled with publicly available in-lake temperature and chemistry (bicarbonate and chloride) model data and landscape scale lake watershed characteristics for over 48,000 individual lakes in the Midwest and northeast U.S.A. January and July–August–September growing degrees days, bicarbonate concentrations, and chloride concentrations correlate with high relative likelihood of occurrence. Our analyses found N. obtusa likelihood of occurrence is high in developed, lake-dense regions with ~ 2000 July–August–September growing degree days, 1–3 mMol bicarbonate, and > 10 mg L −1 chloride. Based on relative likelihood of occurrence predictions, N. obtusa has the potential to spread to new lakes within Midwest and northeast USA states that currently do not have known populations of N. obtusa , including inland lakes in Illinois, Iowa, Ohio, and Pennsylvania.

Fecal pollution is one of the most prevalent forms of pollution affecting waterbodies worldwide, threatening public health, and negatively impacting aquatic environments. Microbial source tracking (MST) applies polymerase chain reaction (PCR) technology to help identify the source of fecal pollution. In this study, we combine spatial data for two watersheds with general and host-specific MST markers to target human, bovine, and general ruminant sources. Two different PCR technologies were applied for quantifying the targets: quantitative PCR (qPCR) and droplet digital PCR (ddPCR). We found that ddPCR had a higher detection rate (75%) of quantifiable samples compared to qPCR (27%), indicating that ddPCR is more sensitive than qPCR. The three host-specific markers were detected at all sites (n=25), suggesting that humans, cows, and ruminants are contributing to fecal contamination in these watersheds. MST results, combined with watershed characteristics, suggest that streams draining areas with low-infiltration soil groups, high septic system prevalence, and high agricultural land use are at an increased risk for fecal contamination. Microbial source tracking has been applied in numerous studies to aid in identifying the sources of fecal contamination, however these studies usually lack information on the involvement of watershed characteristics. Our study combined watershed characteristics with MST results, applying more sensitive PCR techniques, in addition to watershed characteristics to provide more comprehensive insight into the factors that influence fecal contamination in order to implement the most effective best management practices. Highlights ddPCR provided higher sensitivity over qPCR when analyzing environmental samples Human markers had an association with the number of septic systems in a watershed Every site had positive detections for all FIB markers Both ruminant markers were associated with low infiltration hydrologic soil groups Combining watershed characteristics with MST testing improved source identification

Climate change has significant implications for irrigated agriculture and global food security. Understanding how altered precipitation patterns and magnitudes, coupled with rising growing season temperatures, affect irrigation demand and crop production is a prerequisite for formulating effective water resources management strategies. This study evaluated the effects of near-term climate change (centered on 2035) on irrigation demand, green water scarcity, and row crop yields in a major agricultural watershed in southern New Jersey, USA. Downscaled precipitation and temperature from six General Circulation Models (GCMs) for two representative concentration pathways (RCP-4.5 and 8.5) from the Coupled Model Intercomparison Project Phase 5 (CMIP5) were used to drive the Soil and Water Assessment Tool hydrological model. Temperature and precipitation increases resulted in greater surface runoff, lateral flow, groundwater recharge, and total streamflow. Seasonal ET for corn is projected to alter between -3.0 % to 0.5 %, with irrigation demand between -17 % to -1 %, and yield ranges between -4 % to +9 % depending on the GCMs in the RCP-4.5 scenario, with similar patterns projected by RCP-8.5 scenario. For soybean, the simulation also indicates a declining trend of ET and irrigation demand while increasing yield. Increasing yield for both crops is attributed to changes in agronomic management practices combined with genetically improved cultivars and higher soil fertility due to CO2 fertilization. Green water scarcity analysis under future climate change for corn and soybean display a decreased soil moisture stress due to increased water use efficiency resulting from reduced stomatal conductance under elevated CO2 concentration.

Urban development is a well-known stressor for stream ecosystems, presenting a challenge to managers tasked with mitigating its effects. For the past 20 y, streamflow, water quality, geomorphology, and benthic communities were monitored in 5 watersheds in Montgomery County, Maryland, USA. This study presents a synthesis of multiple studies of monitoring efforts in the study area and new analysis of more recent monitoring data to document the primary lessons learned from monitoring. The monitored watersheds include a forested control, an urban control with centralized stormwater management, and 3 suburban treatment watersheds featuring low-impact development and a high density of infiltration-focused stormwater facilities distributed across the watershed. Treatment watersheds were monitored before development, during construction, and after development. Monitoring was initiated to inform adaptive management of stormwater and impervious cover limits within the study area, with a focus on the impacts of distributed stormwater management. Results from our synthesis indicate that distributed stormwater management is advantageous compared with centralized stormwater management in numerous ways. Hydrologic benefits were greater with distributed stormwater infrastructure, demonstrating the ability to mitigate runoff volumes and peak flows and, for small storms, replicate predevelopment conditions. Baseflow temporarily increased during the construction phase in the treatment watersheds. Water-quality benefits were mixed, with declines in baseflow nitrate concentrations but limited changes to nitrate export and increases in specific conductance after development. Substantial topographic changes occurred during construction in the treatment watersheds, including changes within the riparian zone, despite riparian buffer protections. Ecological monitoring indicated that even though index of biotic integrity scores rebounded in some cases, sensitive benthic macroinvertebrate families did not fully recover in the treatment watersheds. Lessons learned from this synthesis highlight the importance of tracking multiple indicators of stream health and considering past land use and that more stormwater facilities distributed across the watershed is beneficial but cannot mitigate the effects of all urban stressors on aquatic ecosystems.

Increased intensity and frequency of floods raise concerns about the release and transport of contaminated soil and sediment to and from rivers and streams. To model these processes during flooding events, we developed an External Coupler in Python to link the Hydrologic Engineering Center-River Analysis System (HEC-RAS) 2D hydrodynamic model to the Water Quality Analysis Simulation Program (WASP). Accurate data transfer from a hydrodynamic model to a water quality model is critical. Our test results showed the External Coupler successfully linked HEC-RAS and WASP and addressed technical challenges in aggregating flow data and conserving mass during the flood event. We ran the coupled models for a 100-year flood event to calculate flood-induced transport of sediment-associated arsenic in Woodbridge Creek, NJ. Change in surface sediment and arsenic at the end of 48-h flood simulation ranged from a net loss of 13.5 cm to a net gain of 11.6 cm, and 16.2 to 2.9 mg/kg, respectively, per model segment, which demonstrates the capability of the coupled model for simulating sediment and contaminant transport in flood.

Soils provide vital ecosystem services, from sequestering carbon to providing food and moderating floods. Soil erosion threatens the provisioning of these services and degrades downstream water quality. Vegetation plays an important role in soil retention: by holding it in place, soil can continue to provide ecosystem goods and services and protect water resources. The aims of this study were to: (1) develop a 30-meter resolution map of erosion in the conterminous United States, and (2) quantify the soil retention service of natural vegetation. Using the Revised Universal Soil Loss Equation and physiographic and remote sensing datasets, we estimated sheet and rill erosion. We also developed a map of sediment delivery ratio to connect erosion to downstream delivery using hydrologic connectivity. The estimated sheet and rill erosion in the conterminous United States was 1.55 Pg yr−1, of which 0.52 Pg yr−1 reached waterbodies. Natural land cover prevents 12.3 Pg yr−1 of sheet and rill erosion and 5.1 Pg yr−1 in delivery to waterbodies. The value of natural land cover in retaining sediment is a function of the land cover, physiographic characteristics, and spatial context. This study has implications for spatial prioritization of natural land cover preservation and agricultural land management to minimize sediment erosion and delivery.

Green stormwater infrastructure implementation in urban watersheds has outpaced our understanding of practice effectiveness on streamflow response to precipitation events. Long-term monitoring of experimental suburban watersheds in Clarksburg, Maryland, USA, provided an opportunity to examine changes in event-based streamflow metrics in two treatment watersheds that transitioned from agriculture to suburban development with a high density of infiltration-focused stormwater control measures (SCMs). Urban Treatment 1 has predominantly single family detached housing with 33% impervious cover and 126 SCMs. Urban Treatment 2 has a mix of single family detached and attached housing with 44% impervious cover and 219 SCMs. Differences in streamflow-event magnitude and timing were assessed using a before-after-control-reference-impact design to compare urban treatment watersheds with a forested control and an urban control with detention-focused SCMs. Streamflow and precipitation events were identified from 14 years of sub-daily monitoring data with an automated approach to characterize peak streamflow, runoff yield, runoff ratio, streamflow duration, time to peak, rise rate, and precipitation depth for each event. Results indicated that streamflow magnitude and timing were altered by urbanization in the urban treatment watersheds, even with SCMs treating 100% of the impervious area. The largest hydrologic changes were observed in streamflow magnitude metrics, with greater hydrologic change in Urban Treatment 2 compared with Urban Treatment 1. Although streamflow changes were observed in both urban treatment watersheds, SCMs were able to mitigate peak flows and runoff volumes compared with the urban control. The urban control had similar impervious cover to Urban Treatment 2, but Urban Treatment 2 had more than twice the precipitation depth needed to initiate a flow response and lower median peak flow and runoff yield for events less than 20 mm. Differences in impervious cover between the Urban Treatment watersheds appeared to be a large driver of differences in streamflow response, rather than SCM density. Overall, use of infiltration-focused SCMs implemented at a watershed-scale did provide enhanced attenuation of peak flow and runoff volumes compared to centralized-detention SCMs.

In this study, the Soil and Water Assessment Tool (SWAT) is coupled with the Bacteria Source Load Calculator (BSLC) to develop alternative scenarios of agricultural best management practice (BMP) implementation and septic system repair to achieve total maximum daily load (TMDL) targets for Escherichia coli (E. coli) in a small agricultural watershed located in Michigan, USA. Due to the uncertainty involved in bacterial source estimation from wildlife and failing septic systems in agricultural watersheds, we propose a method for estimating their respective bacteria loading during SWAT calibration, which was performed at eight sampling locations and validated at a ninth. BMP implementation scenarios were prioritized based on their ability to meet TMDL targets at the lowest cost using a spatially targeted subwatershed ranking identifying the greatest E. coli source contribution. BMP effectiveness is driven by E. coli source locations defined in the BSLC-SWAT linkage. The framework developed here, from linking BSLC to SWAT to ranking subwatersheds, was effective in meeting the TMDL targets and is readily transferable to other watersheds. Agricultural BMPs were unsuccessful in meeting TMDL targets, but the target can be met by repairing about 60 of 110 estimated failing septic tanks.