LIVE - Poster Technical Session & Student Competition #1
Monday, November 8 | 5:30 PM - 6:30 PM EST

IMPORTANT: In order to receive important communication from the AWRA Team about your conference participation, you must add [email protected] and [email protected] to your safe sender list.

Conference Sponsors

Platinum

 

Gold

  

Silver

ATTN: We need you – our conference attendees – to evaluate student participants in the poster competition where one student will receive the People's Choice Award! How do you know who is competing and how do you score them? A blue graduation cap next to a presenter's name indicates they are a student eligible to compete for the People’s Choice Award. As you watch their poster presentations, please make sure to complete the People's Choice Award Evaluation Form.

MOBILE TIP!  Scan the QR Code with your mobile device camera to complete the evaluation form. 

Alternatively if you were contacted by the Conference Planning Committee Student Activities Chair, Allison Lewis, and asked to serve as an official judge for the poster competition, please check your email for the link to the Poster Evaluation Form. Students participating in the poster competition are identified by a blue graduation cap.

Poster Category: Behavior and Water Management & Water and Energy
Zoom Room #1

Moderator: Erich Hersh

Proximity to a Surface Water Body in a Coastal, Golfing Community of South Florida Effects Homeowners’ Willingness to Protect and Fund Remediation

Presenter: Haylie Fine 

Proximity to surface waters influence people’s motives to fund projects that protect and remediate environmental habitats. A private gated community in southeast, coastal Florida, Jonathan’s Landing (JL), consists of an integrated golf course (15,000 sqft) and approximately 1,234 homes on or near the water (JL POA, 2021). The water traps on the golf course have culverts that spill directly into waterways, residential lakefront lawn care is year round, and all manicured facilities lack a riparian buffer. This created odd smells, colors, and algal blooms that alarmed residents. Disconsolate, home owners enlisted aid from students in the Jupiter Environmental Research and Field Studies Academy (JERFSA) at Jupiter Community High School to assist in the creation of a lake association that would educate and engage the community at large.

To provide the community with information on resident attitudes, a survey was modified from Green Mountain Conservation District (2021) and Ossipee Lake Alliance (2020). It was developed in Survey Monkey and distributed anonymously via a link by the Property Owners Association (POA). They placed the link on the POA website and emailed all residents as well. The intent was to determine if homeowners on the lake would be more inclined to act in the protection of lake health. The content contains inquiries on land use, demographics, and visibility of the lake. For example, “How close is your residence to the lake?”. The questions contained references to Tobler’s First Law of Geography (Dempsey, 2014), along with the Theory of Planned Behaviour (LaMorte, 2019), to provide background information for geographic relationships and human beliefs. Understanding what prompts a person to become integrated with environmental restoration and maintenance of a lake will help to understand and ultimately increase community involvement. A specific motivation for willingness to pay v accept is the proximity a potential participant is to the physical feature (Loomis and Pate, 1997). As expected, outcomes indicated that people who have a daily visual of the lake were more willing to take action to protect the lake, and those inland were more willing to accept compromised waters.

Municipal and industrial Water Use in the Oil impacted Bakken Region 2000-2018: Implications for Planning and Management

Presenter: Christina Hargiss
The management of clean energy and water is a pressing societal challenge modeled by the Water-Energy Nexus. As cognizance grows around increased energy demand and dwindling fossil fuels, research into sustainable energy sources has increased. Organic waste-producing companies like agricultural, food/beverage, and wastewater treatment plants rely heavily on water producing high levels of waste, making them ideal candidates for bio-oil creation partnerships. Hydrothermal liquefaction (HTL), a process in which wet, organic biomass is converted into a liquid biocrude at high pressure and temperature, proves viability for converting organic waste into an alternative energy source. Due to varying physical and chemical characteristics of feedstocks such as moisture content and organic content, the objective of this research is to utilize HTL to produce biocrude from co-digestion of organic waste feedstocks including primary sludge (PS), return activated sludge (RAS), winery rose lees (WRL) and brewery trub (BT) in order to optimize these characteristics, and to also conduct a comparative analysis in terms of biocrude yield and toxicity impact of the post-HTL wastewater by-product. The experimental methodology includes raw feedstock characterization; HTL conversion experiment; and characterization of post-HTL products (i.e., biocrude, biochar, and ACP). As-received waste feedstocks showed higher concentrations of moisture content in RAS and PS while BT and WRL showed higher organic content. Post-HTL preliminary results indicate the production of biocrude, along with substantial quantities of ACP. ACP shows high concentrations of nutrients when compared to benchmark wastewaters. It is anticipated that further analysis will elucidate the water impacts of HTL processing.

Evaluating the Hydrothermal Co-liquefaction of Organic Waste Feedstocks into Biofuel

Presenter: Oluwayinka Adedeji
The management of clean energy and water is a pressing societal challenge modeled by the Water-Energy Nexus. As cognizance grows around increased energy demand and dwindling fossil fuels, research into sustainable energy sources has increased. Organic waste-producing companies like agricultural, food/beverage, and wastewater treatment plants rely heavily on water producing high levels of waste, making them ideal candidates for bio-oil creation partnerships. Hydrothermal liquefaction (HTL), a process in which wet, organic biomass is converted into a liquid biocrude at high pressure and temperature, proves viability for converting organic waste into an alternative energy source. Due to varying physical and chemical characteristics of feedstocks such as moisture content and organic content, the objective of this research is to utilize HTL to produce biocrude from co-digestion of organic waste feedstocks including primary sludge (PS), return activated sludge (RAS), winery rose lees (WRL) and brewery trub (BT) in order to optimize these characteristics, and to also conduct a comparative analysis in terms of biocrude yield and toxicity impact of the post-HTL wastewater by-product. The experimental methodology includes raw feedstock characterization; HTL conversion experiment; and characterization of post-HTL products (i.e., biocrude, biochar, and ACP). As-received waste feedstocks showed higher concentrations of moisture content in RAS and PS while BT and WRL showed higher organic content. Post-HTL preliminary results indicate the production of biocrude, along with substantial quantities of ACP. ACP shows high concentrations of nutrients when compared to benchmark wastewaters. It is anticipated that further analysis will elucidate the water impacts of HTL processing.

Application on the Benefit of Micro-hydropower in Agricultural Channels to Consider Water-energy-food Nexus

Presenter: Jen-Chieh Shih

Micro-hydropower is necessary renewable energy to provide baseload, and it has the advantages of sustainable development and reduction of greenhouse gas emissions. There is an interrelationship between irrigation water and energy from micro-hydropower with micro-hydropower development in agricultural irrigation systems. Li et al. (2019) mentioned that the agricultural system estimated the water supply-demand, energy supply-demand, land demand, and food production and was quantitatively analyzed under different scenarios. However, the study of water for electricity generation was neglected in the agricultural system. Zhou et al. (2019) apply small-hydropower into water supply systems to lift renewable power output and uplifting the synergistic benefits of the Water-Food-Energy (WFE) Nexus steered by the optimal water allocation and small-hydropower installation. Still, the adjustment of the water source by the reservoir makes the flow of the water supply system unstable, which leads to inconsistent electricity generation of small hydropower. Gaudard et al. (2018) research that hydropower plants' relationship between water and energy is set upstream. The results show that seasonality slightly affects hydroelectric power generation.

Therefore, the study set up a micro-hydropower generation system in the Linnei channel of the Zhuoshuixi river watershed in the middle of Taiwan. Collected channel background information and used Doppel (Teledyne StreamPro ADCP) to measure the water level, velocity, and discharge of the study site, analyze potential power generation, and evaluate the profit and payback period micro-hydropower generation and the impact of micro- hydropower systems in agriculture. Furthermore, to investigate the relationship nexus of WFE and assess micro-hydropower's effectiveness in reducing carbon dioxide emissions. According to the results of this study, it can be used as a reference basis for setting up other micro-hydropower systems in the future.

Poster Category: Biological Soil Crust & Coastal Sediments
Zoom Room #2

Moderator: Zhang Zhenxing

Chemical and mineralogical characterization of Mobile Bay and eastern Mississippi Sound sediments

Presenter: Steven Young
The Mobile Bay estuary and adjoining Mississippi Sound are important Alabama Gulf Coast marine waters for recreational and commercial fishing, as well as trade and commerce. The Mobile Bay watershed encompasses more than 43,000 mi2 of land in Alabama, Georgia, Mississippi, and Tennessee, and brings a large influx of sediments into coastal Alabama. This study seeks to determine the chemistry and mineralogy of 120 bottom sediment samples collected on a 4km2 grid within Mobile Bay and the Mississippi Sound.

USEPA Method 3051A is used to extract the mobile fraction of the sediment sample chemistry. Digestate solutions are analyzed for major, minor and trace element chemistry by ICP-OES. Sediment mineralogy is determined using X-ray powder diffraction and quantified by Rietveld refinement.

Thus far, chemical and mineralogical data have been collected for 52 sediment samples. Most trace elements (B, Be, Co, Cr, Ni, Pb, V, Zn) are very strongly correlated (r > 0.899) with Al and Fe, suggesting that their distribution is controlled by adsorption on clay minerals and iron oxides. Magnesium is also very strongly correlated with Al, B, Cr, Fe, Ni, K and V. Other very strong correlations exist between Ba and Co, and between Ca and Tl concentrations. Potential sources for trace elements in Mobile Bay sediments include coal mining, agricultural chemicals, industrial processes, and rapid urbanization since the 1970s. Spatial analysis of the chemical data shows higher concentrations of Al, Ba, Be, Cr, Co, Cu, Fe, Ni, Pb and V in sediments near the center of the Bay, away from the urbanized and industrialized northwestern coast, with highest concentrations between the mouth of the Fowl River and Gaillard Island near the western shoreline and near the central eastern shoreline. Local areas of elevated Al, As, Ba, Be, Cr, Fe, Ni, and Pb concentrations also exist in sediments west of Mobile Bay in the eastern Mississippi Sound. Concentrations of Ca and Sr are highest in sediments near the opening between Mobile Bay and the Mississippi Sound. Analysis of spatial and statistical correlations between sediment mineral content and trace element chemistry is underway and will be reported at the Conference.

Ecohydrological Effects of Biological Crusts: An Overview

Presenter: Pushpita Kumkum
Biological soil crusts are the living surface communities that result from the interaction between soil particles and microorganisms such as cyanobacteria, mosses, lichens, algae, microfungi, and bryophytes. They play an important role in hydrologic processes like infiltration, surface runoff, and evaporation. Many studies have been conducted focusing on different parameters or characteristics of biocrusts that have impacts on the hydro-ecosystems. Integration of these studies is essential to fully understand the mechanism and to evaluate the potential of the biological soil crusts in preventing land degradation or contribution in restoring water resources. In this review paper, various factors that take part in effecting the hydrologic behavior of biological crusts such as types or species, soil texture, particle structure, and size distribution, hydrophobicity, absorptivity, porosity, surface roughness, thickness of the crusts, successional stages of development, interaction with the physical crusts present will be discussed. Also, a systematic analysis will be conducted regarding the influences of these factors on hydrologic processes, namely infiltration, evaporation, hydraulic conductivity, carbon/nitrogen ratio, soil-moisture availability, total carbon, and total nitrogen content, soluble ammonium, nitrate, dissolved organic salt, and microbial carbon and nitrogen concentrations. Comparative analysis based on the location and climate of the region of interest will also be investigated. Many researchers have conducted studies using remote sensing methods to monitor the development of biocrusts and to characterize them, this is a useful tool due to the provision of examining the stages without destructions. A comprehensive investigation of these methods will also be presented. The goal of this overview study is to evaluate the potential of the biological soil crusts in alleviating the negative impact of climate change like soil erosion or drought. This study can be helpful in providing a guideline on selecting the most feasible approach to tackle the damaging impacts of climate change on socioeconomics and the natural environment.

Controls on sediment mercury and volatile solids concentrations in Mobile Bay, Alabama

Presenter: Alyssa Kandow
The Mobile Bay watershed is the fourth largest drainage basin in North America, covering 65% of the land area of Alabama plus portions of Georgia, Tennessee, and Mississippi. The estuary is used for recreational and commercial fishing, trade, and commerce. The western and northwestern bay shorelines are highly urbanized and industrialized. These industrialized areas have potential sources of mercury contamination, and it is important to identify affected areas since the bioaccumulation of mercury poses significant health concerns to humans and wildlife. This study investigates the relationships between sediment grain size, mercury content, and volatile solid content and their spatial distribution in Mobile Bay, AL.

Sand, silt, and clay percent fractions were determined for 78 sediment samples using a Bettersizer S3 Plus laser diffraction particle size analyzer (PSA) and loss on ignition (LOI) values were measured by combustion at 550 ᵒC after drying at 105 ᵒC. Mercury concentrations have been measured for 14 of the sediment samples using a Brooks Rand MERX-T DP cold vapor atomic fluorescence mercury analyzer.

Although sediment mercury concentrations and volatile solids contents were expected to be higher in finer grained sediments, statistical correlation analysis of the data collected so far shows that sediment volatile solids contents show only weak positive correlation (r = 0.157) and mercury concentrations show a weak negative correlation (r = -0.231) to sediment silt and clay fraction percentages. Spatially, the highest sediment mercury concentrations occur on the western side of the Bay. This is consistent with the most industrialized and urbanized portions of the Bay being on the northwestern and western shores. In contrast, the highest sediment volatile solid contents occur on the northern and eastern sides of the Bay, which is less urbanized and has greater wetland and herbaceous upland cover than the western and northwestern Bay. Future research will continue to investigate the mercury and volatile solid contents of additional sediments collected throughout Mobile Bay to determine factors that control their distributions.

Poster Category: Nutrient Management, Water Quality, and Wetlands
Zoom Room #3

Moderator: Rachel Rotz

Removal of Microplastics from a Wastewater Treatment System in Midwest USA

Presenter: Bangshuai Han
Microplastics pollution is a pervasive issue within marine and freshwater environments and has recently emerged as a major contamination concern. Bioaccumulation and biomagnification of microplastics within the food web could result in adverse physical and chemical impacts on the ecosystem and human body. Domestic wastewater treatment plants (WWTPs) are major hubs that receive and remove pollutants. Despite being largely removed following treatment processes, wastewater effluent is still a major pathway for microplastics to enter the environment due to the high-water discharge rate from WWTPs. However, the removal efficiency and processes of WWTPs regarding microplastics contamination are not well understood. This study determines how much microplastics are removed per each stage of the wastewater treatment process, and reveals the temporal variations of the abundance of microplastic entering the Muncie Wastewater Treatment Plant (WWTP) in Indiana, USA. Grab samples are collected every other week with duplicates from summer 2021. The samples are then filtered using an assembled stack of sieves with mesh sizes between 400 and 74 µm, as well as using wet peroxide oxidation and density separation techniques to isolate microplastics. The last step involves visual identification under microscopes to categorize microplastics based on morphology. Findings from this research will help the WWTP evaluate future renovations related to the removal of microplastics as well as contribute to cutting-edge knowledge in the field.

Plastic Pollution in Stormwater Ponds

Presenter: Jenna Brooks
Stormwater runoff is widely recognized as an important pathway for microplastics from the terrestrial to the marine environment, though the significance of this pathway and the impacts of stormwater infrastructure on plastic transport are not fully understood. In this study, samples were taken in dry weather from both the water and bottom sediments of six stormwater ponds in Tampa, Florida using a 300 µm neuston net and sediment corer. They were processed using a combination of sodium chloride and sodium iodide density separations, visual sorting, and Raman spectroscopy. Particle properties such as surface area, shape, and composition were determined, as well as pond conditions such as hydraulic loading and retention time. These characteristics were analyzed in relation to plastic concentrations and accumulation in sediments. Concentrations in sediments were consistently higher than in water samples and featured a more diverse profile of plastics in terms of various shape factors and surface area. Hydraulic loading was found to be positively correlated with plastic concentrations in sediments. Stormwater infrastructure, including stormwater ponds, may serve as both sinks and sources of microplastics, respectively mitigating and contributing to their spread to downstream waterbodies. Understanding the factors that influence the tendency of plastics to settle in stormwater ponds is therefore highly relevant to both future modeling efforts and the impacts that particles may have on their various environments.

Simulating the cumulative effect of nutrient management in Lake Okeechobee Watershed

Presenter: Osama Tarabih
Lake Okeechobee is a crucial water source in South Florida, supplying water for human consumption, irrigation, and surrounding wetland ecosystems. However, Lake Okeechobee is identified as nutrient impaired by the Florida Department of Environmental Protection. Lake Okeechobee’s nutrient sources include runoff from the surrounding basins, atmospheric deposition, and internal loading from legacy nutrients stored primarily in the sediments. Northern basins contribute almost 90% of L. Okeechobee’s P and N loads with Kissimmee Basin (Upper and Lower) alone contributes about 35% of TP and TN mass load in the lake. Efforts are underway to reduce nutrient loading to L. Okeechobee through the implementation of Best Management Practices (BMPs) tailored to specific nutrient source locations (e.g., agricultural lands, dairy farms, etc.). This study evaluates the cumulative effect of the distributed placement of BMPs in the Upper and Lower Kissimmee watersheds. The Watershed Assessment Model (WAM) is used to simulate hydrologic and nutrient dynamics of Upper and Lower Kissimmee watersheds. WAM is a spatially distributed model designed specifically to simulate Florida’s unique hydrologic conditions (e.g., an abundance of wetlands, high groundwater table, complex water control structure operations, etc.). WAM simulates surface runoff based on land use and soil type combinations to generate water flow rates and nutrient concentrations and loads at each subbasin outlet within Upper and Lower Kissimmee watersheds. Baseline (i.e., current conditions) WAM model of the Upper and Lower Kissimmee watersheds are calibrated/validated for hydrologic and hydraulic conditions and for nutrient conditions for the periods 1995-2006/2007-2018. Optimized BMP placement scenarios that consider the placement, cost, and effectiveness in the Upper and Lower Kissimmee watersheds are simulated through WAM to identify the best scenario that minimizes nutrient loadings out of the watersheds. Results show that optimal placement of BMPs in the two watersheds reduced P and N loadings into L. Okeechobee significantly, information which could be used to guide the decision-making process in the region.

The Implications Of Park Sanctioned Pesticide Use on Non-Targeted Vegetation in Jonathan Dickinson State Park, Coastal Southeast Florida

Presenter: Mercedes Cassidy
The contamination of surface and groundwaters, as well as non-target vegetation, is occurring due to park sanctioned aerial spraying of pesticides in Jonathan Dickinson State Park (JDSP), South Florida. The Carolina Willow has invaded JDSP and eradicated native wetlands while also creating unnatural swamps. In an effort to eliminate the invasive Carolina Willow, the pesticides glyphosate and imazapyr are applied to the park through the use of helicopters, with indirect aim, and are subjected to surficial and coastal winds. The remnants of these pesticides travel through the surface waters, leach into the groundwater, and adsorb to the soil with little degradation. The pesticides used, glyphosate and imazapyr, reduce the plant’s ability to uptake nutrients through their roots and kill bacteria that are essential to plant growth. Unintended native vegetation is affected throughout the watershed due to the fate and transport of these chemicals. Within a 2-mile area, nine locations were sampled with a surface water grab sample and as a historical baseline. Each of these locations was documented with photos and the longitude and latitude were also recorded. Before the pesticides were applied, testing sites, as well as the sensitive native flora affected by the said pesticides, were photographed, identified, and recorded. Vegetation that was bankside appeared healthy and unaffected, while the native aquatic plants that were downstream from the application area appeared dead. The data suggests that every location, except for location 8, was affected; however, not every single plant in each area was affected. The plants that were most susceptible to the pesticides were saw palmetto, carolina willow, hog plum, wax myrtle, and sawgrass. Results indicate that the non-targeted vegetation was affected by the pesticides glyphosate and imazapyr, but after a year, the non-targeted bankside vegetation recovered.

Poster Category: Innovation in Data Collection, Monitoring, & Modeling
Zoom Room #4

Moderator: Valerie Seidel

Utilizing innovative well pipe to collect partitioning water level and verify 3D hydrogeological parameters

Presenter: Bu-Sheng Lee
In recent years, it is important to estimate the distribution of hydraulic properties of transmissibility (T) and storage coefficient (S) in the confined aquifer by three-dimensional (3D). However, in a single well, it is necessary to repeatedly install multiple sets of well packers to measure the stratified layers' groundwater level and to perform heterogeneous hydraulic properties analysis, multiple well systems need to be established. Thence, data obtained in previous years consumes human resources, work time, and budget. Therefore, the purpose of this study is to design a multi-stage concentric well pipe. This pipe can measure the water level, vertical flow, and vertical hydrogeological parameters in different depths of stratified layers. This innovative pipe not only measures the change of water level in different depth layers but also measures water flow with horizontal and vertical directly at the same time. That is significantly reduced the number of observation wells. To verify the feasibility of this new type of pipe, a 3D aquifer sandbox experiment will be constructed for the pumping tests. Then, estimated the 3D hydraulic properties distribution of the aquifer by hydraulic tomography (HT). The realistic hydraulic properties of the entire groundwater aquifer can be estimated more comprehensively with this concept. This will make a great contribution to the future analysis of changes in regional flow fields, groundwater replenishment patterns, and control of the diffusion of underground pollution.

Application of unmanned aerial vehicle technology for an evaluation of surface drainage to an urban lake at the New Jersey shore

Presenter: Mia Najd
The coastal lakes of Monmouth County are a unique feature of the New Jersey shore, with social and financial value to the surrounding communities. Yet, each summer, at the height of their use, these lakes are subject to harmful algal blooms due to nutrients carried to the lakes. Because of this, it is of utmost importance to evaluate surface drainage and manage harmful runoff to the lakes. Such evaluations should not only construct detailed drainage networks, but also characterize the surfaces (e.g. pervious versus impervious) of the drainage network. Light detection and ranging (LiDAR) terrain data can be used to construct urban drainage networks, but LiDAR surveys are costly and time-consuming. A comparatively low-cost and rapid alternative is the use of unmanned aerial vehicle (UAV) color photography. A UAV survey of Lake Como, New Jersey is compared in this study to the use of LiDAR data. The color photography from the UAV is processed using structure-from-motion to derive a digital surface model of the urban watershed draining to Lake Como. An urban drainage network is then constructed using standard geographic information systems hydrology tools of flow direction (i.e. the direction that water would flow given surface slope) and flow accumulation (i.e. the area upslope of a given grid cell location). The impervious and pervious surfaces of the drainage network and upslope drainage areas are characterized based on the UAV color photography and a maximum likelihood classification. The results of this analysis are then compared to similar products derived from LiDAR data. Differences between the UAV- and LiDAR-derived drainage networks are described (e.g. extent and surfaces) to assess the viability of using the lower cost UAV technology. The methodologies exhibited in this study are of practical use for urban stormwater modeling and management, and could also be applied in comparative studies of water quality differences between urban lakes.

Temporal and Spatial Hydrochemical Characteristics of Rio Grande River Water for Irrigation Applicability

Presenter: Miranda De La Garza
Elevated salt levels in water used for irrigation can lead to poor performance and productivity of crops. Detrimental effects of high salt levels in irrigation water include the prevention of plants to absorb nutrients and can affect crop growth. Water quality variability can pose various management challenges in irrigated fields; thus, it is critical to evaluate the water quality across time and location to ensure efficient water management practices. The Lower Rio Grande Valley (LRGV) in South Texas is heavily dependent upon ditch irrigation with districts being reliant on the Rio Grande River (RGR) to support the production of many crops, including vegetables, cotton, corn, sorghum, and citrus. Irrigation districts utilize an irrigation delivery system to transport water through pipelines, canals, and resacas to deliver water to end users. To date, no detailed studies have been conducted to quantify fluxes in salinity levels and their variations in this region. This research provides temporal and spatial evaluations of the RGR’s hydrochemical characteristics at seven sites within selected irrigation districts in the LRGV. Parameters used in the evaluation include pH, electrical conductivity (EC), temperature, total dissolved solids (TDS), and concentrations of major ions such as calcium (Ca2+), magnesium (Mg2+), sodium (Na+), potassium (K+), sulfate (SO42-), chloride (Cl-), nitrate (NO3-), carbonate (CO32-) and bicarbonate (HCO3-), and chemical indices such as salinity hazard, sodium adsorption ratio, sodium percent, residual sodium carbonate, magnesium hazard, Kelly’s ratio, and permeability index. The temporal and spatial variations of these parameters/indices are evaluated statistically using software such as Minitab and StatistiXL. Our preliminary results show that ions such as Ca2+, Mg2+, and HCO3- have little spatial-temporal variation. Whereas other ions like Cl-, and SO42- increase spatially. Furthermore, the EC measurements indicate that the river water is considered permissible for irrigation purposes. The RGR water is also assessed for agricultural purposes using water classification diagrams such as Piper’s, Richard’s, Wilcox, and Doneen diagrams. Results from this research provide information on variations of the RGR water quality along the river and over different seasons and help growers with water management activities.

A Regional Semi-Distributed Streamflow Model Using Deep Learning

Presenter: Zhongrun Xiang
Recent studies have shown that deep learning models in hydrological applications significantly improved streamflow predictions at multiple stations compared to traditional machine learning approaches. However, most studies lack generalization; i.e. researchers are training separate models for each watershed. The spatial and temporal generalization ability of deep learning models in hydrology that can be gained by training a single model for multiple watersheds is evaluated in this study. We developed a generalized model with a multi-site structure for hourly streamflow hindcasts on 125 USGS gauges. Considering watershed-scale features including drainage area, time of concentration, slope, and soil types, the proposed models have acceptable performance and slightly higher median NSE value than training individual models for each USGS station. Furthermore, we showed that the trained generalized model can be applied to any new gauge in the state of Iowa that was not used in the training set with acceptable accuracy. This study demonstrates the potential of deep learning studies in hydrology where more domain knowledge and physical features can support further generalization. Researchers of deep learning in hydrology and water resources have increased significantly in the past several years, but there are limited considerations on physical features and processes. As shown in this study, deep learning models considering watershed-scale physical features and semi-distributed structures can work on multiple watersheds using regional models with limited sacrifice of accuracy. We reiterate our call to deep learning studies in hydrology to include more domain knowledge and physical features in the future.