Hydrological and empirical modeling framework for farmed prairie potholes in the prairie pothole region of Iowa

Thumbnail Image
Major Professor
Amy Kaleita
Committee Member
Journal Title
Journal ISSN
Volume Title

Closed surface depressions, also known as “potholes” play an important role in the hydrologic cycle and provide multiple environmental services including flood mitigation, water quality improvements, and wildlife habitat. In the Prairie Pothole Region, which covers approximately 715,000 km2, including parts of three Canadian provinces (Saskatchewan, Manitoba, and Alberta) and five states in the U.S. (Minnesota, Iowa, North and South Dakota, and Montana), these potholes are typically farmed and are a dominant feature in the landscape. These potholes are also different than the traditional prairie pothole wetlands as the natural vegetation (Typha spp., Scirpus spp., Carex spp., etc.) has been replaced by agricultural crops (mainly Corn and Soybean). In this study, we evaluated the Annualized Agriculture Non-Point Source (AnnAGNPS) model for simulating the inundation behavior of farmed potholes, in the Prairie Pothole Region (PPR) of Iowa. Performance analyses considered the entire growing season (GS), corresponding to the span in which there was observed data, and only days in which water storage (WS) was observed. Our results demonstrate that the AnnAGNPS model can be used to predict the inundation depth of drained and farmed potholes, which is useful for assessing the landscape impacts of these features. We then investigated the influence of different land use practices on depth, duration, and aerial extent of ponding in the two potholes using AnnAGNPS. Three management scenarios were compared — current: conventionally tilled farmed conditions in corn/soybean rotation with surface inlets in the potholes connecting to a subsurface drainage system; retired: pothole is converted to a mixture of grass, weeds, and low-growing brush, with surface inlets removed and the drainage system underneath the potholes disconnected; and conserved: conservation tillage throughout the field with surface inlets and drainage maintained in potholes. The average annual water depth for the conserved scenario was 7-8% lower than the average annual water depth for the current scenario. It was also observed that the potholes tend to flood more frequently in early stages of plant development, which could lead to delays in management operations and reduced yields.

Next, we assessed the capability of USGS DEMs for modeling pothole inundation in the prairie pothole region of Iowa. We used three DEMs: a 1m DEM prepared from LiDAR data which is readily available for the state of Iowa, USGS 1/9 arc-second DEM (~3m) which covers about 25 percent of the conterminous United States (U.S.) and 1/3 arc-second seamless DEM (~10m) which covers the entire U.S. Modeling performance was evaluated using Nash-Sutcliffe efficiency (NSE), Percent bias (PBIAS), Ratio of the root mean square error (RSR) and R2 statistical performance criteria. Results show that the water depth simulated from AnnAGNPS model based on 1m DEM which is prepared from the LiDAR data gave Nash-Sutcliffe efficiency (NSE) values of 0.77 and 0.24 in the Walnut pothole and 0.56 and 0.30 in the Bunny pothole, for the GS calibration and validation periods, respectively. The estimates of water depths using USGS 3m and 10m DEMs was also found to be very similar to LiDAR 1m DEM based predictions and are also representative of field conditions.

The developed AnnAGNPS model was then used to simulate the water depths for ten years (2007 – 2016) growing season (May to October) in the three potholes termed Bunny, Walnut and Lettuce. An empirical model based on artificial neural network (ANN) technology was developed on the expanded dataset and tested on the actual water depth observations collected in 2018 at another three potholes termed Turkey, Hen, and Plume. The R2 statistics were 0.604 and 0.563 during training and validation periods, respectively. A low root mean square error (RMSE) value of 0.057 and mean absolute error (MAE) value of 0.023 were found during both training and validation of the ANN model. In general, results suggest that the ANN models are able to predict the water depth fluctuations in the potholes during the growing season. These models can be a vital tool to augment the monitoring efforts of prairie potholes and can help stakeholders - farmers and state/federal agencies for management planning and making an informed decision about farming the potholes.

Sat Dec 01 00:00:00 UTC 2018