Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Prairie Pothole Management Support Tool: A web application for evaluating prairie pothole flood risk

Prairie Pothole Management Support Tool: A web application for evaluating prairie pothole flood risk AbbreviationsCcurrent farmed conditionCTconservation tillageDMLDes Moines LobeDSTdecision support toolHDhigh drainageLDlow drainageMDmedium drainageNPnatural prairie conditionsNTno tillagePPMSTPrairie Pothole Management Support ToolRpothole retirementRFrandom forestINTRODUCTIONDecision support tools (DSTs) are effective media for communicating and applying scientific principles in agricultural settings (Ranjan et al., 2020). Within agronomic and conservation‐oriented professionals, the desired utility of these tools includes field‐scale targeting of conservation practices, quantifying environmental or production‐oriented benefits to primary decision‐makers, and facilitating one‐on‐one discussions with primary decision makers (Ranjan et al., 2020). In the agricultural context, the principal roles that DST can fill are educating and informing stakeholders about how land use decisions can affect the hydrologic, biophysical, and environmental outcomes while acknowledging economic implications. Public researchers and private developers have made a suite of DSTs to support efficiency in agriculture, but these are frequently designed for the best land and maximizing profit in those areas. Their applicability in marginal areas is questionable or neglected, as implied by Liu et al. (2020), regarding the need for greater incorporation of waterlogging processes in such models. These DSTs may also fail to consider emerging markets or agronomic alternatives that are more appropriate in marginal areas.Prairie potholes are one example of marginal land and are a point of ambiguity from landowner and legislative perspectives (Schilling & Dinsmore, 2018). Prairie potholes are geographically isolated surface depressions common within the Des Moines Lobe (DML) of Iowa that flood intermittently after precipitation events. Despite the well‐developed use of complex hydrologic modeling (Evenson at al. et al., 2016; McKenna et al., 2018; Upadhyay et al., 2019), there is a scarcity of resources available to stakeholders for assessing the impacts of management on pothole hydrology and the associated economic and environmental implications. Landowners have focused their efforts on artificial drainage to improve crop growth and yield gains (Bishop et al., 1998). Because these potholes cover approximately 7–8% of the DML, and their watersheds represent about 44% of the DML (Green et al., 2019; Miller et al., 2009), their hydrologic and water quality impacts locally and downstream are of interest. However, trading drained land for improved crop conditions incurs many other environmental consequences, extending to the larger watershed (Green et al., 2019; Schilling & Dinsmore, 2018). We focus on these areas for their potential disproportionate impact on the environment. Drained potholes short circuit nutrient cycling by providing passage for sediment, sediment bound phosphorus, and dissolved nutrients such as nitrogen or orthophosphate (Lane et al., 2018). Drained potholes can act as hotspots of pesticide accumulation and export while groundwater nutrient concentrations are similarly elevated (Schilling & Dinsmore, 2018).Furthermore, farmed prairie potholes inconsistently provide financial benefit to landowners (Fey et al., 2016) and have no consistent framework for protection under traditional wetland conservation programs (Schilling & Dinsmore, 2018). Stakeholders need ways to assess economic and environmental trade‐offs. A first step in this is a tool that can evaluate the flooding risk for individual potholes under different management. Pothole‐specific conservation program trials (USDA, 2020) would also benefit from such a tool. Similarly, the potential for pothole wetlands to be included as a Conservation Effects Assessment Project (CEAP) conservation practice (McKenna et al., 2020) suggests the need for decision making tools. The objective of this technical note is to describe a web application developed for to allow users to explore flood risk of prairie potholes within the DML of Iowa based on morphologic and management characteristics.OVERVIEWThe Prairie Pothole Management Support Tool (PPMST) was developed to address the educational need concerning the impacts of topographic and management properties that contribute to the intensity of flooding within farmed prairie potholes. Designed with the R ‘shiny’ package (Chang et al., 2020), the PPMST is a free, simple, interactive web‐based application. The goal of PPMST is to provide a relative risk assessment of farmed potholes on an individual basis and encourage users to assess how land management strategies change the flood risk within potholes. Educational and high‐level information is needed to facilitate discussions with extension professionals and conservation planners regarding the best use of these marginal areas, which can encourage discussions about agricultural profitability and environmental stewardship (Muth, 2014).Core IdeasThere are few to no decision tools to help manage farmed prairie potholes.A developed web application assesses prairie pothole flood risk.The app integrates a random forest machine learning model to predict flood risk.The model requires topographic and land use data provided by the user.The app enables assessments of alternative land management in the Prairie Pothole Region.The PPMST risk evaluation is built upon a random forest (RF) machine learning model trained on a dataset of AnnAGNPS watershed simulation output, of which the models were calibrated to monitored potholes within the DML. A full discussion of the development of the RF model can be found in Nahkala et al. (2022). These scenarios considered the effects of drainage intensity, tillage, and cropping practices on the duration and extent of pothole inundation. These potholes reside within a few miles of each other in Story and Boone Counties, and this work assumes that they are representative of potholes across the DML. Characteristics of the potholes used in the development of the model are shown in Table 1.1TABLECharacteristics of potholes included in the training datasetPotholeMorphologyPothole areaWatershed area (AnnAGNPS)Overflow depthMax. concentrated flow pathMax. watershed reliefhamBunny5.3540.11.09036.1Walnut2.609.60.72984.4Lettuce2.1113.30.82405.4Cardinal1.4912.90.75375.5Hen0.513.50.52304.5Mouth0.949.60.43505.6Note. Morphology illustrates the pothole at its maximum extent (where the surface area would be when the pothole was fully inundated before spilling over at the outlet) within its microwatershed. From Nahkala et al. (2022).The Nahkala et al. (2022) RF model generates a unitless flood risk metric which can range in value from 0.1 to 10, with 0 being no risk and 10 being very high risk. The flood risk metric, the derivation of which is described in Nahkala et al. (2022), incorporates several factors:the fraction of each month in which the pothole is typically flooded, with flooding during the growing season more heavily weighted, acknowledging that more frequent flooding is more likely to result in yield losses;the volume (and thus spatial extent) of flooding that typically occurs in each month, with flooding during the growing season more heavily weighted, acknowledging that the larger the extent of flood‐affected crops, the more yield losses are likely;the typical duration of ponding events in each month, with the growing season more heavily weighted, acknowledging that longer periods of inundation risk more yield loss.The RF model computes flood risk from eight inputs. These include four quantitative values intrinsic to the pothole microwatershed: catchment area to pothole area ratio, maximum pothole depth, maximum watershed flow path, and maximum watershed relief. The four qualitative inputs depend on field management: land use of the pothole, land use of the field, drainage intensity, and tillage. The model predicted the flood risk metric with an R2 of >.9 and >.7 during calibration and validation, respectively, when compared with the risk metric calculated from the output of a more complex calibrated watershed model described in Nahkala et al. (2021).The RF input variables are entered on the user interface and are determined by the user for their specific prairie pothole, which allows customization and versatility. A supportive web map allows users to identify characteristics that are intrinsic to their field and helps provide an accurate assessment of existing conditions. Output includes a singular risk value based on the input scenario and a range of uncertainty on a 10‐point scale. When coupled with a side‐by‐side assessment of alternatives, users can observe the relative effect of parameters on flood risk within their pothole. The tool follows recommended DST design (Cerf et al., 2012; Cox, 1996; Rose et al., 2016) by remaining free and easy to use across a variety of scales, providing applicable results across a range of geographic gradients, and by providing only an assessment of flood risk, which allows users to apply that knowledge to their larger decision‐making context and goals.The PPMST is an open‐source R Shiny application hosted by the shinyapps.io server, developed by RStudio, PBC. The tool can be viewed in web browsers, which function as the graphical user interface (GUI). Programmed GUI reactivity and user interaction are driven by separate commands hosted on the server. The tool features two web tabs, one tab to provide inputs for a baseline assessment of a user‐identified prairie pothole and one tab to provide output for the baseline assessment and options for an alternative analysis of land management strategies.The baseline assessment page features an embedded copy of the user's manual for easy viewing (Figure 1). The manual can also be downloaded and opened with a PDF viewer. Below the manual is a sidebar, which includes 11 input fields that the user may modify, and a main page which hosts a Leaflet web map with basic spatial functionality. The baseline input fields are county ID, field ID, pothole land use, field land use, drainage, tillage, pothole area, watershed area, pothole depth, watershed slope, and pothole shape.1FIGUREMain user interface of the Prairie Pothole Management Support ToolThe county and field identifiers are text inputs that are copied to metadata in the final downloadable report. The subsequent nine inputs are used to determine input parameters for the RF model, which predicts the flood risk of the pothole in its user defined baseline condition. The first four inputs are qualitative factors and are limited to select levels on which the model was trained. The pothole area, watershed area, pothole depth, and slope are numerical inputs. The default values correspond to median values for potholes in the DML (McDeid et al., 2019). The watershed slope is used to estimate two parameters used in the RF model, maximum watershed relief and maximum flow path, and is calculated using the input areas and an assumption about the pothole shape. The two assumed potholes shapes include a perfect circle and an ellipse with a 2:1 ratio between its radii.The Leaflet web map supports three visual layers: default aerial imagery provided by ESRI, and a 2010 aerial imagery dataset and 3‐m digital elevation model from the Iowa Ortho GIS Server, obtained via the web mapping service. The visual layers allow users to identify areas of crop loss or crop stress in their fields. A drawing tool can be utilized to approximate the pothole and watershed areas, which provides a popup of area values.Once the baseline assessment is complete, users are provided with options to read short and long summaries of the flood risk assessment. Any incongruities in the input parameters trigger warning pop ups before this is complete. Users then have the option of assessing alternative land management strategies and can graphically and numerically view how the flood risk of the pothole changes (Figure 2). The risk is provided as an interval to account for uncertainty in the RF model by applying a one RMSE deviation above and below the predicted value. Finally, users have the option of reading about how the flood risk metric corresponds to duration and extent of flooding in the field, as seen from the training of the RF model (Nahkala et al., 2022).2FIGUREUser interface for the land management alternative analysis section of the Prairie Pothole Management Support ToolThe alternative analysis may be downloaded as an html report generated via an RMarkdown script. The report includes metadata about the pothole and analysis, a table of inputs for all scenarios, and graphical and tabular representations of the risk prediction for each scenario. An appendix is included which summarizes the risk once again in terms of flood duration and extent of flooding based on modeling.EXAMPLE APPLICATIONSWe used the PPMST to consider a single farmed prairie pothole under conventional management with no drainage. We used the PPMST to establish the baseline flood risk, and then assessed a series of alternative field management options a farmer might consider modifying the hydrology of their pothole. Next, five of those scenarios were applied to three additional potholes to demonstrate additional comparisons and interpretations that could come from using this tool. Results are summarized in Figure 3. The potholes used in these examples are named Hen, Bunny, Walnut, and Mouth based on their unique aerial image signature, as the potholes in these studies are within the same research fields and thus do not have unique management identifiers for naming (Martin et al., 2019). Hen, Bunny, and Walnut are conventionally farmed within Iowa State University‐managed research farms in Story County, Iowa. Hen has no known drainage, Bunny has two surface inlets with subsurface drainage, and Walnut has a surface inlet with subsurface drainage. Mouth resides >5 km northwest and is planted in perennial vegetation. Bunny has the largest watershed and is the deepest (1.0 m). Its watershed has significantly longer flow paths compared to other potholes. Walnut has a small watershed compared with the ponded area of the pothole but is moderately deep (0.7 m). Hen is the smallest pothole (0.5 ha), while Mouth is the shallowest (0.4 m). Mouth has largest ratio of drainage area to pothole area (10.2). These potholes were chosen for analysis because their watersheds had been previously parameterized in the development of the machine learning model and significant monitoring can more easily bolster the discussion of modeled risk.3FIGUREScenario acronyms include C (current), NP (natural prairie), LD (low drainage), MD (moderate drainage), HD (high drainage), R (retired), CT (conservation tillage), and NT (no till). Panel A: Flood risk of Hen under current conditions and 12 alternative conditions. Panel B: Comparison of flood risk for four potholes across five management scenarios.Example assessment of management on a single potholeThis PPMST is most effective at comparing multiple scenarios for a single pothole. Here, Hen, one of the potholes used to calibrate the RF model, is assessed under its current conditions and 12 alternative scenarios. Current farmed conditions (C) and natural prairie conditions (NP, which considers the entire microwatershed planted with perennial vegetation) are considered potential baseline conditions (Figure 3). All three levels of additional drainage were assessed (LD, MD, and HD), as well as pothole retirement (R), conservation tillage (CT), and no till (NT) management. We also combined practices, assessing low drainage with retirement which represents a strong effort to retire a pothole (LD‐R), low drainage with conservation tillage (LD‐CT), low drainage with no till (LD‐NT), retirement of the pothole and conservation tillage in the field (R‐CT) and retirement of the pothole and no till (R‐NT). We assessed changes to the predicted risk and considered implications and interpretations, discussed in the following sections. Predicted risk is displayed in panel A of Figure 3.Assessing drainage practicesDrainage is compared directly in 4 scenarios, which includes the baseline (current, C) condition, and the three scenarios that only change drainage in the risk prediction model (LD, MD, HD). The undrained flood risk in Hen was predicted to be 5.6 while increasing drainage across our three scenarios reduced this value to 1.5, 0.9, and 0.7, respectively. We see that increasing drainage significantly reduces flood risk. However, there are diminishing returns with adding substantial surface inlet capacity, similarly observed in related water balance studies of farmed potholes (Schilling et al., 2019). While the initial investment in subsurface drainage decreases risk significantly from 5.6 to 1.5, adding surface inlets does not tend to significantly reduce the flooding and flood risk, likely because drain tiles may already be operating at capacity. This is reflected in the minimal decrease from 1.5 to 0.9 and 0.7 based on our flood risk metric. If making this land more farmable is the primary goal, drainage would significantly reduce flooding, but may still carry intermittent risk.Assessing land use practicesLand use practices are directly compared in panel A of Figure 3. These scenarios include retiring the pothole to perennial vegetation (R), implementing conservation tillage (CT) and switching to no till (NT). Compared with the base flood risk of 5.6, retirement and conservation tillage do not change the value at this level of significant digits. Here, we see that if farming this land is the primary goal, these practices do not provide direct benefits based on our modeling. This may change with further model parameterization, training, and representation of tillage practices (in both AnnAGNPS and the RF). However, we see no till reduces the risk by 0.4 to 5.2. This change within our risk scale is small but not negligible and suggest that no till would provide field‐scale benefits and marginally reduce flooding in farmed potholes. We can show that the pothole surface area delineation provides an appropriate boundary for farmed area that could be removed from production.Assessing combined practicesFive scenarios include combinations of low drainage (LD) and land use practices (R, CT, NT). Here, the LD‐R, LD‐CT, and LD‐NT scenarios predicted an equivalent risk to the LD scenario (1.5). This mirrors results from the land use comparison under undrained conditions, but here we see no impact from no till. These highlight interactive effects of combined practices, where drainage dominates flood risk value while land use practices have less effect. While no till provides benefits, the intensity of drainage incurs larger effects on the flood risk and dampens the effect of no till under higher drainage. Ignoring larger contexts, this model shows that creating farmable conditions in prairie potholes can focus solely on drainage. We see in undrained conditions, stacking retirement of the pothole with tillage slightly reduces flood risk. The R‐CT scenario reduces baseline risk from 5.6 to 5.5 and the R‐NT reduces risk to 5.1. These values are 0.1 lower than assessing CT and NT on their own. Overall, we see incremental decreases in flood risk by using multiple land use practices. This may not be highly relevant to drained potholes but may suggest field management is a relevant factor in establishing appropriate vegetation populations during pothole retirement.Trade‐offsThe prior discussion compares alternatives to Hen's farmed conditions, representing potential competing management methods. It does not, however, provide the context for making decisions. Presented with flood risk, a user can then consider related objectives including crop production, financial stability, water quality, or habitat conservation within the aerial extent of a pothole. Users must reference external knowledge of the benefits and consequences of drainage, tillage practices, or land retirement on these additional objectives to make informed decisions. This is because the PPMST is meant to inform but does not make recommendations for the user. Discussions can be facilitated by extension specialists who have specialized knowledge of these practices, as necessary. Users can further compare farmed conditions to natural prairie (NP). Hen's NP predicted flood risk is 2.1, which is of similar magnitude to LD. This requires the entire microwatershed to be planted with perennials, and the actual flood risk value is no longer relevant to farmers. However, this may be relevant to conservation biologists and ecologists who rely on a knowledge of hydroperiod and land extents for population establishment.Example comparisons between potholes under the same conditionsFour potholes, now including Hen, Bunny, Walnut, and Mouth, were modeled in their current condition. The retired (R) and no till (NT) scenarios (Figure 3, panel B) retired the pothole to perennial vegetation and applied no till to the entire planted watershed, respectively. The low drainage and retired (LD‐R) scenario removed surface inlets from drain tile and planted perennial vegetation in the pothole. The NP scenario retired the entire watershed to perennial vegetation, representing either a significant conservation effort or historic natural prairie.Assessing retirementIn our case study, the only change that was made between the C and R scenarios was the removal of the pothole area from production. This land was reverted to perennial vegetation, as might be seen via a conservation mechanism like the Conservation Reserve Program. Figure 3 shows that the flood risk prediction for each pothole does not change, or changes are not sensitive enough to be shown in this model, like our assessment of Hen. Physically, this represents the same extent of frequency of flooding expected under the same weather patterns compared to a completely farmed system. This can be seen positively or negatively by landowners, but regardless, because flood risk is not increasing, this scenario shows that a landowner could confidently retire any of the potholes without being concerned with having to remove extra land out of production where profit margins are more desirable.Assessing intentional conservationThe LD‐R scenario considers a scenario in which these potholes all had drainage installed but are now being retired from production and have surface inlets removed. This might result from more intentional conservation efforts, where surface inlets, if present, are removed to reduce pollutant export (N and P) and grasses are planted to promote wildlife habitat and biodiversity benefits. Here, we observe that risk for Bunny and Walnut increased compared to their current conditions. This is due to the reduced drainage capacity and is an important consideration for landowners. In this case, overestimating the retired land area may be necessary to reduce the risk of crop drown out in the adjacent field, which may become marginally more frequent at the boundary of the pothole.Assessing the inherent risk of a potholeLandowners deal with multiple prairie potholes, and many are managed differently despite their proximity and similar field management. If landowners or conservation specialists (e.g., a watershed coordinator) are assessing multiple potholes, they might compare multiple potholes under the same conditions. Comparing potholes in the NP scenario elucidates potential outcomes. The flood risk in this scenario is considered the inherent risk, reflecting the historical natural drainage and land use conditions of a pothole. In natural conditions, where the entire pothole microwatershed is treated as perennial vegetation, we observe that Mouth has the highest risk while Walnut has the lowest. If one of these potholes were slated for retirement, Mouth might be objectively chosen because it has higher risk morphological characteristics. These characteristics might include its large catchment area to pothole ratio (largest of those studied) or its low overflow depth which leads to frequent inundation of the entire pothole. Anecdotally, reality corroborates this interpretation because Mouth was the only pothole in our study that has been retired. Conversely, Walnut is the better candidate for investment in drainage because its inherent risk is smaller.Assessing incremental gainsNo till assesses the removal of tillage practices from the current management. We observe no change in risk for Bunny. However, Walnut's risk decreased by 0.1, while Mouth and Hen decreased by 0.4. We see modified tillage in highly drained potholes results in minimal to no change while these same activities might have a larger effect for field conditions that are higher risk. This result suggests that the drainage infrastructure we see in potholes can only handle so much natural variation in climate and flood risk will remain as long as drainage remains within the range we have observed and incorporated within the RF model. Potholes will likely become inundated intermittently in wet years, affecting crop yields.Context from modeling dataInquisitive users are provided a table that interprets risk level in terms of watershed simulation data. The predicted risk in our case study ranged from 0.6 to 5.6. Because this scale is multiplicative, anything above a risk of 5 is largely indistinguishable based on the training dataset. These high‐risk conditions flood to 30% of their areal extent almost every year and to 60% of their areal extents in 65–80% of the years. Conversely, low risk conditions flood to 30% of their aerial extent around 80% of years, whereas they reach 60% of their area only 40–50% of the years. Expected average annual days of inundation are 3–12 d for extremely low risk potholes and 16–58 d for extremely high‐risk potholes, but these values can extend to over 100 d (Nahkala et al., 2021). The risk data and modeling study data (described in this section) are provided without management recommendations, allowing users to incorporate their own context for making decisions.Limitations and future developmentFuture development may focus on the integration of raster processing within the web map that automatically calculates watershed parameters from digital elevation models to improve the accuracy of user inputs. Additionally, using larger training datasets with the RF model will ensure that the model captures more variation in prairie pothole characteristics observed across the DML. The model currently captures only a small range of observed land management conditions, and calibrated watershed models (Nahkala et al., 2021), which are what the RF model was trained against, were only available for conventional farming operations. The range of morphologic inputs is limited based on the physical proximity of the potholes used to establish the model, and the known range of these topographic characteristics is larger than included in the model training. Further models could similarly include variation in annual precipitation or climate change projections, as both the geographic and temporal variation may influence current and future flood risk. Additionally, the tool could be paired with a profitability analysis so that users could consider costs and benefits of management options given an expected impact (or lack of impact) on crop yield. Finally, the tool could also be expanded to include an assessment of downstream effects. The focus on an individual pothole ignores the hydrologic connection to downstream waters, and thus increasing the drainage may be impossible or ineffective in some cases if capacity does not exist downstream and may increase water yield and flooding or downstream in other cases. In its current form, the tool is primarily for informational and educational purposes, and a disclaimer to this effect is included with the tool.Software availabilityThe PPMST can be accessed via: https://bnahkala.shinyapps.io/ppmst/. The user's manual is embedded on the PPMST homepage. R code is hosted on GitHub: https://github.com/bnahkala/ppmstACKNOWLEDGMENTSThis work was supported by the US Environmental Protection Agency assistance agreement CD97753901. It has not been subjected to the Agency's review and might not necessarily reflect the views of the Agency. This work was further supported by the United States Department of Agriculture Hatch Projects IOW04414 and IOW05474 and the National Institute for Food and Agriculture (2018‐67019‐27886).AUTHOR CONTRIBUTIONSBrady A. Nahkala: Conceptualization; Data curation; Formal analysis; Methodology; Software; Writing – original draft; Writing – review & editing. Amy L. Kaleita: Conceptualization; Funding acquisition; Resources; Supervision; Writing – review & editing. Michelle L. Soupir: Conceptualization; Funding acquisition; Resources; Supervision; Writing – review & editing. Andy VanLoocke: Conceptualization; Funding acquisition; Writing – review & editing.CONFLICT OF INTERESTThe authors declare no conflict of interest.REFERENCESBishop, R. A., Joens, J., & Zohrer, J. (1998). Iowa's wetlands, present and future with a focus on prairie potholes. Journal of the Iowa Academy of Science, 105(3), 89–93.Cerf, M., Jeuffroy, M. H., Prost, L., & Meynard, J. M. (2012). Participatory design of agricultural decision support tools: Taking account of the use situations. Agronomy for Sustainable Development, 32(4), 899–910. https://doi.org/10.1007/s13593‐012‐0091‐zChang, W., Cheng, J., Allaire, J., Xie, Y., & McPeherson, J. (2020). shiny: Web application framework for R (R package version 1.5.0). The Comprehensive R Archive Network. https://CRAN.R‐project.org/package=shinyCox, P. G. (1996). Some issues in the design of agricultural decision support systems. Agricultural Systems, 52(2–3), 355–381. https://doi.org/10.1016/0308‐521X(96)00063‐7Evenson, G. R., Golden, H. E., Lane, C. R., & D'Amico, E. (2016). An improved representation of geographically isolated wetlands in a watershed‐scale hydrologic model. Hydrological Processes, 30(22), 4168–4184. https://doi.org/10.1002/hyp.10930Fey, S., Kyveryga, P., Connor, J., Kiel, A., & Muth, D. (2016). Within‐field profitability assessment: Impact of weather, field management and soils. In Proceedings of the 13th International Conference of Precision Agriculture. International Society of Precision Agriculture.Green, D. I. S., McDeid, S. M., & Crumpton, W. G. (2019). Runoff storage potential of drained upland depressions on the Des Moines Lobe of Iowa. Journal of the American Water Resources Association, 55(3), 543–558. https://doi.org/10.1111/1752‐1688.12738Lane, C. R., Leibowitz, S. G., Autrey, B. C., LeDuc, S. D., & Alexander, L. C. (2018). Hydrological, physical, and chemical functions and connectivity of non‐floodplain wetlands to downstream waters: A review. Journal of the American Water Resources Association, 54(2), 346–371. https://doi.org/10.1111/1752‐1688.12633Liu, K., Harrison, M. T., Shabala, S., Meinke, H., Ahmend, I., Zhang, Y., Tian, X., & Zhou, M. (2020). The state of the art in modelling waterlogging impacts on plants: What do we know and what do we need to know. Earth's Future, 8, https://doi.org/10.1029/2020ef001801Martin, A. R., Kaleita, A. L., & Soupir, M. L. (2019). Inundation patterns of farmed pothole depressions with varying subsurface drainage. Transactions of the ASABE, 62, 1579–1590. https://doi.org/10.13031/trans.13435McDeid, S. M., Green, D. I. S., & Crumpton, W. G. (2019). Morphology of drained upland depressions on the Des Moines Lobe of Iowa. Wetlands, 39, 587–600. https://doi.org/10.1007/s13157‐018‐1108‐4McKenna, O. P., Mushet, D. M., Scherff, E. J., McKean, K. I., & Mills, C. T. (2018). The Pothole Hydrology‐Linked Systems Simulator (PHyLiSS): Development and application of a systems model for prairie‐pothole wetlands (USGS Report 2018‐1165). USDOI, USGS.McKenna, O. P., Osorio, J. M., Behrman, K. D., Doro, L., & Mushet, D. M. (2020). Development of a novel framework for modeling field‐scale conservation effects of depressional wetlands in agricultural landscapes. Journal of Soil and Water Conservation, 75(6), 695–703. https://doi.org/10.2489/jswc.2020.00096Miller, B. A., Crumpton, W. G., & van der Valk, A. G. (2009). Spatial distribution of historical wetland classes on the Des Moines lobe, Iowa. Wetlands, 29, 1146–1152. https://doi.org/10.1672/08‐158.1Muth, D. (2014). Profitability versus environmental performance: Are they competing? Journal of Soil and Water Conservation, 69, 203A–206A. https://doi.org/10.2489/jswc.69.6.203ANahkala, B. A., Kaleita, A. L., & Soupir, M. L. (2021). Characterization of prairie pothole inundation using AnnAGNPS under varying management and drainage scenarios. Agricultural Water Management, 255, 107002. https://doi.org/10.1016/j.agwat.2021.107002Nahkala, B. A., Kaleita, A. L., & Soupir, M. L. (2022). Empirical tool development for prairie pothole management using AnnAGNPS and random forest. Environ. Modelling & Software, 147, 105241. https://doi.org/10.1016/j.envsoft.2021.105241Ranjan, P., Duriancik, L. F., Moriasi, D. N., Carlson, D., Anderson, K., & Prokopy, L. S. (2020). Understanding the use of decision support tools by conservation professionals and their education and training needs: An application of the Reasoned Action Approach. Journal of Soil and Water Conservation, 75, 387–399. https://doi.org/10.2489/jswc.75.3.387Rose, D. C., Sutherland, W. J., Parker, C., Lobley, M., Winter, M., Morris, C., Twining, S., Ffoulkes, C., Amano, T., & Dicks, L. V. (2016). Decision support tools for agriculture: Towards effective design and delivery. Agricultural Systems, 149, 165–174. https://doi.org/10.1016/j.agsy.2016.09.009Schilling, K. E., & Dinsmore, S. (2018). Monitoring the wildlife, hydrology and water quality of drained wetlands of the Des Moines Lobe, northern Iowa: Introduction to special feature. Wetlands, 38(2), 207–210. https://doi.org/10.1007/s13157‐017‐0989‐ySchilling, K. E., Then, S. R., & Ikenberry, C. D. (2019). Water balance modeling of temporary ponding in a drained prairie pothole wetland. Environmental Modeling and Assessment, 24, 37–48. https://doi.org/10.1007/s10666‐018‐9596‐4Upadhyay, P., Pruski, L. O. S., Kaleita, A. L., & Soupir, M. L. (2019). Effects of land management on inundation of prairie pothole wetlands in the Des Moines Lobe using AnnAGNPS. Agricultural Water Management, 213, 947–956. https://doi.org/10.1016/j.agwat.2018.12.016USDA. (2020). Prairie Pothole Water Quality and Wildlife Program. USDA. https://www.nrcs.usda.gov/wps/portal/nrcs/ia/programs/financial/eqip/a5eaadd4‐eb4b‐4cd9‐b5c9‐11e1b380cb45/ http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Agrosystems, Geosciences & Environment" Wiley

Prairie Pothole Management Support Tool: A web application for evaluating prairie pothole flood risk

Loading next page...
 
/lp/wiley/prairie-pothole-management-support-tool-a-web-application-for-Lroq8idu6R
Publisher
Wiley
Copyright
© 2022 Crop Science Society of America and American Society of Agronomy.
eISSN
2639-6696
DOI
10.1002/agg2.20280
Publisher site
See Article on Publisher Site

Abstract

AbbreviationsCcurrent farmed conditionCTconservation tillageDMLDes Moines LobeDSTdecision support toolHDhigh drainageLDlow drainageMDmedium drainageNPnatural prairie conditionsNTno tillagePPMSTPrairie Pothole Management Support ToolRpothole retirementRFrandom forestINTRODUCTIONDecision support tools (DSTs) are effective media for communicating and applying scientific principles in agricultural settings (Ranjan et al., 2020). Within agronomic and conservation‐oriented professionals, the desired utility of these tools includes field‐scale targeting of conservation practices, quantifying environmental or production‐oriented benefits to primary decision‐makers, and facilitating one‐on‐one discussions with primary decision makers (Ranjan et al., 2020). In the agricultural context, the principal roles that DST can fill are educating and informing stakeholders about how land use decisions can affect the hydrologic, biophysical, and environmental outcomes while acknowledging economic implications. Public researchers and private developers have made a suite of DSTs to support efficiency in agriculture, but these are frequently designed for the best land and maximizing profit in those areas. Their applicability in marginal areas is questionable or neglected, as implied by Liu et al. (2020), regarding the need for greater incorporation of waterlogging processes in such models. These DSTs may also fail to consider emerging markets or agronomic alternatives that are more appropriate in marginal areas.Prairie potholes are one example of marginal land and are a point of ambiguity from landowner and legislative perspectives (Schilling & Dinsmore, 2018). Prairie potholes are geographically isolated surface depressions common within the Des Moines Lobe (DML) of Iowa that flood intermittently after precipitation events. Despite the well‐developed use of complex hydrologic modeling (Evenson at al. et al., 2016; McKenna et al., 2018; Upadhyay et al., 2019), there is a scarcity of resources available to stakeholders for assessing the impacts of management on pothole hydrology and the associated economic and environmental implications. Landowners have focused their efforts on artificial drainage to improve crop growth and yield gains (Bishop et al., 1998). Because these potholes cover approximately 7–8% of the DML, and their watersheds represent about 44% of the DML (Green et al., 2019; Miller et al., 2009), their hydrologic and water quality impacts locally and downstream are of interest. However, trading drained land for improved crop conditions incurs many other environmental consequences, extending to the larger watershed (Green et al., 2019; Schilling & Dinsmore, 2018). We focus on these areas for their potential disproportionate impact on the environment. Drained potholes short circuit nutrient cycling by providing passage for sediment, sediment bound phosphorus, and dissolved nutrients such as nitrogen or orthophosphate (Lane et al., 2018). Drained potholes can act as hotspots of pesticide accumulation and export while groundwater nutrient concentrations are similarly elevated (Schilling & Dinsmore, 2018).Furthermore, farmed prairie potholes inconsistently provide financial benefit to landowners (Fey et al., 2016) and have no consistent framework for protection under traditional wetland conservation programs (Schilling & Dinsmore, 2018). Stakeholders need ways to assess economic and environmental trade‐offs. A first step in this is a tool that can evaluate the flooding risk for individual potholes under different management. Pothole‐specific conservation program trials (USDA, 2020) would also benefit from such a tool. Similarly, the potential for pothole wetlands to be included as a Conservation Effects Assessment Project (CEAP) conservation practice (McKenna et al., 2020) suggests the need for decision making tools. The objective of this technical note is to describe a web application developed for to allow users to explore flood risk of prairie potholes within the DML of Iowa based on morphologic and management characteristics.OVERVIEWThe Prairie Pothole Management Support Tool (PPMST) was developed to address the educational need concerning the impacts of topographic and management properties that contribute to the intensity of flooding within farmed prairie potholes. Designed with the R ‘shiny’ package (Chang et al., 2020), the PPMST is a free, simple, interactive web‐based application. The goal of PPMST is to provide a relative risk assessment of farmed potholes on an individual basis and encourage users to assess how land management strategies change the flood risk within potholes. Educational and high‐level information is needed to facilitate discussions with extension professionals and conservation planners regarding the best use of these marginal areas, which can encourage discussions about agricultural profitability and environmental stewardship (Muth, 2014).Core IdeasThere are few to no decision tools to help manage farmed prairie potholes.A developed web application assesses prairie pothole flood risk.The app integrates a random forest machine learning model to predict flood risk.The model requires topographic and land use data provided by the user.The app enables assessments of alternative land management in the Prairie Pothole Region.The PPMST risk evaluation is built upon a random forest (RF) machine learning model trained on a dataset of AnnAGNPS watershed simulation output, of which the models were calibrated to monitored potholes within the DML. A full discussion of the development of the RF model can be found in Nahkala et al. (2022). These scenarios considered the effects of drainage intensity, tillage, and cropping practices on the duration and extent of pothole inundation. These potholes reside within a few miles of each other in Story and Boone Counties, and this work assumes that they are representative of potholes across the DML. Characteristics of the potholes used in the development of the model are shown in Table 1.1TABLECharacteristics of potholes included in the training datasetPotholeMorphologyPothole areaWatershed area (AnnAGNPS)Overflow depthMax. concentrated flow pathMax. watershed reliefhamBunny5.3540.11.09036.1Walnut2.609.60.72984.4Lettuce2.1113.30.82405.4Cardinal1.4912.90.75375.5Hen0.513.50.52304.5Mouth0.949.60.43505.6Note. Morphology illustrates the pothole at its maximum extent (where the surface area would be when the pothole was fully inundated before spilling over at the outlet) within its microwatershed. From Nahkala et al. (2022).The Nahkala et al. (2022) RF model generates a unitless flood risk metric which can range in value from 0.1 to 10, with 0 being no risk and 10 being very high risk. The flood risk metric, the derivation of which is described in Nahkala et al. (2022), incorporates several factors:the fraction of each month in which the pothole is typically flooded, with flooding during the growing season more heavily weighted, acknowledging that more frequent flooding is more likely to result in yield losses;the volume (and thus spatial extent) of flooding that typically occurs in each month, with flooding during the growing season more heavily weighted, acknowledging that the larger the extent of flood‐affected crops, the more yield losses are likely;the typical duration of ponding events in each month, with the growing season more heavily weighted, acknowledging that longer periods of inundation risk more yield loss.The RF model computes flood risk from eight inputs. These include four quantitative values intrinsic to the pothole microwatershed: catchment area to pothole area ratio, maximum pothole depth, maximum watershed flow path, and maximum watershed relief. The four qualitative inputs depend on field management: land use of the pothole, land use of the field, drainage intensity, and tillage. The model predicted the flood risk metric with an R2 of >.9 and >.7 during calibration and validation, respectively, when compared with the risk metric calculated from the output of a more complex calibrated watershed model described in Nahkala et al. (2021).The RF input variables are entered on the user interface and are determined by the user for their specific prairie pothole, which allows customization and versatility. A supportive web map allows users to identify characteristics that are intrinsic to their field and helps provide an accurate assessment of existing conditions. Output includes a singular risk value based on the input scenario and a range of uncertainty on a 10‐point scale. When coupled with a side‐by‐side assessment of alternatives, users can observe the relative effect of parameters on flood risk within their pothole. The tool follows recommended DST design (Cerf et al., 2012; Cox, 1996; Rose et al., 2016) by remaining free and easy to use across a variety of scales, providing applicable results across a range of geographic gradients, and by providing only an assessment of flood risk, which allows users to apply that knowledge to their larger decision‐making context and goals.The PPMST is an open‐source R Shiny application hosted by the shinyapps.io server, developed by RStudio, PBC. The tool can be viewed in web browsers, which function as the graphical user interface (GUI). Programmed GUI reactivity and user interaction are driven by separate commands hosted on the server. The tool features two web tabs, one tab to provide inputs for a baseline assessment of a user‐identified prairie pothole and one tab to provide output for the baseline assessment and options for an alternative analysis of land management strategies.The baseline assessment page features an embedded copy of the user's manual for easy viewing (Figure 1). The manual can also be downloaded and opened with a PDF viewer. Below the manual is a sidebar, which includes 11 input fields that the user may modify, and a main page which hosts a Leaflet web map with basic spatial functionality. The baseline input fields are county ID, field ID, pothole land use, field land use, drainage, tillage, pothole area, watershed area, pothole depth, watershed slope, and pothole shape.1FIGUREMain user interface of the Prairie Pothole Management Support ToolThe county and field identifiers are text inputs that are copied to metadata in the final downloadable report. The subsequent nine inputs are used to determine input parameters for the RF model, which predicts the flood risk of the pothole in its user defined baseline condition. The first four inputs are qualitative factors and are limited to select levels on which the model was trained. The pothole area, watershed area, pothole depth, and slope are numerical inputs. The default values correspond to median values for potholes in the DML (McDeid et al., 2019). The watershed slope is used to estimate two parameters used in the RF model, maximum watershed relief and maximum flow path, and is calculated using the input areas and an assumption about the pothole shape. The two assumed potholes shapes include a perfect circle and an ellipse with a 2:1 ratio between its radii.The Leaflet web map supports three visual layers: default aerial imagery provided by ESRI, and a 2010 aerial imagery dataset and 3‐m digital elevation model from the Iowa Ortho GIS Server, obtained via the web mapping service. The visual layers allow users to identify areas of crop loss or crop stress in their fields. A drawing tool can be utilized to approximate the pothole and watershed areas, which provides a popup of area values.Once the baseline assessment is complete, users are provided with options to read short and long summaries of the flood risk assessment. Any incongruities in the input parameters trigger warning pop ups before this is complete. Users then have the option of assessing alternative land management strategies and can graphically and numerically view how the flood risk of the pothole changes (Figure 2). The risk is provided as an interval to account for uncertainty in the RF model by applying a one RMSE deviation above and below the predicted value. Finally, users have the option of reading about how the flood risk metric corresponds to duration and extent of flooding in the field, as seen from the training of the RF model (Nahkala et al., 2022).2FIGUREUser interface for the land management alternative analysis section of the Prairie Pothole Management Support ToolThe alternative analysis may be downloaded as an html report generated via an RMarkdown script. The report includes metadata about the pothole and analysis, a table of inputs for all scenarios, and graphical and tabular representations of the risk prediction for each scenario. An appendix is included which summarizes the risk once again in terms of flood duration and extent of flooding based on modeling.EXAMPLE APPLICATIONSWe used the PPMST to consider a single farmed prairie pothole under conventional management with no drainage. We used the PPMST to establish the baseline flood risk, and then assessed a series of alternative field management options a farmer might consider modifying the hydrology of their pothole. Next, five of those scenarios were applied to three additional potholes to demonstrate additional comparisons and interpretations that could come from using this tool. Results are summarized in Figure 3. The potholes used in these examples are named Hen, Bunny, Walnut, and Mouth based on their unique aerial image signature, as the potholes in these studies are within the same research fields and thus do not have unique management identifiers for naming (Martin et al., 2019). Hen, Bunny, and Walnut are conventionally farmed within Iowa State University‐managed research farms in Story County, Iowa. Hen has no known drainage, Bunny has two surface inlets with subsurface drainage, and Walnut has a surface inlet with subsurface drainage. Mouth resides >5 km northwest and is planted in perennial vegetation. Bunny has the largest watershed and is the deepest (1.0 m). Its watershed has significantly longer flow paths compared to other potholes. Walnut has a small watershed compared with the ponded area of the pothole but is moderately deep (0.7 m). Hen is the smallest pothole (0.5 ha), while Mouth is the shallowest (0.4 m). Mouth has largest ratio of drainage area to pothole area (10.2). These potholes were chosen for analysis because their watersheds had been previously parameterized in the development of the machine learning model and significant monitoring can more easily bolster the discussion of modeled risk.3FIGUREScenario acronyms include C (current), NP (natural prairie), LD (low drainage), MD (moderate drainage), HD (high drainage), R (retired), CT (conservation tillage), and NT (no till). Panel A: Flood risk of Hen under current conditions and 12 alternative conditions. Panel B: Comparison of flood risk for four potholes across five management scenarios.Example assessment of management on a single potholeThis PPMST is most effective at comparing multiple scenarios for a single pothole. Here, Hen, one of the potholes used to calibrate the RF model, is assessed under its current conditions and 12 alternative scenarios. Current farmed conditions (C) and natural prairie conditions (NP, which considers the entire microwatershed planted with perennial vegetation) are considered potential baseline conditions (Figure 3). All three levels of additional drainage were assessed (LD, MD, and HD), as well as pothole retirement (R), conservation tillage (CT), and no till (NT) management. We also combined practices, assessing low drainage with retirement which represents a strong effort to retire a pothole (LD‐R), low drainage with conservation tillage (LD‐CT), low drainage with no till (LD‐NT), retirement of the pothole and conservation tillage in the field (R‐CT) and retirement of the pothole and no till (R‐NT). We assessed changes to the predicted risk and considered implications and interpretations, discussed in the following sections. Predicted risk is displayed in panel A of Figure 3.Assessing drainage practicesDrainage is compared directly in 4 scenarios, which includes the baseline (current, C) condition, and the three scenarios that only change drainage in the risk prediction model (LD, MD, HD). The undrained flood risk in Hen was predicted to be 5.6 while increasing drainage across our three scenarios reduced this value to 1.5, 0.9, and 0.7, respectively. We see that increasing drainage significantly reduces flood risk. However, there are diminishing returns with adding substantial surface inlet capacity, similarly observed in related water balance studies of farmed potholes (Schilling et al., 2019). While the initial investment in subsurface drainage decreases risk significantly from 5.6 to 1.5, adding surface inlets does not tend to significantly reduce the flooding and flood risk, likely because drain tiles may already be operating at capacity. This is reflected in the minimal decrease from 1.5 to 0.9 and 0.7 based on our flood risk metric. If making this land more farmable is the primary goal, drainage would significantly reduce flooding, but may still carry intermittent risk.Assessing land use practicesLand use practices are directly compared in panel A of Figure 3. These scenarios include retiring the pothole to perennial vegetation (R), implementing conservation tillage (CT) and switching to no till (NT). Compared with the base flood risk of 5.6, retirement and conservation tillage do not change the value at this level of significant digits. Here, we see that if farming this land is the primary goal, these practices do not provide direct benefits based on our modeling. This may change with further model parameterization, training, and representation of tillage practices (in both AnnAGNPS and the RF). However, we see no till reduces the risk by 0.4 to 5.2. This change within our risk scale is small but not negligible and suggest that no till would provide field‐scale benefits and marginally reduce flooding in farmed potholes. We can show that the pothole surface area delineation provides an appropriate boundary for farmed area that could be removed from production.Assessing combined practicesFive scenarios include combinations of low drainage (LD) and land use practices (R, CT, NT). Here, the LD‐R, LD‐CT, and LD‐NT scenarios predicted an equivalent risk to the LD scenario (1.5). This mirrors results from the land use comparison under undrained conditions, but here we see no impact from no till. These highlight interactive effects of combined practices, where drainage dominates flood risk value while land use practices have less effect. While no till provides benefits, the intensity of drainage incurs larger effects on the flood risk and dampens the effect of no till under higher drainage. Ignoring larger contexts, this model shows that creating farmable conditions in prairie potholes can focus solely on drainage. We see in undrained conditions, stacking retirement of the pothole with tillage slightly reduces flood risk. The R‐CT scenario reduces baseline risk from 5.6 to 5.5 and the R‐NT reduces risk to 5.1. These values are 0.1 lower than assessing CT and NT on their own. Overall, we see incremental decreases in flood risk by using multiple land use practices. This may not be highly relevant to drained potholes but may suggest field management is a relevant factor in establishing appropriate vegetation populations during pothole retirement.Trade‐offsThe prior discussion compares alternatives to Hen's farmed conditions, representing potential competing management methods. It does not, however, provide the context for making decisions. Presented with flood risk, a user can then consider related objectives including crop production, financial stability, water quality, or habitat conservation within the aerial extent of a pothole. Users must reference external knowledge of the benefits and consequences of drainage, tillage practices, or land retirement on these additional objectives to make informed decisions. This is because the PPMST is meant to inform but does not make recommendations for the user. Discussions can be facilitated by extension specialists who have specialized knowledge of these practices, as necessary. Users can further compare farmed conditions to natural prairie (NP). Hen's NP predicted flood risk is 2.1, which is of similar magnitude to LD. This requires the entire microwatershed to be planted with perennials, and the actual flood risk value is no longer relevant to farmers. However, this may be relevant to conservation biologists and ecologists who rely on a knowledge of hydroperiod and land extents for population establishment.Example comparisons between potholes under the same conditionsFour potholes, now including Hen, Bunny, Walnut, and Mouth, were modeled in their current condition. The retired (R) and no till (NT) scenarios (Figure 3, panel B) retired the pothole to perennial vegetation and applied no till to the entire planted watershed, respectively. The low drainage and retired (LD‐R) scenario removed surface inlets from drain tile and planted perennial vegetation in the pothole. The NP scenario retired the entire watershed to perennial vegetation, representing either a significant conservation effort or historic natural prairie.Assessing retirementIn our case study, the only change that was made between the C and R scenarios was the removal of the pothole area from production. This land was reverted to perennial vegetation, as might be seen via a conservation mechanism like the Conservation Reserve Program. Figure 3 shows that the flood risk prediction for each pothole does not change, or changes are not sensitive enough to be shown in this model, like our assessment of Hen. Physically, this represents the same extent of frequency of flooding expected under the same weather patterns compared to a completely farmed system. This can be seen positively or negatively by landowners, but regardless, because flood risk is not increasing, this scenario shows that a landowner could confidently retire any of the potholes without being concerned with having to remove extra land out of production where profit margins are more desirable.Assessing intentional conservationThe LD‐R scenario considers a scenario in which these potholes all had drainage installed but are now being retired from production and have surface inlets removed. This might result from more intentional conservation efforts, where surface inlets, if present, are removed to reduce pollutant export (N and P) and grasses are planted to promote wildlife habitat and biodiversity benefits. Here, we observe that risk for Bunny and Walnut increased compared to their current conditions. This is due to the reduced drainage capacity and is an important consideration for landowners. In this case, overestimating the retired land area may be necessary to reduce the risk of crop drown out in the adjacent field, which may become marginally more frequent at the boundary of the pothole.Assessing the inherent risk of a potholeLandowners deal with multiple prairie potholes, and many are managed differently despite their proximity and similar field management. If landowners or conservation specialists (e.g., a watershed coordinator) are assessing multiple potholes, they might compare multiple potholes under the same conditions. Comparing potholes in the NP scenario elucidates potential outcomes. The flood risk in this scenario is considered the inherent risk, reflecting the historical natural drainage and land use conditions of a pothole. In natural conditions, where the entire pothole microwatershed is treated as perennial vegetation, we observe that Mouth has the highest risk while Walnut has the lowest. If one of these potholes were slated for retirement, Mouth might be objectively chosen because it has higher risk morphological characteristics. These characteristics might include its large catchment area to pothole ratio (largest of those studied) or its low overflow depth which leads to frequent inundation of the entire pothole. Anecdotally, reality corroborates this interpretation because Mouth was the only pothole in our study that has been retired. Conversely, Walnut is the better candidate for investment in drainage because its inherent risk is smaller.Assessing incremental gainsNo till assesses the removal of tillage practices from the current management. We observe no change in risk for Bunny. However, Walnut's risk decreased by 0.1, while Mouth and Hen decreased by 0.4. We see modified tillage in highly drained potholes results in minimal to no change while these same activities might have a larger effect for field conditions that are higher risk. This result suggests that the drainage infrastructure we see in potholes can only handle so much natural variation in climate and flood risk will remain as long as drainage remains within the range we have observed and incorporated within the RF model. Potholes will likely become inundated intermittently in wet years, affecting crop yields.Context from modeling dataInquisitive users are provided a table that interprets risk level in terms of watershed simulation data. The predicted risk in our case study ranged from 0.6 to 5.6. Because this scale is multiplicative, anything above a risk of 5 is largely indistinguishable based on the training dataset. These high‐risk conditions flood to 30% of their areal extent almost every year and to 60% of their areal extents in 65–80% of the years. Conversely, low risk conditions flood to 30% of their aerial extent around 80% of years, whereas they reach 60% of their area only 40–50% of the years. Expected average annual days of inundation are 3–12 d for extremely low risk potholes and 16–58 d for extremely high‐risk potholes, but these values can extend to over 100 d (Nahkala et al., 2021). The risk data and modeling study data (described in this section) are provided without management recommendations, allowing users to incorporate their own context for making decisions.Limitations and future developmentFuture development may focus on the integration of raster processing within the web map that automatically calculates watershed parameters from digital elevation models to improve the accuracy of user inputs. Additionally, using larger training datasets with the RF model will ensure that the model captures more variation in prairie pothole characteristics observed across the DML. The model currently captures only a small range of observed land management conditions, and calibrated watershed models (Nahkala et al., 2021), which are what the RF model was trained against, were only available for conventional farming operations. The range of morphologic inputs is limited based on the physical proximity of the potholes used to establish the model, and the known range of these topographic characteristics is larger than included in the model training. Further models could similarly include variation in annual precipitation or climate change projections, as both the geographic and temporal variation may influence current and future flood risk. Additionally, the tool could be paired with a profitability analysis so that users could consider costs and benefits of management options given an expected impact (or lack of impact) on crop yield. Finally, the tool could also be expanded to include an assessment of downstream effects. The focus on an individual pothole ignores the hydrologic connection to downstream waters, and thus increasing the drainage may be impossible or ineffective in some cases if capacity does not exist downstream and may increase water yield and flooding or downstream in other cases. In its current form, the tool is primarily for informational and educational purposes, and a disclaimer to this effect is included with the tool.Software availabilityThe PPMST can be accessed via: https://bnahkala.shinyapps.io/ppmst/. The user's manual is embedded on the PPMST homepage. R code is hosted on GitHub: https://github.com/bnahkala/ppmstACKNOWLEDGMENTSThis work was supported by the US Environmental Protection Agency assistance agreement CD97753901. It has not been subjected to the Agency's review and might not necessarily reflect the views of the Agency. This work was further supported by the United States Department of Agriculture Hatch Projects IOW04414 and IOW05474 and the National Institute for Food and Agriculture (2018‐67019‐27886).AUTHOR CONTRIBUTIONSBrady A. Nahkala: Conceptualization; Data curation; Formal analysis; Methodology; Software; Writing – original draft; Writing – review & editing. Amy L. Kaleita: Conceptualization; Funding acquisition; Resources; Supervision; Writing – review & editing. Michelle L. Soupir: Conceptualization; Funding acquisition; Resources; Supervision; Writing – review & editing. Andy VanLoocke: Conceptualization; Funding acquisition; Writing – review & editing.CONFLICT OF INTERESTThe authors declare no conflict of interest.REFERENCESBishop, R. A., Joens, J., & Zohrer, J. (1998). Iowa's wetlands, present and future with a focus on prairie potholes. Journal of the Iowa Academy of Science, 105(3), 89–93.Cerf, M., Jeuffroy, M. H., Prost, L., & Meynard, J. M. (2012). Participatory design of agricultural decision support tools: Taking account of the use situations. Agronomy for Sustainable Development, 32(4), 899–910. https://doi.org/10.1007/s13593‐012‐0091‐zChang, W., Cheng, J., Allaire, J., Xie, Y., & McPeherson, J. (2020). shiny: Web application framework for R (R package version 1.5.0). The Comprehensive R Archive Network. https://CRAN.R‐project.org/package=shinyCox, P. G. (1996). Some issues in the design of agricultural decision support systems. Agricultural Systems, 52(2–3), 355–381. https://doi.org/10.1016/0308‐521X(96)00063‐7Evenson, G. R., Golden, H. E., Lane, C. R., & D'Amico, E. (2016). An improved representation of geographically isolated wetlands in a watershed‐scale hydrologic model. Hydrological Processes, 30(22), 4168–4184. https://doi.org/10.1002/hyp.10930Fey, S., Kyveryga, P., Connor, J., Kiel, A., & Muth, D. (2016). Within‐field profitability assessment: Impact of weather, field management and soils. In Proceedings of the 13th International Conference of Precision Agriculture. International Society of Precision Agriculture.Green, D. I. S., McDeid, S. M., & Crumpton, W. G. (2019). Runoff storage potential of drained upland depressions on the Des Moines Lobe of Iowa. Journal of the American Water Resources Association, 55(3), 543–558. https://doi.org/10.1111/1752‐1688.12738Lane, C. R., Leibowitz, S. G., Autrey, B. C., LeDuc, S. D., & Alexander, L. C. (2018). Hydrological, physical, and chemical functions and connectivity of non‐floodplain wetlands to downstream waters: A review. Journal of the American Water Resources Association, 54(2), 346–371. https://doi.org/10.1111/1752‐1688.12633Liu, K., Harrison, M. T., Shabala, S., Meinke, H., Ahmend, I., Zhang, Y., Tian, X., & Zhou, M. (2020). The state of the art in modelling waterlogging impacts on plants: What do we know and what do we need to know. Earth's Future, 8, https://doi.org/10.1029/2020ef001801Martin, A. R., Kaleita, A. L., & Soupir, M. L. (2019). Inundation patterns of farmed pothole depressions with varying subsurface drainage. Transactions of the ASABE, 62, 1579–1590. https://doi.org/10.13031/trans.13435McDeid, S. M., Green, D. I. S., & Crumpton, W. G. (2019). Morphology of drained upland depressions on the Des Moines Lobe of Iowa. Wetlands, 39, 587–600. https://doi.org/10.1007/s13157‐018‐1108‐4McKenna, O. P., Mushet, D. M., Scherff, E. J., McKean, K. I., & Mills, C. T. (2018). The Pothole Hydrology‐Linked Systems Simulator (PHyLiSS): Development and application of a systems model for prairie‐pothole wetlands (USGS Report 2018‐1165). USDOI, USGS.McKenna, O. P., Osorio, J. M., Behrman, K. D., Doro, L., & Mushet, D. M. (2020). Development of a novel framework for modeling field‐scale conservation effects of depressional wetlands in agricultural landscapes. Journal of Soil and Water Conservation, 75(6), 695–703. https://doi.org/10.2489/jswc.2020.00096Miller, B. A., Crumpton, W. G., & van der Valk, A. G. (2009). Spatial distribution of historical wetland classes on the Des Moines lobe, Iowa. Wetlands, 29, 1146–1152. https://doi.org/10.1672/08‐158.1Muth, D. (2014). Profitability versus environmental performance: Are they competing? Journal of Soil and Water Conservation, 69, 203A–206A. https://doi.org/10.2489/jswc.69.6.203ANahkala, B. A., Kaleita, A. L., & Soupir, M. L. (2021). Characterization of prairie pothole inundation using AnnAGNPS under varying management and drainage scenarios. Agricultural Water Management, 255, 107002. https://doi.org/10.1016/j.agwat.2021.107002Nahkala, B. A., Kaleita, A. L., & Soupir, M. L. (2022). Empirical tool development for prairie pothole management using AnnAGNPS and random forest. Environ. Modelling & Software, 147, 105241. https://doi.org/10.1016/j.envsoft.2021.105241Ranjan, P., Duriancik, L. F., Moriasi, D. N., Carlson, D., Anderson, K., & Prokopy, L. S. (2020). Understanding the use of decision support tools by conservation professionals and their education and training needs: An application of the Reasoned Action Approach. Journal of Soil and Water Conservation, 75, 387–399. https://doi.org/10.2489/jswc.75.3.387Rose, D. C., Sutherland, W. J., Parker, C., Lobley, M., Winter, M., Morris, C., Twining, S., Ffoulkes, C., Amano, T., & Dicks, L. V. (2016). Decision support tools for agriculture: Towards effective design and delivery. Agricultural Systems, 149, 165–174. https://doi.org/10.1016/j.agsy.2016.09.009Schilling, K. E., & Dinsmore, S. (2018). Monitoring the wildlife, hydrology and water quality of drained wetlands of the Des Moines Lobe, northern Iowa: Introduction to special feature. Wetlands, 38(2), 207–210. https://doi.org/10.1007/s13157‐017‐0989‐ySchilling, K. E., Then, S. R., & Ikenberry, C. D. (2019). Water balance modeling of temporary ponding in a drained prairie pothole wetland. Environmental Modeling and Assessment, 24, 37–48. https://doi.org/10.1007/s10666‐018‐9596‐4Upadhyay, P., Pruski, L. O. S., Kaleita, A. L., & Soupir, M. L. (2019). Effects of land management on inundation of prairie pothole wetlands in the Des Moines Lobe using AnnAGNPS. Agricultural Water Management, 213, 947–956. https://doi.org/10.1016/j.agwat.2018.12.016USDA. (2020). Prairie Pothole Water Quality and Wildlife Program. USDA. https://www.nrcs.usda.gov/wps/portal/nrcs/ia/programs/financial/eqip/a5eaadd4‐eb4b‐4cd9‐b5c9‐11e1b380cb45/

Journal

"Agrosystems, Geosciences & Environment"Wiley

Published: Jan 1, 2022

References