Assessing the Potential Flood Mitigation Services of a Former Golf Course with a Focus on Flood Storage

Abstract

Golf course closures in the U.S. have exceeded openings since 2006, creating an opportunity for reuse that maximizes ecosystem service. Flood storage potential on a former course in Ohio was evaluated as part of a planning process for its future use. Flow through culverts that historically provided an outlet for excess surface water to drain to an adjacent stream was reversed, reconnecting the stream to its floodplain. Installation of shallow monitoring wells, surface water and groundwater level monitoring, and mapping of flood events provided the data necessary to assess flooding and flood storage potential. This study illustrates two methods for estimating flood storage, a culvert flow model based on head differences between the stream and ponding on the former course, and a GIS flood volume model based on high watermarks. As much as 103 K cubic meters and as high as 2.1 percent of stormflow was stored on the former course on the surface for a given flood event, with additional storage in the subsurface. Ecosystem services associated with stormflow are also provided, including water purification, soil formation, and nutrient cycling.

1. Introduction

Annual course closures in the U.S. have exceeded new course openings since 2006. On average, 0.8 percent of the total supply in 18-hole equivalents has closed over the last 30 years, increasing to as high as 2.1 percent immediately prior to COVID-19 [1]. Closed golf courses are mostly repurposed for residential and commercial real estate but some are converted for agricultural or recreational uses. Because of the large areas of open space they comprise, closed golf courses provide an unmatched opportunity to maximize the ecosystem services they provide [2]. In a national study of golf course closures across the United States, Cederberg [2] found that 42 percent of the closed courses examined had no clear plans for future use. Of the courses that had been converted to parks and open spaces, the results indicated that large areas of former courses were converted to open spaces or natural areas with services limited to trail walking, nature viewing, and outdoor education [2].

Since Petrosillo et al. [3]. make the case that urban planners consider early in the planning phase the ecosystem services that new golf courses may provide, it is equally, if not more, important in planning for course closures. Particularly with respect to publicly owned courses facing closure, leaders need to consider whether a closed golf course can support critical ecosystem services beyond cultural services associated with health, well-being, and education. Golf courses account for as much as 29 percent of all urban greenspace and are the most accessible form of greenspace for approximately 3.4 percent of the U.S. urban population [4]. Course closures, by virtue of their extent alone, provide a unique opportunity to integrate those services with others that more directly address current and future challenges that face a community (e.g., water regulation in the case of communities with combined sewer overflows or climate amelioration in urban heat islands).

In the upper Great and Little Miami River watersheds, comprising an area of approximately 8000 km², of the 53 golf courses identified, nine have closed since 2012. One of these courses was converted for residential use and another for agricultural use, but the remaining seven are green spaces, mostly fallow since the time of closure. Only one of them (the former Larch Tree Golf Course west of Dayton, Ohio) was assessed for its conservation value and converted into a wetland mitigation bank. In addition to trails, wildlife, and educational opportunities, the former course provides wetland habitat and improves water quality, regulates runoff, and reduces erosion in addition to the host of other regulating and supporting services. Another one of the closed courses, the former Snyder Park Golf Course (SPGS), an 18-hole municipal course in Springfield, OH, is the subject of this study. It was closed in 2014 without a post-closure plan.

The goal of this chapter is to provide an example of how a select ecosystem service can be assessed to maximize the sustainable reuse of a golf course. The former golf course could provide a number of ecosystem services, including provisioning services (food, fresh water, fiber, and fuel), regulating services (climate regulation, water regulation, erosion control, water purification, and pollination), cultural services (education, recreation, and tourism), and supporting services (soil formation, nutrient cycling, and primary production) [5]. The focus of this study is on the role the former golf course plays in water regulation. Two different processes are associated with water regulation, flood prevention, associated with runoff generation occurring at the watershed scale, and flood mitigation, associated with storage of water once it has accumulated in the stream [6]. In cases of prevention, the need for vegetation biomass and forest to prevent runoff is critical, whereas available storage space for stormflow along the stream corridor is critical for mitigation [6]. In this study, I examine the stormflow conditions under which the former course floods, the surface water-groundwater interaction during flooding, and the volume of flood storage it provides.

2. Study area

Snyder Park Golf Course opened in 1920 as an 18-hole public golf course operated by the City of Springfield and National Trail Parks and Recreation District. The original course was constructed on the floodplain at the confluence of Mad River and Buck Creek, utilizing in its design extant floodplain and channel features that were preserved following channelization and levee construction on Buck Creek. Figure 1 illustrates the topography of the former course. It is higher to the north and generally slopes downward toward the confluence at a gradient similar to the current streams, 0.175 percent or 1.75 m/km. Former channels of Buck Creek are especially evident in Figure 1, recognized by their narrow sinuous features. Following closure in 2014, the higher elevations were developed into community gardens, but the lower elevations, which experienced periodic flooding (Figure 2), were left fallow. Currently, approximately 40 acres to the north and northeast are mowed to maintain a park-like setting, with a portion of it developed into test gardens for landscape plants and an arboretum. Walking trails are maintained at lower elevations, but because of the flooding, it is otherwise rewilding with a mix of native and invasive vegetation and noxious weeds.

Figure 1.

DEM of the former Snyder Park golf course in Springfield, Ohio illustrating the floodplain and channel topography on which it was built. The location of monitoring wells and water level loggers used in this study are shown in relation to key features. Following closure in 2014, the higher elevations (brown area to the northeast) were developed into community gardens, but the lower elevations (green and light blue area to the southwest), with periodic flooding, were left fallow. The relatively low, narrow sinuous features are former channel positions of Buck Creek.

Figure 2.

Flooding in March 2020 covered approximately 18 acres of the former course. Buck Creek is separated from the course by a levee.

The former course is underlain primarily by glacial sand and gravel outwash and more recent stream alluvium. Although mapped as Westland silty clay loam [7], its field soil description suggests it is more likely a Tremont silt loam (Figure 3). The Tremont silt loam exhibits gleyed colors with reddening around concentrations of organic matter indicative of occasional flooding with a seasonally high-water table. It is rated as a hydric soil. A narrow ridge that runs the length of the southeastern border of the course (Figure 1) is underlain by glacial sand and gravel and soil profile development is indicative of an Eldean silt loam. It offers a striking contrast in soil profile development to that of the Tremont silt loam (Figure 3).

Figure 3.

Two soils are present on the former SPGC, Eldean (left) and Tremont (right) soil series. The Tremont soil series is rated as a hydric soil in Ohio and comprises more than 98 percent of the former course. The sand and gravel that form the parent material for the Eldean underlie the entire area.

The decision to focus on water regulation and, specifically, flooding was based on the topographic setting, the extent of hydric soils, and early observations of flooding. When the golf course was in operation, excess water was drained to Mad River through a network of ponds connected by a tile network and two side-by-side, culverts that were installed to drain the golf course to Mad River (Figure 4). The culverts are 82.6 cm (30 in) in diameter, approximately 57.0 m (187.0 ft) in length, with a fall of 0.35 m (13.8 in) sloping toward Mad River (Figure 4) with a fall of 0.35 m (13.8 in) over that length. Closed flap gates on the Mad River side of the culverts prohibited back-flooding from the Mad River. Following closure, the flap gates were inactivated, in effect hydrologically reconnecting the river to its floodplain under stormflow conditions. The ponds and tile network have not been modified so flow is currently governed by relative elevation of water on Mad River and the former course. The Mad River watershed at this point is 337 sq. km. When Mad River exceeds 50 cm (1780 cfs) at the St. Paris Pike at Eagle City, OH stream gage (USGS gage 03267900) just upstream from the culverts, the stage is sufficient for flow onto the course. Once water floods the course, it is distributed through a series of ponds and water hazards that are connected by a network of culverts at cart crossings and tiles, particularly one connecting pond with a central water hazard (Figure 1).

Figure 4.

Illustration of the relation of the culvert between Mad River and the former course showing a stormflow event on Mad River at a higher stage than surface water on the course. The elevations shown are surface water elevation on the former course at peak stormflow elevation on Mad River according to the hydrographs in Figure 5.

3. Methodology

From 2014 to the present, surface water and groundwater elevations in various areas of the course have been surveyed or measured during and after floods (Figure 1). Water-level loggers were installed in various configurations to examine (1) groundwater elevations on the former course and its fluctuation in response to streamflow and flooding, and (2) surface water elevations and head differences between Mad River and the former course during flood events (Figure 1). At any given time, between two and four Solinst Leveloggers were deployed during the course of the study at various locations to measure surface water and groundwater water depths at 15- to 60-min intervals. Depths were compensated for barometric pressure and converted to water-level elevations using surveyed water levels during the period of measurement. Shallow monitoring wells were hand-augured to depths of approximately 2 m. While some were equipped with water-level loggers, depth to water was periodically measured in others using a steel-measuring tape and converted to water-level elevations. Open bodies of water (i.e., water hazards on the former course) were considered to represent groundwater level because all were excavated in sand and gravel outwash.

During the initial stages of this study, I realized flooding from stormflow events on Mad River was more frequent than anticipated, presenting an opportunity to study flood storage potential. Immediately following many of these flood events, the elevations of high-water marks, indicated by the highest accumulations of organic debris or surface ice, were surveyed. Flood storage potential was estimated by two means, modeling culvert flow based on the head difference between the Mad River and surface water elevation on the course and calculating the volume difference between the golf course surface and high-water elevation of a given flood.

Under flood conditions on Mad River, the culverts connecting Mad River with the former Snyder Park Golf Course exhibit full barrel flow [8]. In this case, the culverts are in pressure flow throughout their length, driven by the positive head between the river and the golf course. For the floods modeled in this study, datalogger data confirm that the inlet and outlet are fully submerged. This condition is often assumed in calculations [8] and is the basis for the nomograph estimations of discharge presented here. They also note that the equations and nomographs for submerged, full flow can be reasonably applied to culverts with no slope or an adverse slope. The culvert in this study has an adverse slope (Figure 4). Culvert discharge was calculated at a 15-min interval using a spreadsheet solution to the flow equation in [8] based on the head difference (Figure 5), an entrance loss coefficient (Ke), and culvert length and diameter. A value of 0.9 for Ke, representing a corrugated metal pipe projecting on either side, was used.

Figure 5.

To examine flood storage potential, water level was logged at 15-min intervals on either end of the culvert and converted to elevation. The difference in elevation, or head, was used in calculating culvert flow.

Flood storage was also estimated using the Cut Fill tool in ArcGIS ArcMap 10.8.2. The Cut Fill tool is used to calculate a volume difference between two input raster surfaces at two different time periods, prior to flooding and at peak flood. The surface prior to flooding is the bare earth digital elevation model (DEM) based on LiDAR. The surface at peak flood is created for each flood based on high-water marks or peak surface water elevation. Flood surfaces for two events are shown in Figure 6. Peak flood elevation is assumed flat for calculation purposes, but there is a negligible slope from the culverts to distant points of flooding.

Figure 6.

The lower elevations of the former course frequently flood. Two floods are depicted here from the sequence of floods used in the study to estimate flood storage. The November 2, 2018 and may 17, 2019 floods inundated 13 and 17 acres, respectively.

4. Results

4.1 Groundwater fluctuations

Depth to groundwater varied seasonally and as a result of floods, from 0 m during floods to greater than 2 m during extended dry periods in late summer and fall. Figure 7 shows variation in water level along a central transect of monitoring wells (Figure 1) at representative times during the study. Three surveys are illustrated. The May 27, 2022 survey occurred during a wet period, concurrent with a minor stormflow event (peak discharge 29 m³s⁻¹) on Mad River. The water table is fairly flat at an elevation of 272.3 m (Figure 7). The Jun 28, 2022 survey occurred 15 days after a significant flood event but with trace rainfall since then. Groundwater lies below the central water hazard (Figure 7) and is falling. The Jul 13, 2022 survey occurred 7 days after a stormflow event (peak discharge 131 m³s⁻¹) on Mad River sufficient to flood lower elevations of the former course. Surface water in the central water hazard is still elevated as it is the pond along the west side of the course (at 0 m on the transect) with standing water along the western side of the transect. Groundwater elevations on either side of the central hazard are lower (Figure 7).

Figure 7.

(a) Surface water and groundwater variation for selected dates along the central transect of monitoring wells shown in Figure 1. Depending on the magnitude of flooding, stormflow moves overland or through a tile connecting ponds on the western edge of the former course to the central water hazard. (b) Groundwater hydrographs for two floods, June 2022 (day one is 06-03-2022) and March 2023 (day 1 is 02-22-2023) from a monitoring well equipped with a groundwater level logger east of the central water hazard (Figure 1).

The groundwater response to two flood events, shown as a range of elevations, is also shown in the cross section in Figure 7 for the monitoring well along the eastern edge of the transect. The respective groundwater hydrographs for stormflow events on Mad River are also shown (Figure 7). The base groundwater elevations prior to the flood events were similar (272. 1 m). The Jun 13, 2022, stormflow event, which peaked on Mad River at 118 m³s⁻¹, is associated with a peak groundwater elevation of 272.9 m (Figure 7). The Mar 3, 2023, stormflow event, which peaked on Mad River at 173 m³s⁻¹, is associated with a peak groundwater elevation of 273.6 m. High-water marks for this event surveyed on the course ranged from 273.33 m to 273.49 m; the average high-water mark, 273.42 m, is shown on the transect (Figure 7). The entire transect was flooded during this event. It is important to note that the average high-water mark is approximately 15 cm below the maximum peak recorded by the water level logger (Figure 7).

4.2 Flood storage on the former course

From November 2018 to June 2019, 20 streamflow events occurred on Mad River upstream of the former course with peak discharges in excess of 50 m³s⁻¹, the approximate flow magnitude necessary for Mad River to backflow through the culverts connecting Mad River to the former course. Seven events, two of which exceeded floods with a return period of more than 5 years, were selected for the study of flood storage potential. Seven stormflow events on Mad River (circled, Figure 8) were selected for the study of flood storage potential on the former course. Peak discharge and total stormflow discharge for these events are summarized in Table 1. The total stormflow of these events accounts for 1845–13,219 K m³ (Table 1). Of that, it is estimated that 37–103 K m³, or 0.4–2.1 percent of total stormflow, was stored on the former course based on modeling of culvert flow (Table 1). The volume of storage tends to increase with the peak discharge. GIS estimates based on the difference between the elevation of the course and a flood surface representing the elevation of the flood are similar: 29–103 K m³, or 0.3–1.6 percent of total stormflow (Table 1). To put this in a different perspective, the stormflow volumes reported would amount to the equivalent of storing between 5 and 8 minutes of peak streamflow discharge for the events studied.

Figure 8.

From November 2018 to June 2019, 20 stormflow events occurred on Mad River at St. Paris pike at Eagle City, OH stream gage (USGS gage 03267900) with peak discharges in excess of 50 m³s⁻¹, the approximate flow magnitude necessary for the Mad River to flood onto the former course. Flood storage was estimated for the seven events circled events.

			Culvert-based Volume		GIS-based Volume
Date(s) of Stormflow Event	Peak Discharge^1 (m^3/s)	Total Stormflow Discharge^2 (m^3)	Stormflow Volume^3 (m^3)	% Stormflow Stored	Stormflow Volume^4 (m^3)	% Stormflow Stored
Nov 1–2, 2018	130.5	9,782,216	59,303	0.6	48,675 (49,843)	0.5
Dec 31, 2018 -Jan 1, 2019	126.0	10,272,094	37,993	0.4	34,245	0.3
Jan 23–24, 2019	111.3	1845,330	38,399	2.1	29,406	1.6
Feb 6–7, 2018	104.8	4,304,795	37,110	0.9	31,651	0.7
Feb 7–8, 2019	206.7	13,219,644	93,245	0.7	103,442 (103,537)	0.8
Feb 12–13, 2019	103.6	8,709,809	42,191	0.5	43,776	0.5
May 17, 2019	205.0	13,057,374	103,445	0.8	103,443 (103,623)	0.8

Table 1.

Flood storage on the former Snyder Park golf course for seven flood events between November 2018 and July 2019.

5. Discussion

With golf course closures exceeding openings in the recent past and their location and extent in predominantly urban and peripheral-urban areas, evaluating the potential reuse of golf courses for water regulation or other ecosystem services is critical. Water regulation services are considered from two perspectives, flood prevention and flood mitigation [6]. The factors most often associated with flood prevention, increased biomass, proportion of permeable surface area, infiltration capacity, and soil quality [6], also characterize golf courses. Former courses in upland urban areas are particularly well-suited to serve this function. However, it is not unusual for golf courses to be developed on “unbuildable” land, even land that is or was formerly floodplain (e.g., see the discussion thread in [9], a site dedicated to golf course architecture). Reconnecting streams to former floodplains are widely recognized as a sustainable means of flood mitigation with the added benefits of increasing floodplain goods and services and resiliency to climate change impacts [10]. The former Snyder Park Golf Course, situated on a former floodplain at the confluence of two streams, was reconnected to Mad River utilizing the existing side-by-side culverts that had historically drained excessive standing water from the operating golf course.

Flood storage was estimated by two different methods with remarkably similar results (Table 1). During stormflow events of sufficient stage-discharge, stormflow stored on the course amounts to as much as 2.1 percent of the total stormflow volume on Mad River (Table 1). Results from either method of estimation are conservative as both methods rely on the stage elevation of water on the former course and do not account for infiltration as the flood wave inundates the course (i.e., the stage elevation of water does not include these concurrent losses) as it will be discussed later. There is also loss through the culverts as the flood elevation exceeds the stage elevation of Mad River as the flooding wanes. The largest stormflow event for which flood storage was estimated occurred on May 2019, amounting to 103, 455 cu m of flood storage (Table 1). With a high-water mark of 273.7 m, the May 2019 flood inundated more than 39.6 acres of the former course, about 91 percent of that part of the course that is not mowed or in gardens. The maximum flood storage, without impacting current maintained areas, with a high-water mark of 274.0 m, would amount to 162, 437 cu m of stormwater. The former course could receive additional stormflow, but it is ultimately limited by the volume of water able to move through the culverts.

Flood storage is likely underestimated as noted previously. Water-level data from the monitoring wells indicate that groundwater and surface water are hydraulically connected during flooding events, with surface flooding causing rapid changes in groundwater levels. By implication, the former course offers both surface and subsurface flood storage, the latter not included in the flood storage estimations. Groundwater levels rise in wells nearest Buck Creek and Mad River even during stormflow events that do not exceed the threshold necessary for surface flooding of the course, similar to bank storage [11] with some of the same benefits for flood mitigation. The two groundwater hydrographs in Figure 7 illustrate this for events that flooded the former course. Elevation of the flood surface during June 2022 was sufficient to push water, through the tile drain, to the central water hazard. Seepage into the banks of the water hazard produced a 0.78-m rise in groundwater elevation in a monitoring well approximately 16 m further east of the central water hazard. For the May 2023 flood event, the central water hazard and well location were both inundated (Figure 7). The monitoring well was submerged by approximately 0.5 m of water. Though it is not possible to know whether the subsurface at the location of the monitoring well was recharged by lateral flow from the central water hazard or vertically from rapid surface flooding, it is clear that flood storage at this point on the course included approximately 1.1 m of groundwater storage and 0.5 m of surface water storage (Figure 7). The hydraulic connection between surface water and groundwater on the former course is facilitated by the sand and gravel outwash deposit which underlies the entire course. Diminishing groundwater levels in the days and weeks following stormflow events and flooding suggest that groundwater beneath the former course is gradually supplying baseflow on Buck Creek and Mad River.

Flood storage on the former course is fairly frequent. Streamflow discharge in excess of the 50 m³s⁻¹ needed to overflow onto the course occurs with an annual exceedance probability of less than 50 percent and an average recurrence interval of less than 2 years [12]. Ecosystem services directly or indirectly related to flooding and flood storage are also potentially (and frequently) provided by the former golf course. Several ecosystem services were observed or inferred, but not measured, here. Given the nature of stormflow in Midwestern U.S. watersheds dominated by agricultural uses (e.g., excessive sediment and nutrient loads), the former course provides supporting services like soil formation and nutrient cycling. Sandy deposits nearest the culverts and cores from ponds nearer the culverts showing fine sand, silt, and clay, and fine silt recently deposited on reduced muck and sand and gravel, and a thin layer of clay coating vegetation on flooded surfaces all indicate sediment accumulation. Carbon storage on the reconnected floodplain provides climate regulation, especially as the formerly hydric soils are more routinely saturated due to flooding. The increased residence time associated with flooding and the infiltration of stormflow would result in increased water quality, which is returned to Buck Creek and Mad River as baseflow. A conceptual plan for the park district managing restoration of the former course is based on these services in its early stages. It includes excavating a waterway near the culverts to promote flow onto the course, removing barriers between, and enlarging, ponds on the western edge to increase storage, and lowering the topography in and adjacent to the central water hazard to restore it as a wetland. In addition to walking trails and wildlife observation decks, different land covers include woodland, prairie, wet meadow, and pond.

6. Conclusions

Golf course closures exceed new course openings in the U.S. and many comprise large tracts of green space in urban areas without clear plans for future use. Course closures in urban areas and the periphery provide an opportunity for maximizing ecosystem services beyond typical cultural services (i.e., trail walking, nature viewing, and outdoor education), increasing community resilience. The focus of this study is a closed golf course on a former floodplain at the confluence of two streams, Buck Creek and Mad River, the latter of which is connected to the course through two culverts that historically drained excess water from the course. This study reports the results of two methods for estimating flood storage, a culvert flow model based on head differences between the stream and ponding on the former course and a GIS flood volume model based on high watermarks. Streamflow on Mad River in excess of 50 m³s⁻¹, a relatively minor flood with an exceedance probability of greater than 50 percent, is sufficient to flow onto the former course. For the stormflow events modeled, the former course stores 0.3–2.1 percent of total stormflow with good agreement between the two methods, but they likely underestimate actual storage. Concurrent with surface flooding, stormflow is being infiltrated causing groundwater levels to rise in flooded areas as well as in areas without surface flooding.

Acknowledgments

Students in my watershed hydrology course helped to install monitoring wells during preliminary investigations. Andrew Francis and Christian Smith maintained, collected, and analyzed water-level data from them as well as helped with water level surveys.