Annual Water Budgets for a Seasonally Inundated Sinkhole Wetland
- First Online:
- Cite this article as:
- Hill, A.J. & Neary, V.S. Wetlands (2012) 32: 963. doi:10.1007/s13157-012-0331-7
- 245 Views
Annual water budgets spanning 2 years, 2004 and 2005, are constructed for a sinkhole wetland in the Tennessee Highland Rim following conversion of 13 % of the watershed area to impervious surfaces. Surface runoff was the dominant input, with a contribution of 56.4 % of the total. An average of 18.9 % of gross precipitation was intercepted by the canopy and evaporated. Deep recharge varied from 55.5 % (2004) to 52.2 % (2005) of total outflow. Evapotranspiration accounted for 46.2 % of the total losses, with an average of 50.3 % lost from soil profile storage. The annual water budgets indicate that deep recharge is a significant hydrologic function performed by isolated sinkhole wetlands, or karst pans, on the Tennessee Highland Rim. Continued hydrologic monitoring of sinkhole wetlands are needed to evaluate hydrologic function and response to anthropogenic impacts. The regression technique developed to estimate surface runoff entering the wetland is shown to provide reasonable annual runoff estimates, but further testing is needed.
KeywordsWetland hydrologyEvapotranspirationWater budget
In karst landscapes, wetlands commonly form within topographic depressions created from the collapse of the underlying limestone bedrock and are referred to as sinkhole wetlands. Sinkhole wetlands are often geographically isolated (i.e., completely surrounded by upland, Tiner 2003) and fall into the depressional hydrogeomorphic wetland class (Brinson 1993; Hill et al. 2006). Other classes of sinkhole wetlands are distributed throughout the United States, particularly the southeastern portion (Tiner 2003). Sinkhole wetlands, like other depressional wetlands, perform many important functions including surface-water storage/flood-water protection, nutrient transformation and cycling/water-quality maintenance, aquatic productivity, and wildlife habitat (Tiner 2003).
Several studies have focused on the hydrology of wetlands located in karst settings (Hendricks and Goodwin 1952; Greear 1967; Wolfe 1996; O’Driscoll and Parizek 2003; Hill 2007; O’Driscoll and Parizek 2008, and Moran et al. 2009). A comprehensive treatment of hydrologic transfer processes and water budgeting specific to karst terrains is given by White (1988). These studies reveal great variety and complexity of hydrologic function in karst terrains. For example, O’Driscoll and Parizek (2008) examined the influence of geologic substrate on pool hydroperiod for a chain of wetlands in an Appalachian karst valley in central Pennsylvania, USA. Unexpectedly, they found hydroperiod to be longer for sandy pools compared to clay pools. O’Driscoll and Parizek (2003) evaluated groundwater hydrology for the same chain of wetlands. They found the hydrologic catchment area of the ponds to be significantly smaller than catchment area based on topography.
Comprehensive investigations in other regions have reported similar findings. For example, the correlation between greater sinkhole depth and better internal drainage observed by Wolfe (1996) has also been reported by Greear (1967) for a similar class of wetlands in northwest Georgia, known as sagponds. However, Hendricks and Goodwin (1952) found no relation between sinkhole depth and internal drainage in their study of limesinks in southwestern Georgia.
The objective of this study is to construct annual water budgets for a sinkhole wetland on the Eastern Highland Rim, presumably a karst pan in the preceding classification (i.e., shallow and lacking visible internal drains), but geographically distant from the study sites of Wolfe (1996).
Study Site and Monitoring Program
The Tennessee Highland Rim is a division of the larger Interior Low Plateau physiographic province (Wolfe 1996). Sinkhole wetlands are common along the central and southern portions of the Eastern Highland Rim and the northwestern portion of the Western Highland Rim. Most of the Tennessee Highland Rim is underlain by Mississippian age limestone with interbedded cherts and shales (U.S. Department of Agriculture 1992). A notable concentration of karst wetlands occurs in The Barrens (Fig. 1), known for its low relief compared to the rest of the Highland Rim (Wolfe 1996).
Most of the Highland Rim is covered by a layer of loess, which probably originated in the western part of the state during the glacial ages, underlain by a layer of clay or cherty clay residuum. Most soils on the Highland Rim contain a fragipan layer near the boundary between the loess mantle and the limestone residuum (U.S. Department of Agriculture 1992).
The wetland selected for this study is located at N36°10′53″ and W85°27′21″ in Algood, TN, near the Eastern Highland Rim escarpment, approximately 161 km west of Knoxville, TN. The location of the site relative to the study sites of Wolfe (1996) is shown in Fig. 1.
Dominant vegetation at the site is red maple (Acer rubrum L.), black gum (Nyssa sylvatica L.), willow oak (Quercus phellos L.), and green ash (Fraxinus pennsylvanica Marsh) in the overstory, red maple and Virginia willow (Itea virginiana L.) in the midstory and shrub layers, and various herbaceous Carex species in the understory. The spatial arrangement of plant community species is closely related to the hydroperiod and individual species tolerance to inundation and saturation.
A detailed soil characterization was conducted at two locations within the wetland. The first location was outside the extent of seasonal flooding and Location B was within the region seasonally inundated with surface water. The soil horizon sequence at both locations was similar. Both contained an A horizon, followed by a series of Bt horizons. A Btx horizon (i.e., fragipan) was encountered at both locations at a depth of approximately 1 m and the thickness ranged from 20 cm to 66 cm. A C horizon followed the fragipan and limestone bedrock was encountered at a depth of approximately 2 m at both locations. The surface horizon (A) had a silt loam texture and the remaining horizons were a silt clay loam.
The fragipan horizon influences site hydrology by restricting downward root penetration, thus decreasing plant transpiration, and promoting saturated soil conditions in the overlying horizons. It was observed during monitoring well installation that fragipan expression and thickness varied significantly throughout the wetland.
Water Budget Equations
A monitoring program was initiated for the Algood wetland to collect information on meteorological conditions, surface and groundwater levels, and the spatial distribution of precipitation. Although some pre-development hydrologic data was collected, hydrologic data reported in this paper represents post-development conditions. Meteorological conditions, including temperature, solar radiation, humidity, atmospheric pressure, and precipitation, were measured with a Davis GroWeather weather station located 55 m southeast of the wetland in an open area (Fig. 2). A Global Water WL15 water level sensor was installed in the deepest part of the depression to measure surface water levels. Both meteorological variables and surface water level were measured at 15-min time intervals.
A total of 15 recording rain gages were installed at 12 stations to measure gross precipitation and throughfall (gross precipitation minus interception) as indicated in Fig. 2. Duplicate gages were installed at three stations for quality control purposes (RGGA/RGGB, RG1A/RG1B, and RG2A/RG2B in Fig. 2). Stations WS1 and RGGA/RGGB were located in open areas separated by a distance of 250 m to measure gross precipitation. The remaining 10 stations were located under the canopy to measure throughfall. Stations were located to provide good site coverage without consideration of canopy characteristics (e.g., cover, plant community, etc.). This resulted in an average separation distance of 40 m for the 10 throughfall gages. Each gage was mounted on a platform 62 cm above the ground surface. Event data loggers were connected to each gage and stored under the platform. Gage funnels, debris screens, and buckets were cleaned routinely throughout the monitoring period.
Groundwater monitoring wells were installed at 11 locations throughout the wetland and adjacent upland to monitor groundwater levels (Fig. 2). Wells were constructed and installed following Sprecher (2000) using 5.1 cm (2 in) PVC pipe. During field installation, the installation depth was dictated by the location of the restrictive soil layer, or fragipan. Online Resource 1 provides a summary of monitoring well properties. Groundwater levels were measured manually using a Heron water level indicator accurate to the nearest 0.25 mm (0.01 in). The sampling interval ranged from 1 day to 1 week.
Daily estimates of throughfall for the Algood wetland were obtained by averaging throughfall measurements at each rain gage station under the mixed forested plant community (10 stations). Interception was computed as the difference between daily gross precipitation and throughfall. Gross precipitation estimates were obtained by averaging measurements at stations WS1 and RGGA/B. Precipitation records were analyzed meticulously to identify erroneous data. Regression relationships between daily gross precipitation Pg and throughfall were developed to fill in missing records early in 2004 before the rain gage network was in operation (Hill 2007). Shrubs and herbaceous ground cover were sparsely distributed in the seasonally flooded portion of the wetland and assumed not to affect interception losses. Stemflow, the portion of intercepted water that flows down tree trunks and plant stems, was also not measured.
Surface Runoff Estimation
The ET demand was first satisfied by available canopy storage then the remaining (ETp−Ec) satisfied by surface water and soil water storage. Potential ET from soil profile storage was reduced using a soil moisture reduction factor that depends on the soil moisture relative to the saturated value (i.e., porosity). Water availability was generally not a limiting factor on ET throughout the 2 year monitoring period, so ET occurred at or near the potential rate. Late in the growing season of 2005, a year of below average precipitation, water availability was limiting and a reduction factor was applied. Hill (2007) used a water budget model to simulate the temporal variability in soil moisture content during this period. The model accurately predicted the return of flooded conditions following the dry period late in the growing season. Since soil moisture was not measured directly at the site, simulated values for the soil moisture reduction factor were used.
Since all terms in Eq. 3 except the net seepage (Gi−Go) were either measured directly (Pn) or estimated from ancillary data (ΔSsw, Ri, ETsw, and Ro), the residual represents the seepage magnitude and the sign represents the direction (+ for net discharge to wetland and – for net recharge from wetland). Winter (1981) cautions that groundwater terms computed as a residual can differ from independent estimates by more than 100 %. Independent estimates of wetland seepage reported by Hill and Neary (2007) are used to evaluate the budgetary residuals computed in this study.
The long-term climatic conditions for the nearest climatological station (National Climatic Data Center, COOP ID 402009) were analyzed. The average annual precipitation for a 43 year record from 1952 to 2005 was 144 cm. Average monthly temperatures over the same period were used to calculate an average potential evapotranspiration of 61.1 cm using the Thornthwaite method (Ponce 1989). The monitoring period (2004–2005) includes 1 year of “below average” annual precipitation and 1 year of “above average” annual precipitation (Hill 2007).
The water table was nearly 0.5 m below the ground surface by the end of May 2004, but returned to a ponded condition following a 2.90 cm rainfall event and again dried completely by mid-June. The rapid decline in wetland stage is a result of increased evapotranspiration and seepage losses. Seepage losses are greater during the growing season due to large transpiration losses near the outer edge of the surface water body. Over a 40 day period early in the growing season, the wetland stage reached the crest elevation and dried completely four times. A ponded condition returned mid-October and persisted until May of 2005. The hydroperiod (days with ponded surface water) during 2004 was 255 days and decreased to 151 days in 2005 due to below average precipitation.
These relationships were used to fill in missing records early in 2004 before the rain gage network was in operation. Equations 10 and 11 were applied to a 47-year record (NCDC COOP station 402009) of daily gross precipitation and resulted in an average annual interception of 17.6 %. Monthly interception and precipitation totals are shown in Online Resource 2 for 20 months in 2004 and 2005. Monthly interception ranged from 8.3 % in the dormant season to 39.3 % during the growing season.
Reference evapotranspiration estimates predicted with the FAO Penman-Monteith equation are shown in Online Resource 3. The expected seasonal distribution is evident, with peak growing season values approaching 5 mm/day and dormant season values generally less than 1 mm/day. The assumed temporal distribution (dashed line) is shown in Online Resource 4 with the calculated values. A comparison of the ET estimates is shown in Online Resource 5. The mean absolute error was 0.04 m.
Summary of water budget for 10 months in 2004 and 2005 with ponded surface water
Seepage + error (m3)
Average seepage rate (mm/day)
No ponded surface water
No ponded surface water
During 2004, a year of above average precipitation, 18.9 % of gross precipitation was intercepted by the canopy and subsequently evaporated. Inputs to the surface water body were dominated by surface runoff (63.3 %). Seepage into the soil profile was the dominant water loss mechanism, accounting for 79.5 % of the total loss from surface water storage. Total ET, consisting of canopy evaporation (Ec) and surface water and soil evapotranspiration (ETsw and ETsoil), accounted for 44.5 % of the total water loss. The remaining loss (55.5 %) is attributed to deep recharge. A total of 17.7 % of the annual ET was canopy evaporation. Surface water ET and ET from soil profile storage accounted for 42.5 % and 39.8 %, respectively.
Canopy interception was 18.8 % during 2005, a year of below average precipitation. Surface runoff contributed 59.5 % of the total input to the surface water body. Seepage into the soil profile accounted for 86.7 % of the total loss from surface water storage. Although 2005 was a year of below average precipitation, ET occurred at or near the maximum (i.e., potential) rate for most of the year. ET demand was satisfied first by water in canopy storage, followed by water in surface water and soil profile storage. Since the hydroperiod was shorter in 2005 (151 days) than 2004 (255 days), ET from soil profile storage (ETsoil) was dominant in 2005. Deep recharge accounted for 52.2 % of the total water loss, with the remaining 47.8 % lost through ET. Due to the short hydroperiod in 2005 (151 days), the majority of ET (60.6 %) was from soil profile storage.
Groundwater and Surface Water Interaction
Monitoring well data are shown in Online Resource 6, which includes the reemergence of flooded conditions mid-September of 2004. Water level elevations in all wells were lower than the surface water elevation, indicating a recharge condition. The second period (Online Resource 7) includes 7 months during 2005. On several occasions in 2005, particularly April, discharge conditions occurred on the south side following precipitation events. During May of 2005, discharge conditions occurred on the north side (well NC) with recharge conditions occurring on the south (well SCS), east (well 4), and west (well W) sides.
Water surface elevations in wells SCS and SCD, which are screened in the top and bottom aquifers (Online Resource 1) at the same location (Fig. 2), differed by as much as 26 cm. The water elevations in the deeper well (SCD) exceeded those in the shallow well (SCS) for many of the readings in Online Resources 6 and 7. The monitoring well in the south pasture (SP) was dry until April 2005 and the water table remained below the upland fragipan horizon throughout the monitoring period. Wells were not installed beyond the forest along the north, east, and west directions (permission denied by property owners), although exploration with a soil probe indicated the absence of a water table above the upland fragipan soil horizon.
The hydraulic gradient out of the wetland varied significantly with stage and direction as indicated in Online Resource 8. At high stages, the hydraulic gradient out of the wetland varied less with direction. The high gradients at low stage are due to increased ET losses from perimeter vegetation.
Discussion and Conclusions
Water budgets are subject to considerable uncertainty, especially when one or more terms are estimated as a budgetary residual (Winter 1981; Favero et al. 2007). The present study employed a network of rain gages much denser than comparable studies reported in the literature. The coefficient of variation of throughfall measurements at the study wetland were reported by Hill (2007) to approach 10 % for larger events (Pg > 2 cm) and increase to a maximum of 89 % for a smaller event (Pg = 0.4 cm). In addition, stemflow and interception by ground cover and scrubs were not measured. Based on a review of the stemflow literature, Johnson and Lehmann (2006) found stemflow, as a percentage of gross precipitation, to range from 0.07 to 22 % for a wide range of species and environmental conditions. A value of 5.6 % was reported for red maple, a dominant tree species at the Algood wetland. The annual totals reported in Fig. 6 for Pg, Pn, and Ec are uncertain, but probably within 5–10 % of the true values.
The seepage estimates reported in Fig. 6 contain the measurement error of all other inflows and outflows to surface water storage. Surface runoff represents the greatest source of uncertainty for the annual water budgets in Fig. 6. The linear regression model shown in Fig. 5 does not account directly for antecedent moisture or rainfall characteristics (duration, intensity, etc.).
Sensitivity of seepage rates to varied surface runoff (Ri) for the monthly water budgets in Table 1. All other variables were held constant while Ri was varied from (Ri +/−25 %) to (Ri +/−75 %)
Average seepage rate (mm/day)
The annual ET estimates reported in Fig. 6 (164 cm for 2004 and 117 cm for 2005) are abnormally high, well above the long-term average potential ET estimate of 61 cm calculated with the Thornthwaite method as detailed by Ponce (1989). Standard ET models (i.e., Penman Monteith equation) assume large expanses of uniform vegetation. The application of these models to wetland environments with contrasting moisture and roughness can lead to erroneous results. Advective transport of energy to the wetland is enhanced by contrasting moisture (oasis effect) and roughness (clothesline effect), leading to increased ET rates. It is well established that ET rates from small wetlands may be increased by oasis and clothesline effects (Allen et al. 1998; Kadlec and Wallace 2009). The ET rates reported by Hill and Neary (2007) were postulated to be enhanced by both effects. Since parameterization of the ET model in the current study relied on these results, the annual estimates reported in Fig. 6 are consistent with the working hypothesis of Hill and Neary (2007) that significant enhancement of ET rates is occurring locally.
A distinguishing feature of the soil profile at the study wetland was a fragipan horizon at a depth of ~1 m from the ground surface throughout the wetland that extends into the adjacent upland. It was observed during soil explorations and well construction that fragipan expression and thickness varies considerably throughout the wetland. Soil morphological features of reduction indicate a perched water table above the fragipan horizon for portions of the year. Following the return of standing surface water in the fall, the water table migrates downward, eventually leading to a fully-saturated, or near-saturated soil profile.
Despite the restrictive nature of the subsurface conditions (i.e., presence of fragipan soil horizon), the annual water budgets indicate that deep recharge is the dominant pathway out of the wetland and accounts for over 50 % of the total output. Although seasonal variation in the recharge function is expected, monitoring well readings and water budget calculations indicate that the site is predominantly recharge (i.e., seepage from surface water body to local groundwater system). Limited lateral seepage of ponded surface water occurred beyond the forest buffer surrounding the wetland. One exception is when the wetland stage is high and the pasture west of the wetland is flooded. As discussed previously, this occurred only seven times and is not considered a dominant outflow pathway. The groundwater recharge function of certain wetland classes has been previously documented, especially in the semi-arid northern prairies (van der Kamp and Hayashi 1998). Millar (1971) evaluated seepage rates from a large number of wetlands of various sizes, and estimates mid-summer seepage rates to be as high as 50 l per day per meter of shoreline. This translates into less than 5 mm/day at the study wetland, well below the growing season seepage rates reported by Hill and Neary (2007). It is also important to note that deep recharge increases as ET from surface water and soil water storage decreases. If the working hypothesis of elevated local ET rates is in error, greater deep recharge to the underlying groundwater system would be expected.
The seepage rates reported herein and elsewhere (Hill and Neary 2007) are quite large, especially considering the subsurface conditions at the site. Water levels in wells screened only below the fragipan horizon were observed to increase rapidly in response to rainfall events (Online Resources 6 and 7). This suggests a rapid pathway to the lower aquifer below the fragipan horizon. Extensive soil cracking was observed during the growing season throughout the monitoring period (Hill 2007), especially late in the growing season. The importance of soil cracking to infiltration and seepage into fine-textured, swelling soils has been demonstrated (e.g., Arnold et al. 2005). The effect of soil cracking on infiltration and seepage is expected to be of greater significance at the study wetland due to (1) the absence of vegetative ground cover over much of the wetland and (2) the multiple rapid drawdowns that are a characteristic feature of the hydrologic regime.
A consequence of both elevated seepage and ET rates is limited lateral migration of the shallow water table above the fragipan soil horizon. Seepage from ponded surface water moves laterally and is removed by either transpiration in the densely vegetated buffer around the perimeter of the wetland or deep seepage below the fragipan horizon . O’Driscoll and Parizek (2003) also found that the groundwater basin was significantly smaller than the catchment area for shallow karst depressions in central Pennsylvania, USA. Maintenance of the forested buffer at the study wetland has been suggested as the most crucial aspect of future land management (Hill and Neary 2009).
An erratic hydrologic regime was observed with multiple cycles of flooding and drawdown throughout the growing season. Hill et al. (2006) simulated a pristine condition for the study wetland and report a long-term average hydroperiod of 212 days. Comparison with the observed hydroperiods reported in this study (255 days for 2004 and 151 days for 2005) indicates that natural variability in climatic conditions has a greater effect on wetland hydrology than the increased surface runoff from the watershed developments. The critical role of climatic conditions in regulating wetland hydrology has been demonstrated previously with modeling efforts for a variety of wetland environments (Poiani et al. 1996, Mansell et al. 2000, Hill et al. 2006, and Sun et al. 2006). This also highlights the potentially devastating impact of global climate change predictions on wetland environments (Brooks 2009).
The regression technique developed to estimate surface runoff entering the wetland has not been tested rigorously. A semi-quantitative evaluation of the annual runoff estimates is possible by examining the sensitivity of the water budget residuals to a change in surface runoff. If the surface runoff depths reported in Fig. 6 are increased by 50–100 %, the resulting water budget residuals representing seepage fall well outside the range of diurnal curve estimates. This indicates the runoff estimates are reasonable, but further testing of this technique is needed. Specifically, under what conditions can event seepage be neglected and how does model bias change with the length of the averaging period?
It is still unknown how atypical the findings at the study wetland are relative to other sinkhole wetlands in the region. Additional hydrologic studies of sinkhole wetlands are needed on the Tennessee Highland Rim for comparison, particularly for sites with minimal watershed development.