4.1. Water Balance Analysis
The 2018 water year was selected for focus because it was the most recent year with a complete record of water level data. Massachusetts generally averages between 1,000 and 1,200 mm of precipitation a year (U.S. Climate Atlas 2019). In 2018, 1,440 mm was recorded, which made this year higher than average in both precipitation and discharge. Air temperature followed typical patterns for the area, and minimum and maximum air temperatures were consistent with most other years. Effects of the inter-annual variability typical of the New England region can be expected to have a similar influence on vernal pool hydroperiod, the annual streamflow hydrograph, and water level fluctuation in Black Gum Swamp.
Precipitation was consistent (on a monthly basis) throughout the water year, with several large events occurring in summer and late fall. As expected, discharge was a subdued reflection of rainfall and snowmelt patterns. The amount of water that passed the stream gage was about half of the annual precipitation total, which, again, is typical in the southern New England region. The climatograph illustrates the dynamic relationship and compensatory changes in Q and ET in response to rain and snowmelt events (Fig. 2). Normalized change in storage (ΔS) is superimposed on the climatograph to show the temporal patterns of water availability.
Figure 2 Daily climatograph (Rain, Snowmelt, Evapotranspiration, and Discharge [from the upland watershed to the vernal pool]) and normalized change in storage representative of conditions in South Deerfield, MA (Data sources: Orange Municipal Airport Weather Station, and USGS 01174500 stream gage)
As described in the Methods section (3.3.3), total storage is calculated as a water balance residual (Eq. 3.2) with a lower boundary condition based upon the wilting point water content of the dominant soil type. The absolute range of the estimated storage term was 61 mm (6/26/18) to 308 mm (10/29/17) across the water year. As expected, the strongest seasonal influence on storage is evapotranspiration. The water year (1 October to 30 September) can be divided into a temporal sequence of events and seasonal patterns.
4.1.1 Fall Recharge
The beginning of the water year is the transition from growing season to dormant season, also known as Fall recharge. Early in the water year at the South Deerfield pool, storage was high, reaching a peak as a 65 mm late-October storm completed recharge. Most of the water from this precipitation event entered the soil, becoming temporary storage. Because evapotranspiration is low at this time of year in relation to dormant vegetation, and decreasing air temperatures and daylength, comparatively little water leaves the system via this pathway.
4.1.2. Snow Accumulation and Melt
Precipitation in the form of rain has an immediate effect on wetland water level and storage. Snow has an equivalent effect, but is also linked to temporal patterns of accumulation and melt. During early and mid-winter, cycles of thawing and re-freezing were evident in the snowpack (Fig. 6). Colder temperatures yielded more consistent patterns of snow accumulation. Soil water content (the dominant component of total storage) declined during this period as drainage proceeded without new inputs (i.e., temporary storage in the snowpack).
Snowmelt begins during late winter and early spring, whenever air and snowpack temperatures rise above 0 °C. Evapotranspiration at the pool was still low and trees were dormant. An early-March snowmelt event caused an abrupt increase in the amount of water moving into storage.
4.1.3. Spring Transition
During the transition between dormant season and growing season, evapotranspiration begins to have a more substantial effect on the fate of water in storage. While precipitation at the pool remained consistent with earlier periods, plant activity and increasing air temperatures caused a larger proportion of inputs to be taken up or evaporated. At this point in the water year, storage generally decreased, but was dynamic—reflecting the interplay of precipitation inputs and evapotranspiration outputs. While spring in southern New England begins in late-March, most plants do not begin to leaf out until late-April or early-May. Changes in the amount of water in storage became more pronounced during the growing season, when foliage matured and plant growth, nutrient uptake, and water use all reached their annual maximum.
4.1.4. Growing Season
Estimated storage generally decreased as the year progressed, reaching the lowest point in July. At this point in the growing season, vegetation was in full leaf and air temperatures (total energy available for ET) were highest. The cumulative effect of high ET on soil water content (and vernal pool water level) was, as expected, inversely related. The effect of evapotranspiration on the amount of water in storage was most evident at this time of year. A large rainstorm (65 mm) in June 2018 caused the amount of water in storage to rise sharply. There was an immediate discharge response to the storm, yet the estimated subsurface flow hydrograph returned to antecedent baseflow conditions soon after the storm ended. This discharge response was notably smaller than the response to a rain event of the same size in October 2017 (65 mm), as well as a smaller event in January 2018 (44 mm). This indicates that precipitation entering the soil was quickly redirected to evapotranspiration, rather than remaining in detention storage or becoming discharge (QSS and/or streamflow). Predictably, as evapotranspiration decreased (entering the dormant season) the amount of water in storage increased.
4.1.5. Fall Transition
At the end of the water year, though storage is still highly variable due to the countervailing precipitation inputs and evapotranspiration outputs, the senescence of plants decreases the amount of water lost via transpiration. Fall storms, such as the large September 2018 event (79 mm), led to the typical fall recharge increase in storage.
These trends are observed in the fluctuation of vernal pool water level as well as in our storage estimate (Fig. 3). For this reason, storage is sometimes used as a proxy for water level change in wetlands. However, due to the unique hydrology of vernal pools, storage is not necessarily an appropriate approximation of water level in these systems. This point is addressed below.
Figure 3 The hydrologic seasons in a Massachusetts vernal pool (South Deerfield, MA) during the 2019 water year. The change in water level in these systems varies throughout the year depending on the relative seasonal influences of precipitation, shallow subsurface flow from adjacent uplands, leakage from the bottom of the pool, and evapotranspiration. These photos depict a pool on the same site, landscape position, and parent material as the pool analyzed for this study, which simply has vegetation more conducive to the visual demonstration of seasonal water level changes. The photos reflect general New England hydrologic trends, and are comparable to the study pool, but do not represent the specific data analyzed in this study
4.3. Exploring the Drivers of Water Level Change
In many cases, the direct relationship between rain events and wetland water level can be represented by the estimated storage term depicted in Fig. 3. This effect and response time is rapid and readily observed in small, closed vernal pool systems (Fig. 4). However, the weekly and daily water level measurements for the South Deerfield pool and Black Gum Swamp clearly diverge from watershed storage estimated as a water balance residual (Figs. 4, 5, and 6). Before this is discussed in more detail, the drivers of water level change in these unique systems need to be identified and explored. Figures 4 and 5 document the influence of rain, snowmelt, and evapotranspiration on vernal pool water level during the 2018 water year.
Figure 4 Relationship of vernal pool water level (South Deerfield, MA), Black Gum Swamp (Harvard Forest, Petersham, MA) water level, and rain and snowmelt (Orange Municipal Airport, Orange, MA) during the 2018 water year. Water level measurements were adjusted to a common datum, then normalized based on the maximum water level measured during the 2018 water year. Pressure transducer and manual measurements vary slightly because the pressure transducer data are averaged from 4-hour time step data, while the latter are measurements from a single point in time
4.3.1. Precipitation Effects
The water level observed in the pools is clearly dependent on rain and snowmelt inputs, but the persistence of these effects varies seasonally. Fall recharge has a distinct effect on the water level in both the pool, as recorded by the pressure transducer, and Black Gum Swamp. The clear, short-term effect of precipitation events can be seen when Tropical Storm Philippe (65 mm, 10/24/17) caused the water level in both systems to rise sharply, then recede to a more consistent level by mid-November. During the dormant season, as snow accumulated, water levels decreased in relation to little or no input to the soil, consistent with estimated total storage. However, during snowmelt and spring transition, storage and water level diverged, with vernal pool water level reaching its maximum as estimated storage dropped rapidly. During this period, snowmelt in the upland travels down to the pool through the soil mantle of the watershed as shallow subsurface flow (QSS), filling it to the maximum extent. This is also evident in the discharge hydrograph in Fig. 3, which rises as estimated storage decreases. During the spring transition period and the growing season, vernal pool and Black Gum Swamp water levels further deviate from total storage (Fig. 5).
Figure 5 Relationship of vernal pool water level (South Deerfield, MA), Black Gum Swamp (Harvard Forest, Petersham, MA) water level, and Hamon PET for the 2018 water year
4.3.2. Evapotranspiration Effects
As expected, evapotranspiration has a strong and persistent influence on water level trends in both Black Gum Swamp and the South Deerfield vernal pool. The primary period of interest regarding evapotranspiration is when the spring transition progresses to the growing season. The previously stable water level of the pool becomes more dynamic. The cumulative influence of evapotranspiration begins to overpower precipitation inputs, increasing inversely with vernal pool water level. While evapotranspiration occurs on the scale of millimeters a day, without consistent precipitation inputs, the amount of water lost this way can substantially affect water level (Table 1).
Weekly precipitation and evapotranspiration totals for South Deerfield, MA during the late growing season into the Fall transition period.
Weekly Precipitation (mm)
Weekly Evapotranspiration (mm)
On 8/24/18, in concert with evapotranspiration far exceeding recorded precipitation, the water level in the pool began to drop. This trend continued during the week of 8/31/18, when evapotranspiration remained at a similar level, and precipitation was limited to trace amounts. However, during the weeks of 9/7/18 and 9/14/18, several sizable storms occurred as evapotranspiration was decreasing for the season. This produced a notable increase in water level, bringing the pool to its almost full level, as anticipated for the Fall transition period (Table 1; Fig. 5).
The snow accumulation and snowmelt periods also effectively demonstrate the relationship between precipitation and evapotranspiration (Fig. 6). As noted earlier, precipitation from Tropical Storm Philippe caused a substantial increase in water level. This storm occurred early in the water year, when temperatures remained above 0°C, but the forest had already entered dormancy, so available energy produced a melting, then re-freezing cycle in the snowpack, causing water level to decrease. Over the next several months, air temperatures remained below 0 °C and inputs were generally retained in the snowpack. When the main snowmelt event occurred in early-March, the pool water level rose rapidly in response (Fig. 6).
Figure 6 South Deerfield vernal pool water level, calculated upland storage, and accumulated snowpack. Increased water in storage and vernal pool water level rise in February 2018 can be attributed to melting snowpack
4.3.3. Water Level and Storage
In late-February, the pool reached the highest annual level as a result of the snowpack melting. At this point, our estimate of storage as a water balance residual was also at a near-high point. However, after this point, our estimate of storage began to diverge from the recorded water level data. We determined that this was a result of storage being calculated for the site’s upland area. Estimated storage declines steadily as a result of the upland contributing area draining (QSS), and plant uptake rising, while the vernal pool (lacking an outlet) remains at its “brim full” condition for another ~ 75 days before the combined effects of evapotranspiration and precipitation become more evident in water level fluctuations. This distinct, characteristic water conservation effect of vernal pools is not represented by the estimated change in storage for the upland contributing area. While it is observable, this conclusion was also confirmed by attempting to correlate both Black Gum Swamp and vernal pool water levels with the change in storage water balance residual, which resulted in weak correlations. As a result, we concluded that storage calculated as a water balance residual did not realistically describe the patterns of water level fluctuation in either Black Gum Swamp or our vernal pool.
4.3.4. Vernal Pool Storage
While some components of upland and wetland storage are similar, the differences are pronounced enough for the resulting estimates to vary considerably. Using a conceptual model of the factors defining the hydrological regime of upland contributing areas and vernal pool systems, the terms specific to each system can be defined (Fig. 7).
Figure 7 Watershed and vernal pool inputs and outputs. Discharge from the watershed travels as shallow subsurface flow to the vernal pool, where it is detained. Water is not lost from the vernal pool as discharge since it has no outlet. Water losses from leakage (~ 2 mm/day) or deep seepage, contribute to cumulative changes in vernal pool water level
There is no standardized equation for determining storage in wetlands, or vernal pools specifically. However, with the understanding of the basic hydrologic structure of these systems, depicted in Fig. 7, the water balance calculation for vernal pool storage can be written as follows.
S i+1 = Si + QSS – ET – L + P*, (4.1)
S i = storage at the beginning of the day;
S i+1 = storage at the end of the day;
Q SS = shallow subsurface flow;
ET = evapotranspiration;
L = leakage;
P* = direct precipitation input adjusted to pool size.
Vernal pool storage required a 1 October initial value that would not result in a negative value at any point during the year, as was required for the upland storage equation. We iteratively estimated the initial storage at 400 mm. The corresponding calculated (non-negative) annual minimum was 7 mm.
Leakage was also considered in the calculation of vernal pool storage. We arrived at an in situ estimate of approximately 2 mm/day by reviewing the pool’s water level time series data to find physically and mathematically useful conditions (Axthelm 2019).
The final adjustment was the modification of the precipitation term. Although little precipitation is likely to fall directly in the pool, given that it makes up only 3% of the watershed area, during large events the volume of precipitation that lands in the pool is not insignificant, and should be represented. This is a key difference from the pools studied in Montrone (2013). Hence, to account for this addition to storage, daily precipitation was multiplied by 0.03 and added to Eq. 4.1.
Figure 8 Vernal pool water level, with both vernal pool storage (Eq. 4.1) and watershed storage (Eq. 3.2) calculated as water balance residuals
The calculation of vernal pool storage using these adjusted terms resulted in an estimate that more closely followed field measurements of water level, including accounting for the prolonged period of standing water during snowmelt and spring transition (Fig. 8). From 1/12/18, when pool water level rose due to snowmelt, to 7/21/18, when the countervailing effects of precipitation and evapotranspiration caused more pronounced fluctuations in pool water level, the correlation between pressure transducer measurements and estimated vernal pool storage was 0.84. Unsurprisingly, our storage model did not capture the complex water level changes of the growing season, and has certain limitations (Axthelm 2019). Because this time period is important for obligate vernal pool breeding organisms, the development of an accurately timed estimate is useful in both a research and regulatory sense, and holds promise for further refinement of vernal pool hydroperiod models.
The majority of existing research on wetland water balance analysis pertains to systems that differ from New England vernal pools in multiple ways. This includes research on non-vernal pool wetlands with different hydroperiods (permanence) or surface water connectivity, or non-New England vernal pools – such as lakes and wetlands in the Great Lakes Region (Mishra et al. 2010), prairie potholes in the Great Plains Region (Hayashi et al. 2016), and vernal pools in California (Montrone 2013). The water balance equation for wetlands outlets contain an additional mathematical term (Q, streamflow) that is not directly relevant to the movement of water within a vernal pool system. We addressed this by converting Q to QSS, to represent subsurface movement of water from the contributing area to the pool itself. However, without this adjustment (as in the above studies), comparison of wetlands with surficial connections to vernal pools is challenging. Water balance studies specific to vernal pools, including Montrone (2013), Pyke (2104), Hanes and Stromberg (1998), and Boone et al. (2006), are restricted to regions outside of New England. In Montrone (2013), the main input is direct precipitation, which plays a relatively small role in the water balance of New England pools. The California pools also have relatively small contributing areas, substantially diminishing the influence of our QSS term (Montrone 2013). This study also takes into account loss of water from overflow (O), which is not a pathway of loss in the eastern United States (Montrone 2013). Pyke (2004) and Hanes and Stromberg (1998) have merit as examples of adjusted water balances, removing the streamflow (Q) component, but are still not comparable to our study area due to location. Boone et al. (2006) comes closer to representing climatically similar conditions to the South Deerfield pool, and was able to successfully model vernal pool hydrology during the high water period, but the water balance analysis used in this study incorporates both surface water inputs and outputs, which are not relevant to our studied pool. Our model is specific to New England vernal pool conditions, and incorporates terms that model the movement of water into (P*, QSS) and out of (L, ET) the pool without discounting the characteristic lack of inlet and outlet in these systems. Additionally, our study successfully models relative water level rise and fall in our vernal pool during the high water period.
We used multiple data sets from credible sources to ensure the reliability of our collected and generated data. Meteorological and streamflow data from federally maintained and monitored NOAA and USGS stations were used. Since our water balance analysis was derived from these data, we can be confident in the reliability of our calculated values (i.e. AET, PET) as well. Comparison with a Long Term Ecological Research site with the National Science Foundation (Black Gum Swamp) was used to further validate the reliability of our data, and detect erroneous values and trends. As an additional measure in ensuring the reliability of our results, we incorporated both automatically and manually collected water level data from the study pool into our analysis. Both datasets communicate the same hydrologic trends, though minor variances exist due to frequency of data collection differing by method (as discussed in section 4.3).