Why To They Draw Down Lakes In The Winter?

Lake water level fluctuations compose a natural hydrological regime whose intra- and interannual variability in magnitude, timing, rate, frequency, and duration of high and depression water level events influence biogeochemical patterns and processes in littoral zones of natural lakes (Wantzen et al. 2008a). Water level fluctuation regimes in lakes with outflow control structures can significantly deviate from water level regimes in natural lakes (Kennedy 2005). Consequently, changes to the magnitude, timing, duration, rate, and frequency of high and low water level events can alter and degrade lake and coastal zone ecological weather (Wantzen et al. 2008b, Miranda et al. 2010, Zohary and Ostrovsky 2011). The direction and strength of various ecological responses to contradistinct lake h2o levels depend on the specific hydrologic metrics and resident biota (Hill et al. 1998, Eloranta et al. 2018). Therefore, reliable prediction of ecological responses and mitigation of negative impacts to littoral zone communities require authentic quantification of regulated water level fluctuations.

Almanac winter water level drawdowns (WDs) are an example of a regulated h2o level authorities that is regularly performed in temperate and boreal lakes as a event of wintertime power demand in hydroelectric lakes or to provide jump flood storage (Hellsten 1997). In recreational lakes of Massachusetts and other states in the northeastern United states, WDs are used to improve recreational value (east.g., canoeing, swimming) by attempting to reduce nuisance densities of macrophytes and protecting shoreline structures (e.g., docks, retaining walls) from water ice damage (Mattson et al. 2004). Water levels in WD regimes are lowered in autumn, reach target drawdown levels in wintertime, and are refilled in the spring (east.thousand., Mjelde et al. 2013, Carmignani and Roy 2017). Previous studies, primarily from hydroelectric and storage lakes in Scandinavia and Canada, have characterized WD hydrology to explain patterns in coastal zone communities predominantly as a function of WD amplitude and referred to hereafter as magnitude (e.thousand., White et al. 2011, Mjelde et al. 2013). For example, Sutela et al. (2013) quantified WD magnitude equally the xx yr mean deviation between the highest and lowest water level per winter in 16 regulated lakes in Finland. Ecological quality indices, represented by macrophytes, macroinvertebrates, and fish littoral assemblages, decreased with drawdown magnitude.

In dissimilarity to hydroelectric lakes, the spatiotemporal variability of WD regimes in northeastern U.s. recreational lakes has non been quantified despite its widespread and historical prevalence. Due to differences in climate, watershed and lake hydromorphological features, and water level direction goals, WD regimes in recreational lakes in the northeastern United States probable differ compared to hydroelectric lakes in due north temperate and boreal zones. Consequently, ecological responses to WDs in recreational lakes may differ, requiring quantification of h2o level fluctuation to reliably estimate potential ecological impacts (e.g., Carmignani et al. 2019). Furthermore, while virtually studies focus on WD magnitude, other hydrological components, such equally timing, duration, water level recession and refill rates, and degree and duration of lakebed exposure, may help to better predict ecological responses (Carmignani and Roy 2017, Hirsch et al. 2017). For example, the timing of low spring water levels contributes to low recruitment of spring spawning fish species that reproduce in shallow depths of the littoral zone (Linløkken and Sandlund 2016).

Nosotros aimed to improve understand the hydrology of almanac winter drawdowns in recreational lakes in Massachusetts, through continuous water level monitoring of WD lakes over several years. Our objectives were to (ane) appraise the interlake and interannual variability of WD timing, magnitude, rate, and duration, and (ii) evaluate the correspondence of empirical WD metrics with the magnitude, timing, and recession rate performance standards issued by the Massachusetts Partitioning of Fisheries and Wildlife for WD events (MassWildlife 2002) and restated in the Massachusetts Generic Environmental Impact Written report on Eutrophication and Aquatic Constitute Management (Mattson et al. 2004). Mattson et al. (2004) provide full general guidance to implement and perform WDs in Massachusetts to minimize impacts to in-lake and downstream nontarget organisms (e.k., mollusks, amphibians, reptiles, spawning fish species, mammals) and water-supply availability in local wells, while managing macrophytes. Hydrologic information nerveless in this study tin provide insight on how WDs are performed in Massachusetts and guide future WD management in northeastern US recreational lakes to assist balance ecological sustainability and recreational value, and to help guide realistic WD implementation in the face of climate change.

Written report site

Nosotros selected 18 lakes in western and central Massachusetts with current WD regimes and iii lakes with no history of annual winter drawdowns (Figure 1, Tabular array 1) using a stratified random approach to primarily capture a WD magnitude gradient (see Methods S1 in the Supplement for total details). Briefly, we stratified a set of 271 lakes of area ≥ 0.035 km² with historical WD information past lake surface area, lakeshore residential development inside a 100 m riparian buffer, and reported WD magnitude information, creating several grouping combinations from natural breaks in data distributions. We also targeted lakes in all level 4 U.s. Ecology Protection Agency (USEPA)-defined ecoregions within the level Iii Northeastern Highlands and 2 level IV ecoregions in the level 3 Northeastern Coastal Zone to assist reduce h2o quality variation amid waterbodies based on watershed land cover and geology for a related report effort on macrophytes (Methods S1). From stratification groups we randomly selected 16 WD lakes and 4 non-drawdown lakes. Nosotros were unable to admission 4 WD and one non-drawdown lakes and therefore replaced them with six additional WD lakes within the study ecoregions and within the range of stratification criteria.

Figure 1. Map of study lake locations inside the state of Massachusetts. Circles represent lakes with almanac winter drawdown water level regimes (WD) and triangles stand for lakes with no history of WDs. USEPA boundaries are level III ecoregions.

Tabular array i. Morphometric features and winter drawdown history of study lakes. Non-WD = non-winter drawdown lake, WD = winter drawdown lake, WL = water levels, NK = data not known, TP = total phosphorous. Drainage ratio = watershed:lake expanse.

All study lakes possess an outflow command construction and tin can be divers every bit natural drainage lakes (n = 14) or impounded lakes (due north = seven; Whittier et al. 2002). We define natural drainage lakes as lentic systems that had an original state of >10 ha before lake surface area enhancement via factory operations, water supply, and/or alluvion storage, amidst other reasons. We define impounded lakes equally dammed stream systems constructed for similar reasons. These lake types are common in Massachusetts and are also representative of the ii most prevalent lakes types in the Northeast region (Whittier et al. 2002). Time to come, both waterbody types are referred to equally lakes. Nigh lakes are outfitted with a valve or gate (due north = 14) to control lake water levels and outflows, while a minority utilise wooden stopboards (n = 7).

Inland Massachusetts has a continental temperate climate with iv seasons. Mean minimum/maximum July and January temperatures for western Massachusetts tend to exist 1–3 C lower than in central Massachusetts (Griffith et al. 2009), and winter precipitation averages 21.6–25.four cm (1981–2010) beyond these general regions (https://www.ncdc.noaa.gov/cdo-web/datatools/normals). Lake water ice-on typically occurs from December to January and ice-off between February and early May in Massachusetts lakes (Hodgkins et al. 2002). Stream inflows generally are highest in spring months (Mar–May), decline throughout the summertime (Jun–Sep), increase in the fall (Oct–November), and decline again through the winter (Huntington et al. 2009). Watersheds of report lakes have mixed land use with variable urban development ranging from 2 to xl% (median = ix%) with a general increase from west to due east, and relatively small proportions of pasture (0–15%) and agronomics (0–8%). Concomitantly, total watershed woods encompass ranges from 20 to 83% (median = 64%) among lakes (MRLC 2018). Forests are primarily composed of mixed deciduous and conifer stands, including northern, cardinal, and transition hardwoods (Griffith et al. 2009). Watersheds are underlaid past various geologies across the report area. Lakes located in the Northeast Highlands are characterized by fibroid-loamy to loamy soils and metamorphic bedrock- or limestone-derived coarse-loamy soils and calcareous bedrock (Griffith et al. 2009). In central Massachusetts or the Northeast Coastal Zone, lakes are underlain with sedimentary bedrock and alluvium soils, metamorphic bedrock with fibroid-loamy soils, or coarse-loamy and sandy soils (Griffith et al. 2009).

Materials and methods

Water level monitoring and quality control

H2o levels were monitored continuously from fall 2014 to fall 2018 at 18 drawdown and 3 not-drawdown lakes. We deployed paired nonvented pressure transducers (Onset HOBO U20L-01, Bourne, MA) in 14 lakes in September–October 2014 and in 6 lakes in September–Nov 2015 to collect pressure and temperature at bihourly intervals (Table 1). Water level data for Otis were provided by the Massachusetts Section of Conservation and Recreation, where information started in March 2012 upwards to May 2018. Water level information collection ceased in May–November 2018, resulting in ii–4 twelvemonth of winter water levels per lake (half-dozen for Otis). We more often than not followed methods from Stamp et al. (2014) for pressure level transducer (i.e., logger) installation and monitoring. In each lake we installed paired loggers next to the point of lake outflow, one underwater and one higher up water on shore. If admission was limited, we installed underwater loggers side by side to admission points (due east.yard., bridges, culverts) in other parts of the lake. All loggers were sheltered in PVC housing. Underwater loggers were fixed to dam or bridge abutments and suspended on nonstretch cablevision within a PVC pipage. If we could not attach an underwater logger to a fixed construction, loggers were fixed to a wood stake or metallic pipe that was anchored into the lakebed. All loggers were set to record at 2 h intervals. We downloaded data from loggers at least twice per year, pre- and post-drawdown result, and recorded relative peak from a secondary fixed location (eastward.g., staff gauge, spillway, dam abutment) to assist identify unintentional logger movement from ice formation/melt and instrument accuracy migrate.

Paired pressure level measurements were converted to h2o levels using HOBOWarePro software (version 3.7.8, Onset Computer Corporation, Bourne, MA) and imported into R software. Nosotros used the ContDataQC packet (Leppo et al. 2017, version 2.0.2.9001) in R (R Core Team 2017, version 3.4.2) to identify potential inaccurate water level records based on water level modify and minimum and maximum records. We flagged records with an absolute change of ≥3 cm and adjusted preceding information to account for credible transducer move or drift derived from discrete h2o elevation measurements from secondary locations. Nosotros removed negative water level records and values <1 cm that are unreliable measurements. Additionally, we examined water temperature information coupled with pressure level measurements to assistance place inaccurate water level records, such that records with water temperatures <0 C were flagged for inspection. To compensate for lost barometric air pressure readings at Wyola (nineteen Jun 2017–2 Nov 2018) and hence gauge water levels, we used predicted air force per unit area records generated from the closest study lake at Leverett (seven.2 km from Wyola).

Water level metrics

We defined two general h2o level time periods to calculate water level metrics: the WD catamenia or issue and the summer or the non-drawdown period. We farther divide the WD menstruum into iii fourth dimension frames or phases: h2o level decline (recession phase), drawndown water levels (stable stage), and the period of refill to predefined normal pool levels (refill phase, Figure 2). We first isolated WD periods by visually identifying the recession initiation date as the get-go record of consistent water level decline in the fall and the refill phase end date as the first record of increasing h2o levels that culminate in predefined summer pool levels in winter–spring. Summertime or normal pool water levels were defined as the median water level from non-drawdown periods in 2015 (n = 15) or from spillway elevations (due north = vi). Within the WD period, the end of h2o level recession or the start of the stable phase was marked past no consistent water level increase or decrease. The start of the refill phase was marked by a consistent visual water level increment in the hydrograph with no clear water level decline earlier reaching reference water levels. These water level phase boundaries were visually inspected over a xiv d flow. These definitions allowed for the inclusion of precipitation or melting events to influence recession and refill phases. For non-drawdown lakes, we divided water level records into spring/summer and fall/winter menstruum that covered 2 April–30 September and 1 Oct–1 April, respectively, to more often than not stand for to summertime and WD periods in drawdown lakes. For the summer menstruation and each of the WD period phases (eastward.g., recession, stable, refill), we calculated basic summary statistics including elapsing, minimum, maximum, medians, and selected quantiles.

Figure 2. (A) Instance hydrograph and associated wintertime drawdown (WD) metrics calculated for a unmarried WD menstruum. Water levels (y axis) are relativized to reference water level (e.thou., summer/normal pool level) such that relative h2o level = 0 represents normal pool level. WD flow phases (in italics and gray shades) include the recession, stable, and refill phases. Vertical dotted lines and changes in background color indicate the start and finish dates for WD phases. These dates are used to calculate WD elapsing, recession and refill rates, and WD magnitude. Duration exposed for a given depth (e.g., 0.5 thousand, i k) corresponds to elapsed time when relative waters exceeded this depth. (B) Example of recession and refill rates through time for a WD period, with boxplot displaying interquartile range and farthermost values >ane.5 times the interquartile range; this tin can be inferred from plot. (C) Photos corresponding to changes in water level throughout a WD period as labeled in (A).

For each WD event, nosotros quantified magnitude, recession and refill rates, and WD duration, and identified the timing of each WD stage (Figure ii). We calculated magnitude as the difference between reference pool level and the (1) maximum (i.eastward., everyman) h2o level recorded during the entire WD period, and (2) mean water level during the stable phase. Rates of recession and refill were calculated using consecutive records and summarized into median, minimum, and maximum values, and scaled from cm/h to cm/d for ease of interpretation. Durations were adamant in days for the entire WD period (i.e., recession start to refill end) and for each stable phase. Further, we calculated duration of exposure/emersion at 0.25 thousand depth intervals from 0.25 to 2.0 m depth contours relative to reference water levels. These WD metrics were calculated using bihourly records except for daily h2o level data at Otis between October 2015 and May 2018. Results are reported using median WD metric values and ranges per WD event.

Bathymetry collection and analysis

Nosotros sampled depths for all lakes in Apr–June 2015 or 2016 when water levels were at or to a higher place normal puddle levels. Following a cross-hatched design over the lake surface, depths were measured using a Garmin GPSMAP 431s with 1309–48,803 sample points per lake, depending on expanse. We used empirical Bayesian kriging in ArcGIS 10.three to interpolate unsampled depths from empirical depths (Krivoruchko 2012; come across Methods S2 in the Supplement for details).

Nosotros estimated the maximum depth of macrophyte colonization every bit a surrogate of coastal zone boundaries to decide the lakewide littoral zone surface area modified from Perlelberg et al. (2016). We established four to 21 transects scaled by lake area <10 m in depth to sample macrophyte presence from 29 August to nine September 2017. Transects were equally spaced along the 10 m depth contour or deepest depth contour within a lake or singled-out lake basins. We sampled macrophytes along transects perpendicular to depth contours at 0.5 k depth intervals using a double-headed rake suspended by rope. The rake was dragged approximately 0.5–one yard along the lesser at each sampling signal and was then inspected for macrophyte or macroalgae presence. Maximum depth values per transect were averaged for each lake and incorporated into littoral area exposure calculations for given WD events. If macrophytes were sampled at the deepest betoken of a lake, nosotros assumed the littoral zone was equivalent to the unabridged benthic surface area.

Calculation of lakebed and littoral zone exposure required coupling interpolated depths with h2o level records and applying WD magnitudes to bathymetry data. First, depths were matched to contemporaneous water level records at the fourth dimension of bathymetry surveys. Then nosotros found the simple difference between reference water levels and h2o levels at matching times. If the difference in water levels was greater than the accuracy of the pressure level transducers (one cm), we practical the difference to magnitudes to estimate exposure surface area metrics more than accurately. We calculated area of lakebed and littoral expanse exposure as the number of 1 thousand² depth cells for lake and littoral areas less than the maximum magnitude for a given WD outcome. Areas exposed were expressed every bit pct of the whole lake and coastal areas to compare across lakes.

We besides determined ratios of watershed to lake area (drainage ratios) to potentially explain WD water level metric variability. To calculate watershed areas, nosotros delineated watersheds from the signal of lake outflow using spatial analyst tools in ArcGIS 10.3 (ESRI 2015) based on the 3 grand US Geological Survey Digital Elevation Model. Finally, we calculated Spearman rank correlation coefficients among WD medians of period and phase durations, magnitude, exposures, rates, and drainage ratios.

Comparison to state guidelines

We compared observed water levels to the magnitude, timing, and recession rate guidelines of MassWildlife (2002) and Mattson et al. (2004). These performance standards were developed to protect in-lake and downstream ecological integrity using the best scientific evidence at the time while even so meeting the proposed goals of WD, namely, macrophyte control. The guidelines recommend magnitudes <3 feet (i.e., 0.914 yard), for recessions to start subsequently i Nov, reach target stable phase water levels by 1 December, and to refill to normal lake levels by 1 Apr, and for recession rates to not exceed 3 inches/d (i.e., 7.62 cm/d). Initiating WDs after i Nov prevents reductions in already low dissolved oxygen levels within shallow vegetated basins and flushing this water downstream, which can cause in-lake and downstream fish kills. Meeting target stable stage water levels past 1 December enables hibernating biota (e.g., amphibians, reptiles, beavers, muskrats) to relocate earlier substrate freezing and lake ice-on. Refilling to normal puddle level by 1 Apr will ensure express impact to available fish spawning habitat of spring littoral spawning species (east.g., concatenation pickerel, Esox niger; yellow perch, Perca flavascens). Lastly, recession rates are capped to maintain natural flows downstream and forbid stranding of fish and other aquatic organisms. If lake managers aim to deviate from the WD functioning standards, municipal conservation commissions in coordination with land agencies tin can permit special WD performance weather to meet management goals of macrophyte control. Several lakes are permitted to initiate WDs by 1 October (Boon), xv Oct (Goose, Otis, Wickaboag, Watatic), or sometime later Columbus Day (i.e., the second Monday in October; Hamilton) earlier the i November country recommendation. Additionally, several lakes are permitted to perform WDs with magnitudes >0.914 m (Otis, Goose, Onota, Garfield). Although several lakes possess special WD functioning weather that differentiate from country guidelines, we do not assess whether these lake-specific allow atmospheric condition were met. Rather, we highlight these special cases within the results when comparing against the original MassWildlife (2002) guidelines.

For WD magnitudes we identified the number and proportion of WD events of >0.914 m. For timing, we identified the percentage of WD phases that did and did not meet corresponding phase timing guidelines. For recession rates, we calculated internet h2o level rates over a 24 h moving window to compare confronting the recommended ≤iii inches/d (i.e., 7.62 cm/d) of water level refuse. We determined the percentage of internet daily recession rates ≥7.62 cm/d for each recession phase and determined the number of recession events for which the median recession rate was ≥7.62 cm/d of h2o level decline.

Results

Nosotros captured two–4 complete WD events per WD lake and 3–4 yr of water level data for non-drawdown lakes. Overall, nosotros collected water level data on 69 complete WD events across 18 lakes. Due to the timing of logger installation and logger failure, we did not capture complete recession phase durations in 2014–2015 for Brookhaven and Silvery, in 2015–2016 for Hamilton, Wickaboag, and Wyola, and stable and refill phases at Cranberry Meadow in 2015–2016.

Drawdown versus non-drawdown lakes

Overall, hydrology of WD lakes differed from that of non-drawdown lakes, particularly during winter months (Figure 3). In non-drawdown lakes, median h2o levels in winter months (i.eastward., 1 October–one Apr) ranged from 13.2 cm below reference pool level to 62.4 cm to a higher place reference pool level. The lowest winter water levels ranged from v.7 to 31.six cm below reference pool levels, with the extreme everyman h2o levels occurring in the 2016–2017 winter beyond all non-drawdown lakes. In comparison, median h2o levels in WD lakes across WD periods ranged from −202.4 to 0.ane cm. Median summertime water levels varied across years, with the lowest water levels in 2016, but were like across WD and non-drawdown (WD: median = 0.1 cm, range = −58.ix to 45.6 cm; non-drawdown: median = 0.5 cm, range = −30.0 to 102.iv cm).

Figure 3. H2o level fourth dimension series for report lakes. Water levels are expressed relative to reference pool level (relative water level = 0, dotted line). Black lines point water level medians, and gray lines represent the range per twenty-four hour period of year over 3–4 yr. Note that y-axis scale varies by lake. Come across Figure 1 for lake locations.

Magnitude

Nosotros captured a magnitude gradient with interannual median of stable stage water levels ranging from 0.001 to 2.16 m with an overall median of 0.66 k across lakes (Figure 4, Table S1). Median maximum magnitudes (i.e., lowest water levels below reference levels) ranged from 0.09 to 2.23 m with a lowest maximum magnitude of 0.07 m at Silver and the highest at 2.66 grand at Onota (Figure 3). Median h2o levels during stable phases were consequent among years for nigh lakes, varying <10 cm for ix lakes and <20 cm for fourteen lakes. Onota showed the highest interannual variability in maximum magnitude (1.67 m) because of a authorities with a sequence of 2 shallow drawdowns and one deep drawdown in successive years. Maximum magnitudes exceeded the 0.914 thousand magnitude guideline recommended past MassWildlife (2002) in 6 of 18 WD lakes and twenty of 74 WD periods (27%) consistently (east.g., Otis, Onota, Garfield, Goose) or variably (due east.1000., Stockbridge 3 of 4, Wyola ane of 3) amidst years. Median stable phase water levels for 5 lakes also variably exceeded this threshold among years (east.yard., Otis, Onota, Garfield, Goose, Stockbridge). Four of these lakes (Otis, Onota, Garfield, Goose) were permitted to exceed the magnitude guideline.

Figure 4. Interannual magnitudes categorized as median (dark grey bars) and maximum (light gray bars) h2o levels during the stable phase. Not-drawdown lakes are Quacumquasit, Leverett, and Congamond.

Area exposed

Interannual median lake exposure ranged from 1.three% to 34.2% across lakes (Figure 5, Tabular array S2). Median littoral exposure ranged from 9.three% to 64.8% across lakes (Effigy v). Lake area and littoral area exposed was largely inside 10% exposure deviation amidst years for most lakes, except for Onota, Stockbridge, Wyola, and Wyman. Onota displayed the highest interannual variability in lake and littoral percent exposure. The highest maximum magnitudes typically equated to the highest coastal and lake area exposed. All the same, relatively small-scale magnitudes at a few lakes resulted in relatively high percent littoral and lake area exposed (e.g., Silver, Watatic). Conversely, several lakes with moderate to loftier magnitude had relatively low percent exposures (e.g., Goose).

Figure v. Median (± range) percent lake expanse and littoral expanse exposed at maximum drawdown magnitudes. Lakes are ordered by decreasing mean drawdown magnitude.

Durations

WD menstruum durations ranged from v to 246 d with an overall median of 171 d (Figure vi, Table S2). Otis exhibited the longest median elapsing at 236 d and Wyman the shortest at 17 d. WD duration varied interannually within lakes from 24 to 175 d with a median of 54 d. The recession phase comprised xix.4%, the stable phase 62.5%, and the refill phase xiii.5% for a median WD period. WD phases too exhibited wide variability. The recession phase varied from 3 to 70 d (median = 24 d), the stable phase from 0 to 215 d (median = 98 d), and the refill phase from 0 to 139 d (median = 17 d, Tabular array S2).

Figure 6. WD period duration and timing for 3 or 4 drawdowns per lake (color coded by year). Each WD period is divided into recession, stable, and refill phases past line types. Vertical dashed lines represent the Generic Ecology Impact Written report guidelines recommended for WD start and terminate dates. For Wyman, ii–3 WDs are conducted per wintertime year. Elapsing values can be institute in Tabular array S2. Lakes are ordered by decreasing mean drawdown magnitude.

Forth the magnitude gradient, depth contours were variably exposed across lakes, and this exposure varied interannually inside lakes (Figure S1). The 0.25 m depth contour was exposed in xvi WD lakes, 0.v thousand contour in thirteen lakes, 1 chiliad contour in 6 lakes, 1.v m contour in 4 lakes, and 2 m profile in 2 lakes. Median duration exposure lasted from one to 127–229 d for any given depth profile beyond lakes.

Timing

WD events were initiated between 1 October and 1 December (Figures six and 7), excluding late WD events from Wyman that occurred in Feb–April. Overall median recessions ceased, and stable phases started on 24 November and ranged from 7 October to nine Jan. Median stable phase ended, and refill started on 12 March and ranged from 4 January to v June. Refills and the entire WD flow ended (i.due east., refill end) between 13 January and 26 June with a median of 13 April. There was variability in timing across years (Figure vii). The median recession start dates varied from 21 October to 29 Oct and end dates varied from 16 November to 1 December. The median refill start dates varied from 27 Feb to 23 March and stop dates varied from iv April to 23 April.

Figure 7. Density of recession (left) and refill (correct) start and end dates (solid, dashed) aggregated past lake and paneled by winter year (eastward.g., 2014–2015). Points along the x axis correspond to start (filled) and end (open) dates. Dashed vertical lines represent MassWildlife (2002) recommendations for WD initiation start (1 Nov) and recession stop dates (1 Dec) and refill end date (1 Apr). Notation difference in x-axis fourth dimension scale between recession and refill graphs. Phase dates from late winter–bound WD periods in Wyman are not included.

Relative to the MassWildlife (2002) WD timing guidelines, 83.1% of WD events were initiated before 1 November, with 8 distinct WD periods that occurred in Wyman in February to Apr. Half dozen lakes were permitted to initiate recessions in early to mid-Oct and comprised 35.2% of recession phases started before 1 November and thirty% of total recession phases. Stable phase water levels were reached (i.east., recession end) before 1 December for 63.6% of WD events. Lastly, 70.half dozen% of WD periods did not attain reference water levels by 1 April (Figure 7).

Rates

Sequential recession and refill rates varied across lakes and years (Tabular array S3). Annual median recession rates varied from 0.9 to 5.eight cm/d with an overall median of 2.four cm/d across lakes. Interannual lake variation of median recession rates ranged from 0 to ix.half dozen cm/d and the highest median recession rates per lake ranged from 1.ii to 12 cm/d with interannual variation ranging from 0 to nine.6 cm/d across all lakes. Overall, the highest recorded recession rates occurred at Onota with 188.4 cm/d, followed past 73.ii cm/d at Wickaboag, and 71.7 cm/d at Otis. During recession phases, water levels also increased, most notably during 2017–2018 when a relatively big precipitation event occurred during the recession phase.

Annual median refill rates varied from 0.9 to 11.half-dozen cm/d across lakes. Lake interannual variation of median refill rates ranged from 0 to 35.4 cm/d with the highest median rate of 35.4 cm/d and the lowest at 0 cm/d. The highest overall refill rates occurred in Stockbridge (315.6 cm/d), Garfield (126 cm/d), and Greenwater (98.four cm/d). Similar to recession rates, declines in h2o level occurred during refill phases. Several lakes reached reference pool level after a strong precipitation/melting effect in January 2018 and did not attempt water level recession again.

Of the 71 recession periods, 39 possessed net daily recession rates that exceeded the 7.62 cm/d standard (MassWildlife 2002, Figure viii). Several lakes exceeded the 7.62 cm/d standard consistently beyond WD periods, including Watatic (5.1–30.2% of time), Otis (iv.5–27.8% of time), Garfield (eight.three–17.1% of time), Brookhaven (1.iii–30.1% of fourth dimension), Wyola (2.2–34.8% of time), and Hamilton (1.eight–27.0% of time). Other lakes also exceeded this threshold merely not consistently beyond WD periods (eastward.g., Onota, Ashmere, Stockbridge), and few lakes did non exceed this threshold overall (Silvery, Goose, Boon, Buel). There were 2 recession events in Wyman where median internet recession rates exceeded 7.62 cm/d.

Figure 8. Density and range of median daily cyberspace recession rates aggregated across WD events per lake. The horizontal dashed line is the recession rate guideline (−7.62 cm/d) from Mattson et al. (2004). Lakes are ordered by ascending WD magnitude.

Metric correlations

Mean lake WD magnitude was positively correlated with mean lake proportions of lake area exposed and littoral area exposed (Figure 9). Generally, these magnitude metrics were correlated with duration metrics. Notably, magnitude was positively correlated with refill duration and recession duration, and negatively, admitting weakly, with stable phase duration. Duration and magnitude metrics exhibited relatively weaker correlations with recession, including faster recession rates with increasing magnitude. Slower refill rates were correlated with decreasing drainage ratios. Lastly, decreasing drainage ratios were moderately correlated with longer recession and refill durations.

Effigy 9. Paired correlations and scatterplot matrix for selected WD metrics in WD lakes (n = 18). Numbers in the upper diagonal are Spearman rank correlation coefficients, with darker colors representing higher negative or positive correlations. Points in scatterplots are interannual medians by lake with linear trendlines and 95% confidence intervals. Diagonal plots stand for WD metrics probability density distributions. WD metrics abbreviations and respective units are: DrainRatio = drainage ratio (watershed area/lake area), Mag = bihourly hateful drawdown phase water level (1000), LakeExp = proportion of whole lake expanse exposed, LittExp = proportion of estimated littoral area exposed, WDDur = duration of entire WD period (d), RCDur = duration of recession phase (d), DDDur = duration of drawdown phase (d), RFDur = duration of refill phase (d), CRCRate = mean cumulative recession rates on a natural log scale (cm/d), RFRate = mean refill rates on a natural log calibration (cm/d). Positive changes in cumulative recession rates equate to faster water level declines.

Discussion

Recreational lakes in Massachusetts exhibit wide-ranging WD magnitude, duration, rates, and timing; this study is the first to certificate this variability over multiple lakes and years, providing essential information for understanding ecological impacts of WD. Almost lakes possessed magnitudes of <0.914 m that remained consistent across years; yet, differences in lake bathymetry and h2o quality (i.e., transparency) translated to variable lake and littoral zone exposure. Timing and elapsing of WD refill phases varied widely across years, suggesting the importance of seasonal-specific precipitation and temperature events. The bulk of WD events deviated from Massachusetts timing and recession charge per unit performance standards, which may accept unintended ecological impacts to nontarget species (e.thousand., express fish spawning habitat, mollusk stranding). Understanding the timing, duration, and rates of WD events in add-on to magnitude could exist critical for predicting WD impacts on lake ecosystems and managing WDs nether climate change.

Potential drivers and ecological implications of WD regimes

Most WD events and lakes had magnitudes less than 0.91 1000 (0.001–2.16 m, mean = 0.66 m), in line with the guidance of Mattson et al. (2004). Magnitudes that exceeded this standard were found in five lakes that obtained local and state government approval. In comparison to hydroelectric lakes, WD in Massachusetts recreational lakes are typically smaller. For example, in hydroelectric lakes of Canada and the U.s. states of Maine and New Hampshire, studies study magnitudes of 0.3–seven.2 m (n = 15, mean = iii.0 grand, Trottier et al. 2019) and 0.8–10 chiliad (northward = 24, White et al. 2011). This suggests WD direction context is likely an important driver for magnitude decisions. Many WD regimes are implemented in recreational lakes to dewater shoreline structures (e.thousand., docks, retaining walls, dams) earlier ice-on to prevent damage from ice erosion, to reduce nuisance densities of macrophytes that may impede recreational activities (Clayton 1996), or to forestall the spread of nonnative invasive species (Hussner et al. 2017). Thus, virtually magnitudes are relatively small to correspond to shallow depths of shoreline infrastructure, but larger magnitudes may be conducted to maintain dam integrity (e.grand., Otis) or expose a significant portion of a nonnative invasive species like Eurasian watermilfoil (Myriophyllum spicatum; eastward.chiliad., Mattson et al. 2004).

Recent WD-ecology research derives mostly from hydroelectric and storage lakes that can experience relatively large WD magnitudes (e.1000., >2–3 m), which are greater than those observed in our study. Typically, these larger magnitudes have more pronounced impacts to populations and communities (e.g., Haxton and Findlay 2009, White et al. 2011). However, even relatively modest WD magnitudes can have substantial ecological impacts. For instance, within a subset of the current study lakes, Carmignani et al. (2019) found WD regimes with <i thou magnitudes limited freshwater mussel distributions deeper than stable phase h2o levels, presumably due to their low mobility and susceptibility to desiccation. Also, short pulses of large and rapid h2o level recession as observed in our study may betrayal loftier mussel densities on shallow benthic shelves (due east.g., Onota). Although rare, these relatively extreme events tin accept a lasting impact to nontarget biota populations (Richardson et al. 2002).

Although WD magnitude was moderately correlated with coastal and lake exposure, these relationships were not strong, emphasizing the importance of morphometry in determining exposure. In shallow lakes or lakes with expansive shallow benthic shelves, relatively pocket-size to moderate magnitudes can betrayal a significant proportion of lakebeds. In contrast, lakes predominantly composed of steep-sided basin slopes show small whole-lake exposure even at high magnitudes observed in this written report. Therefore, biotic assemblages occupying shallow uniform depths are more than vulnerable to direct WD impacts, such every bit desiccation and freezing, compared to biotic assemblages distributed along moderate to high sloping lake basins.

Similarly, narrow littoral zones are more susceptible to exposure even at small WD magnitudes because of a lake's bathymetry (Duarte and Kalff 1990). Food availability and factors that influence water transparency, including phytoplankton and nonalgal suspended solids (Brezonik et al. 2019), will affect littoral zone depth boundaries (i.e., macrophyte colonization) and hence littoral zone exposure. Because littoral zones can provide unduly loftier energy and habitat resources for a diversity of consumers across lake morphometries (Vander Zanden et al. 2011), it is of import to estimate littoral zone exposure. Although deep and steep-sided lake morphometries may be less sensitive to overall lake area exposure, valuable benthic-littoral habitat and energy resources are naturally constrained to relatively small areas (Vadeboncoeur et al. 2008) and hence are particularly susceptible to regulated water levels (Eloranta et al. 2018). Fifty-fifty at WD magnitudes of <0.91 1000, large proportions of littoral zone habitat were exposed in our study. Accurate estimation of lake and coastal exposure areas will crave fine-scaled bathymetry data to generate expanse exposed and volume lost and will require depth estimations of littoral zone boundaries during summer months.

Typically, WD periods recorded in our report lasted >120 d, such that h2o levels were receding, refilling, or in stable phases for most of the nonsummer months (due east.1000., Oct to Apr–Jun). Although entire WD period elapsing and magnitude were weakly correlated, recession and refill phase durations were moderately to strongly correlated, indicating that more time is needed to reach target stable drawdown water levels and particularly summer pool levels. Consequently, stable phase water levels are maintained for shorter durations with increasing magnitudes. These patterns tin can be attributed to variable interlake WD direction decisions to maintain stable water levels up to unlike dates, and reflect interlake hydrogeomorphic differences (i.e., inflows, outflows, residence time) in response to precipitation events. This is supported by drainage ratio correlations with refill and recession durations, and refill rates in our written report that suggest inflows and lake water storage mediate WD regimes.

The timing of WD phases resulted in timing that did not run into the recommendations of the MassWildlife (2002) standards. The majority of WD events were initiated before the 1 Nov guideline and reached reference pool levels later on ane April. In contrast, the bulk of WD recessions ended by the outset of December, per recommendation; however, this might be the effect of relatively early WD initiation dates. Recessions were predominantly initiated before 1 November across years, which suggests lake managers largely dictate and control recession starts. Of the 18 WD lakes, vi lakes had special operation atmospheric condition to initiate water level recession before the 1 November guideline in early on and mid-October. Potential reasons for intentional early get-go dates include meeting target stable phase water levels before ice-on, given Massachusetts state guidelines on downstream flows and in-lake recession rates, and permitting shoreline property maintenance for shoreline residents. Reaching stable phase h2o levels before ice-on enables targeted benthic areas for macrophyte control to dewater and become exposed to freezing temperatures. Additionally, continued h2o level drawdown after water ice-on tin be a rubber business organization for wintertime water ice recreational activities (Mattson et al. 2004).

The college interannual variability for the timing of recession end and for refill offset and end dates implies less water level control and more influence of environmental factors such equally precipitation and water ice cook. For example, sustained cold winter temperatures into late March and April of the 2014–2015 winter synchronously delayed refill phases into mid-April to May across many of our study lakes. In contrast, the timing of the start of refill phases and reaching normal pool levels in 2017–2018 was highly variably beyond lakes. Furthermore, longer refill durations and slower refill rates correlated with decreasing drainage ratios indicate that lake inflows likely influence the timing to meet summer pool levels. Thus, WD phase timing differences amid lakes and betwixt years demonstrate the interaction between different WD management practices and lake water budgets.

Since the MassWildlife (2002) guidelines are to help minimize ecological impacts, the general incongruity with timing standards may have ongoing negative ecological effects. The one Apr refill guideline is in part to ensure access to disquisitional shallow-water spawning habitat for spring spawning species (MassWildlife 2002) such every bit yellow perch, chain pickerel, and northern pike (Esox lucius). Impacts to annual recruitment volition depend on the amount of spawning habitat bachelor below drawdown water levels and the disturbance to eggs from fluctuating water levels and wave activity (Larson et al. 2016). Future investigations can assistance assess the availability of spawning habitat (e.g., water temperature, substrate) nether different refill scenarios (Papenfuss et al. 2018) and for different fish species that require unlike spawning substrates. The i November recession start guideline delays WD until colder weather to assist preclude in-lake and downstream fish kills resulting from warm, poorly oxygenated h2o, particularly in shallow, macrophyte-dominated lakes (MassWildlife 2002), but more research is warranted to appraise the potential for fish kills. Recession initiation dates earlier 1 November, by contrast, may benefit benthic species susceptible to exposure. Warmer h2o temperatures in mid-Oct could allow for more efficient move of benthic organisms (e.g., mussels, gastropods), given that recession rates are not extreme. Additionally, earlier recession initiation could enable amphibians and turtles to select overwintering benthic areas that would remain submerged during a drawdown and prevent winterkill events. Lake direction volition demand to consider and residuum these potential impacts, given their downstream and lake community composition. Overall, more research could help appraise the effects of variable timing on nontarget biota and habitat to help refine WD performance timing guidelines.

Recession and refill rates were similar across most lakes and years in our written report; however, the ranges of rates had several key insights. Outset, we documented relatively extraordinary rates within a few recession and refill phases. For instance, nosotros observed maximum cyberspace recession rates >25 cm/d for 4 recession phases reaching up to 62.9 cm/d. 2nd, recession phases often contained rates ≥7.62 cm/d, the MassWildlife (2002) guideline. Although, the per centum of these rates largely equanimous a minority of rate records, several lakes consistently barbarous within or exceeded the recession rate guideline across WD periods. Few studies take investigated the event of recession rates on ecological responses, simply depression-mobility organisms like freshwater mussels are especially susceptible to rapid dewatering. Galbraith et al. (2015) found most mussels were stranded under 4 cm/d and eight cm/d recession rates but with variable species-specific bloodshed afterwards stranding. Given that many WD events in the current report possessed net daily recession rates >four cm/d, increases in magnitude with like recession rates volition likely impact existing mussel assemblages, for which the distributions are already restricted past ongoing WD regimes (Carmignani et al. 2019). Also, rapid recession rates may trap fish in shallow pools, leading to mortality via stranding or because of stressful overwintering weather (Nagrodski et al. 2012). More field-based studies are needed to estimate the effect of typical recession and extreme recession rates on littoral communities. Furthermore, more than research is needed to gauge the impact of loftier outflows to downstream communities associated with WD recession phases, every bit these menstruation patterns are likely atypical of natural streamflows during fall months.

WD management implications

The level of annual water level fluctuation is determined by a lake's water budget (eastward.k., inflows, outflows, residence time, evapotranspiration), and can exist coarsely predicted by drainage ratios such that lakes that incorporate an increasing percentage of watershed area will have smaller water level magnitudes (Keto et al. 2008). Based on our results, we hypothesize larger WD magnitudes in lakes with smaller drainage ratios restrict control on the timing, duration, and rates of recession and refill phases considering of higher dependency on local precipitation and temperature events as compared to smaller magnitudes in lakes with larger drainage ratios. Therefore, larger WD magnitude regimes with small drainage ratios may accept more difficulty meeting WD performance standards for timing and rates. To aid balance WD direction goals and lake ecological integrity, simulating WD magnitude scenarios under various water budget atmospheric condition can help approximate the duration, timing, and rates of WD phases needed to meet or define functioning standards. Furthermore, the use of easily determined hydrogeomorphic metrics similar drainage ratios tin can assist classify the hydrological status of lakes as developed in Republic of finland'southward hydroelectric lakes (Keto et al. 2008). Further inquiry could atomic number 82 to more detailed hydrological nomenclature of lakes and potentially adapt specific WD performance guidelines according to hydrological classification.

The effectiveness of WD regimes equally a macrophyte command strategy in drawdown exposure zones strongly depends on winter weather conditions and the resistance of target species to freezing and desiccation (Cooke 1980). Most WDs monitored in this study were initiated in October, reached target water levels earlier or in the beginning of December, likely before ice-on, and were refilled in April. This timing and duration allow ample exposure to rhizome-damaging conditions. Lonergan et al. (2014) experimentally found that sediment temperatures at −five C sustained for ≥24 h, or below a sediment h2o content threshold for ≥48 h, prevented regrowth of Eurasian watermilfoil, a widespread invasive species in the Northeast. However, the presence of water ice and snow cover concurrent with freezing and desiccated soil will dictate the level of rhizome mortality (Lonergan et al. 2014). Early freezing of exposed lakebed followed by snow encompass can sustain frozen soil atmospheric condition that may result in effective macrophyte rhizome bloodshed. In contrast, snow encompass earlier the onset of freezing temperatures tin effectively insulate sediment to a higher place freezing and regulate freeze–thaw cycles (Huntington et al. 2009 and references therein). Thus, sufficient fourth dimension is needed to allow sediment dewatering before ice formation, along with consecutive subzero freezing days to command susceptible nuisance species. Monitoring of exposed soil temperature and wet and of water ice and snowfall comprehend durations during WD periods could help determine the timing of refill once macrophyte mortality conditions are met and the lake is ice-gratuitous (Lonergan et al. 2014). Additionally, incorporating fine-scale estimates of bathymetry could help place benthic areas of high topographic heterogeneity that possess variable moisture and temperature conditions, and hence may exist less vulnerable to macrophyte mortality.

Likely changes in lake h2o level regimes from climate change are a tiptop concern amidst lake management stakeholders (Magee et al. 2019). Climatic change is projected to increase winter temperatures, increase winter rainfall, reduce the extent and duration of snow cover, increase the frequency of brusque-term droughts, and shift the timing of bound floods in the northeastern U.s. (Hayhoe et al. 2008, Huntington et al. 2009). Additionally, the current trend of earlier water ice-out dates (Hodgkins et al. 2002) is expected to continue in the future, forth with the potential of shorter ice comprehend durations and reduced ice thickness (Huntington et al. 2009).

These projected changes pose challenges to the use of WD regimes as a macrophyte control strategy that also minimizes ecological impacts and maintains recreational value. Specifically, warmer and wetter winters may limit macrophyte mortality by keeping exposed sediment above mortality threshold temperatures and past keeping sediments moist from rainfall and associated water level fluctuations. Another major business organisation associated with climate change is delayed or incomplete refill to reference pool levels because of a spring drought (Magee et al. 2019). In several lakes in Connecticut, McDowell (2012) documented refill phases that did non reach summertime puddle levels until mid to tardily May, as a event of a springtime drought in 2012. Delayed refill extending into summertime months could also decrease recreational opportunities for boating and fishing (Miranda and Meals 2013) and may subtract lakefront property values (Hanson et al. 2002).

Anticipation of climate-related changes in precipitation and temperature regimes could help guide WD management strategies that accommodate the magnitude, duration, and even frequency of WDs to sustain ecological integrity and maintain recreational value. Heterogeneous watershed characteristics (e.grand., land use and embrace, gradient, drainage density) and lake-specific factors (morphometry, residence time) that regulate lake h2o levels (Molinos and Donohue 2014) could help decide lake-specific adaptation strategies for WD regime direction (Magee et al. 2019).

Data needs and conclusions

The scarcity of lake water level records and water level monitoring efforts poses a large challenge to assessing WD impacts on lake ecosystems and understanding the role of anthropogenic stressors interacting with natural controls (eastward.one thousand., climate change, watershed land cover, lake morphometry). Increased monitoring of lake levels at ecologically relevant temporal resolutions and scales is needed to understand a lake's natural hydrological character, and to help explain current and time to come ecological patterns (Magee et al. 2019). Our bihourly measurements of lake water levels enabled the documentation of short-term extreme events (due east.thou., high recession rates) and captured the overall variability of WD regimes inside and between years. Predicted climate changes suggest that winter h2o level regulation may progressively carry over into summertime months (Magee et al. 2019). To sympathise the ecological effects of such a shift, we strongly recommend that water levels be monitored twelvemonth-circular, as recent testify suggests summer water level fluctuations impact h2o quality more (e.one thousand., cyanobacteria blooms; Bakker and Hilt 2016) than winter drawdowns (Elchyshyn et al. 2018). Integrating knowledge of the natural range of variability of unregulated lake levels over long time scales (i.e., decades; Hofmann et al. 2008, Molinos et al. 2015) could help to predict future water level changes in regulated lakes within like hydromorphic characteristics and direct management to mitigate and anticipate related h2o quality issues (Lisi and Hein 2019). We also need increased modeling efforts to sympathise the drivers and patterns of lake water level fluctuations beyond unregulated and regulated h2o level regimes and lake types (east.m., drainage, seepage) to help u.s.a. define the natural range of water level fluctuations and fix WD management expectations (Magee et al. 2019). Application of recently adult models can meliorate our agreement of lake h2o budgets at local and regional scales and assistance to estimate the hydrological impacts of varying WD regimes in combination with watershed land cover and state utilise (Hanson et al. 2018). Knowledge of cardinal lake characteristics including lake morphometry, water transparency, food status, and watershed hydrogeomorphic attributes and how those characteristics command in-lake abiotic and biotic dynamics can help provide context for WD direction and its effectiveness in the face of ongoing climate change.