Northeastern Naturalist 460 A. Tiwari and J.W. Rachlin 22001188 NORTHEASTERN NATURALIST 2V5(o3l). :2456,0 N–4o7. 83 Ecological Effects of Road De-icing Salt on Adirondack Forests and Headwater Streams Athena Tiwari1,* and Joseph W. Rachlin2 Abstract - We collected water samples from upstream and downstream sites over 3 years on 18 study streams in the Adirondacks, NY, and conducted analysis via ANOVA for the presence of road-salt runoff, as measured by chloride ion content. Streams crossed by state roads received more road-salt runoff than streams crossed by county roads, as shown by higher mean chloride loads across different sampling years (P ≤ 0.01). The chloride load in streams was not reliably higher downstream from a road as opposed to upstream from a road for either state or county roads but varied in different sampling years (P ˂ 0.001–P ˃ 0.05). We collected a total of 1259 nymphs of Ephemeroptera, Plecoptera, and Trichoptera during water sampling. High levels of road-salt runoff were not associated with lower levels of Plecoptera or Trichoptera. Neither numbers of individuals nor numbers of genera of Ephemeroptera, Plecoptera, or Trichoptera collected per month showed any pattern when regressed on stream-chloride level. However, we detected no Ephemeroptera above a relatively high level of road-salt runoff (154 mg/L chloride ion). We employed the point-centered quarter method to assess forest composition on 10 transects above and below state roads. We conducted further analysis on trees in the lowest quartile of circumference in each transect as a representation of tree-species recruitment. Mean chloride-ion content of study streams, indicating adjacent forest exposure to road-salt runoff, was associated with greater recruitment of Abies balsamea (Balsam Fir), and lower Shannon–Weiner diversity. At the highest chloride levels, there was almost no recruitment of any species but Balsam Fir. Soil-cation analysis and linear regression, however, did not indicate concomitant depletion of plant nutrients, and therefore, we did not confirm the cause of the apparent relationship between higher road-salt runoff and higher Balsam Fir recruitment. Introduction Approximately 20 million tons of road salt per year are applied in the US (Anning and Flynn 2014), and ~22 tons per lane mile in New York State (Kelting and Laxson 2010, USGS 2011). Road-salt pollution is increasing (Williams et al. 1999), is known to persist in both surface and groundwater (Demers and Sage 1990, Williams et al. 1999), and is considered a serious threat to freshwater ecosystems (Kaushal et al. 2005). Road-salt runoff into streams can be measured as chloride-ion (Cl-) concentration because Cl- tends to remain in solution (PMRA 2006). Juvenile stages of the Ephemeroptera, Plecoptera, and Trichoptera (EPT), used in biomonitoring for their sensitivity to water quality (Lenat and Penrose 1996, Norris and Georges 1Laboratory for Marine and Estuarine Research (LaMER), Lehman College, Bronx, NY 10468. 2The Graduate Center, City University of New York and LaMER, Lehman College, City University of New York, New York, NY 10017. *Corresponding author - athenatiwari13@gmail.com. Manuscript Editor: Hunter Carrick Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 461 1993, Wallace et al. 1996), may be exposed to residual chloride from road-salting. Sodium ions tend to bind to soil particles and may displace important plant macro- and micronutrient cations (Granato et al. 1995, Kelting and Laxson 2010, PMRA 2006). The Adirondacks of New York State present an opportunity to compare the results of very different road-salting regimes. County roads in the Adirondacks are maintained by county highway departments, which follow the traditional local practice of allowing a snowpack of a few inches to form on top of the road asphalt. After plowing, workers apply a mixture of salt (92%) and sand (8%; to keep the sand pile from freezing) to this snowpack (Craig Donaldson, Harrietstown Highway Supervisor, Harrietstown, NY, pers. comm.). In at least 1 Adirondack town, no salt is added to sand piles; thus, county roads receive sand only. In this case, local residents had expressed concern that sodium might contaminate local wells (Mark Yandon, Town and Hamlet of Newcomb Superintendent of Highways, Newcomb, NY, pers. comm.). In contrast, state roads—maintained by the New York State Highway Department—are plowed down to bare pavement after each snowfall, and receive only salt, but no sand (Craig Donaldson, Harrietstown Highway Supervisor, Harrietstown, NY, pers. comm.). Chloride-ion concentrations far in excess of background levels have been measured in Adirondack forest streams 100 m below points at which the streams were crossed by a road, and months after winter road-salt application or spring melt (Demers and Sage 1990). There remains a need for studies that compare road-salt impact between streams crossed by roads that receive different winter treatments. In this study, we examined the effects of road-salt runoff on ecosystem health by comparing ecosystem parameters at study sites in Adirondack streams and forests above and below roads that receive different winter treatments. Field-site Description We chose for study within the Adirondack Park, NY, eighteen 1st- or 2nd-order streams (Strahler 1957) that are crossed by roads. We established 2 study sites per stream—1 each 30 m upstream and downstream from the road. We chose 12 streams crossed by county roads. Of the 12 study streams crossed by a county road, 8 were crossed by a road that received the traditional mixture of 92% sand, 8% salt above a snowpack of a few inches and 4 streams were crossed by a road that did not receive any salt treatment in winter, only sand on top of the snowpack. We chose 6 streams crossed by state roads, which were plowed to bare pavement in winter, then salted. Four of our study streams are in Franklin County, in the Saranac River Watershed— part of the larger St. Lawrence River Watershed—and are located within 9.5 km (6 mi) of each other (Fig. 1). Two of these streams are crossed by State Route 3 (streams 3-1 and 3-2) and 2 of the study streams in this watershed are crossed by County Road 45, also called Panther Mountain Road (P-1 and P-2). We collected water samples from the 8 study sites on these 4 streams monthly throughout 2007 and 2008 and collected macroinvertebrate samples monthly from May through September (sampling details below) Northeastern Naturalist 462 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 We redesigned the study in 2009 by eliminating 1 stream, adding 14 streams, and sampling in April, June, and August only. The impetus for this additional sampling period was twofold: (1) we had detected fairly low chloride levels during our previous water sampling at the 2 streams crossed by State Route 3 (see Results), and we wondered if streams crossed by a different state road might yield water samples with higher chloride concentrations; (2) one of the 2 streams crossed by a county road, P-2, had shown anomalous chloride spikes in some months, whereas P-1 had shown a very stable chloride profile. We wanted to know whether other Figure 1. Watersheds containing study streams crossed by state and county roads in the Adirondack Park. Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 463 streams crossed by county roads also showed chloride spikes. The number of study streams had become unwieldy; thus, we decided not to sample the stable stream (P-1) in 2009. Six of the streams added in 2009 are located in St. Lawrence County in the Raquette River Watershed—also part of the larger St. Lawrence River Watershed— and are within 4 km (2.5 mi) of each other. The 6 streams are crossed by 2 county roads: MA-1, at the foot of Mount Arab, is crossed by County Road 62, and streams MA-2–MA-6 are crossed by Mount Arab Road (Fig. 1). The other 8 streams added in 2009 are located in the Upper Hudson River Watershed. Four of the streams are crossed by State Route 28 (streams 28-1–28-4) and are within 6.4 km (4 mi) of each other (in Hamilton County). The other 4 streams in this watershed are crossed by Goodnow Flow Road (streams G-1–G-4) and are within 2 km (1.2 mi) of each other (in Essex County). Goodnow Flow Road is a county road that receives no salt in winter. US Geological Survey staff have sampled groundwater in these watersheds for Cl- and other contaminants. In the St. Lawrence River Watershed, results from 20 test wells showed a median value for Cl- concentration of 6.62 mg/L. Chloride levels in wells in sand and gravel aquifers varied from 3.72 mg/L to 206 mg/L, while wells in bedrock varied from 0.36 mg/L to 58.4 mg/L chloride (Nystrom 2012). Values for the Upper Hudson River Watershed (20 wells) had a median value of 12.4 mg/L chloride and varied from 2.23 mg/L to 105 mg/L Cl- for sand and gravel wells, and 0.71 mg/L to 1440 mg/L Cl- for wells in bedrock. The latter very high value was an extreme outlier (in a Hamilton County well ~27.4 km [17 mi] SSW of the Route 28 study sites) caused by a municipal road salt stockpile that had formerly been in the area (Scott and Nystrom 2014). The surficial geology at all the study sites was glacial till, with the exception of site 28-1, in the Upper Hudson River Watershed, which had alluvial inwash (Cadwell and Pair 1991). The major bedrock type at all study stream locations is gneiss (metamorphic rock), with the exception of the Upper Hudson River Watershed at streams 28-1 and 28-2, where the bedrock is undivided metasedimentary rock and related magmatite (Isachsen and Fisher 1970). The Upper Hudson River Watershed at stream 28-3 is interlayered metasedimentary rock and granitic, charnockitic, mangeritic, or syenitic gneiss. At streams 28-4 and the Goodnow Flow streams G-1– G-4, the Upper Hudson River Watershed bedrock is charnockite, mangerite, pryoxene-(hornblende)- quartz syenitic gneiss. All of the Saranac River Watershed streams are on bedrock composed of metanorthosite and anorthositic gneiss. The Raquette River Watershed stream at the foot of Mt. Arab, MA-1, is crossed by County Route 62, and rests on interlayered amphibolite and granitic, charnockitic, mangeritic, or syenitic gneiss. The other Raquette River Watershed streams are located in ascending order up the slope of Mt. Arab. Streams MA-2 and MA-3 rest on biotite and/or hornblende granitic gneiss, locally pyroxenic; commonly with subordinate leucogranitic gneiss, biotite-quartz-plagioclase gneiss, other metasemdimentary rocks, amphibolite, migmatite. Farther up the mountain, streams MA-4–MA-6 rest on bedrock of mangerite, pyroxene syenitic gneiss, pyroxene-(hornblende) syenitic gneiss; mesoperthite is common (Isachsen and Fisher 1970). Northeastern Naturalist 464 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 We also sampled a small human-made wetland for high road-salt input under conditions of low drainage. “Ampersand Slough” begins at the outflow of a plastic drainage pipe that is buried beneath Route 3; the wetland is parallel to Route 3 and ~1 m downgradient from the roadway for about 20 m before it becomes an intermittent stream traveling down to Middle Saranac Lake. This stream only reaches the lake during high flow; at other times the wetland has no real outflow (A. Tiwari, pers. observ.). All study streams were forested, except for the strips next to the road that were maintained by highway departments. No buildings were visible from 36 of the 38 stream sites. Structures (a house, a driveway) were visible downstream from the downstream sites on P-1 and P-2. Methods Water quality We collected water samples by the grab method in deionized water-rinsed, 500-ml wide-mouth polyethylene bottles (Wildlife Supply Company, Yulee, FL), 1 bottle per sample. We analyzed the water samples within 15 d at the Laboratory for Marine and Estuarine Research, Lehman College of the City University of New York. We divided each sample into 5 subsamples and titrated each subsample for Cl- concentration by the silver nitrate method, using a Hach digital titrator, model 16900 (Hach Company, Loveland, CO; Yoder 1919). Other studies examining road salt have also employed this titrator and method (Williams et al. 1999). We report the mean of these 5 subsamples. Ephemeroptera, Plecoptera, Trichoptera We collected a benthic sample at every site each time we took a water sample. We employed a Surber sampler, a net with a 500-μm mesh size and an attached frame enclosing ~30 cm2 (~1 ft2) to collect benthic invertebrates. We used the Surber design with an attached cod bucket (Wildlife Supply Company). We agitated substrate within the attached frame by hand, and the water current swept disturbed macroinvertebrates into the net. We collected, placed in a labeled jar at the site, and preserved in 75% ethanol within a few hours 1 Surber-net’s worth of substrate at each study site per sampling day. We later separated invertebrates from sand and pebbles by floatation with saturated calcium chloride and picked out leaf and twig debris. We identified EPT taxa to genus mainly using keys in Merritt and Cummins (1978) and Peckarsky et al. 1990, but we consulted additional keys for Plecoptera (Stewart and Stark 1988), and Trichoptera (Wiggins 1996). We were also aided by images of Ephemeroptera in Schweibert (2007). We aggregated and regressed on stream-chloride level EPT data by order and month (number of individuals collected per month, number of genera collected per month). We combined all sampling years by month. We compared Shannon–Weiner diversity across all stream sites for upstream versus downstream numbers of EPT via the Shannon diversity t-test (Hutcheson 1970) in the Past V3 program (Hammer et al. 2001). We compared upstream versus downstream numbers separately by Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 465 order (Ephemeroptera, Plecoptera, or Trichoptera) for total individuals and for total genera, for August, and, separately, for June in streams crossed by state roads in all sampling years. We conducted data analyses in Microsoft Excel and PAST Paleontological Statistics (Hammer et al. 2001). Forest-tree assemblage To determine the tree species present and tree species recruitment, we selected forest stands on either side of state roads that cross study streams. To differentiate the levels of salt exposure for stretches of forest land, we compared the mean Cl- content of water at stream sites in April, June, and August of 2009. We employed point-centered–quarter analysis (Cottam and Curtis 1956, Mitchell 2007) to create 10 transects on either side of 2 state roads; transects were 30 m from the road. We set up 8 transects on Route 28, which crosses 4 of the study streams. Each transect was parallel to the road and centered on 1 of the study streams, with 15 randomly obtained points on either side of the stream. When using the point-centered quarter method (Cottam and Curtis 1956, Mitchell 2007), at each point a line is drawn perpendicular to the transect, creating four “quarters”, or areas for sampling, and the nearest tree in each quarter is measured and identified. We measured the 4 trees found at each sampling point (diameter ≥1 cm) at 130 cm above ground level rather than at breast height (Brokaw and Thompson 2000, Mitchell 2007). On Route 28, Stream 1295 is 3.06 km (1.9 mi) north of Stream 1313, which is 1.77 km (1.1 mi) north of Stream 1324, which is 1.45 km (0.9 mi) north of Stream 1333. Transects along Route 28 were ~260–300 m long. We undertook a slightly different process on Route 3, where the 2 study streams it crosses lie close enough to each other that it was possible to start a transect 30 m beyond 1 of these streams, proceed parallel to the road, cross that stream, keep going, eventually cross the second study stream, and end the transect 30 m beyond the second stream. We created transects on either side on Route 3 and encompassing both streams (average of 48 points per transect). We estimated the length of this transect given the fact that the 2 streams are 1.13 km (0.7 mi) apart, according to roadside mile markers. Therefore, on the upstream or downstream side, each transect was ~1186 m long and 30 m from the road. We followed Cottam and Curtis (1956) and Mitchell (2007) to analyze data from the point-centered–quarter method (PCQM) sampling. If a tree had multiple trunks at 130 cm from the ground, as often happens with Alnus incana (L.) Moench (Speckled Alder), we computed the basal area for each trunk and summed the results (Mitchell 2007). We sampled 10 transects: 1 above and 1 below Route 3 and 4 above and 4 below Route 28 and ran 2 separate analyses on the data from each transect. After the first analysis of a transect’s data, we created a new database that included the 1st-quartile circumference (the quartile of smallest trees by circumference) and performed all PCQM calculations on this quartile to obtain a measure of recruitment. For data visualization, we multiplied the Shannon–Weiner index of each transect by 100. Northeastern Naturalist 466 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 Soil cations We took soil samples at sites along the transects. There were 8 soil sampling sites on Route 28, where each of the 4 transects was centered on a study stream; soil-sampling sites were 30 m in from the road, and 30 m from either side of a study stream. On Route 9, where the transect went from one stream to another, there were 2 soil-sampling sites: 30 m in from the road, and 30 m in from the inside transect edge of each stream. We collected soil samples and replicates at 2 depths: 0–4 cm and 15–19 cm. We collected a total of 80 soil samples (10 sampling sites, 4 samples at 2 depths per site). We analyzed samples for plant macro- and micronutrients at the Cornell Nutrient Analysis Laboratory by the 1060 Modified Morgan soil fertility test package (CNAL 2015). Analyses compared 10 soil cations for concentrations above versus below the road, or identified correlations with other factors, such as chloride level in the nearest stream. We compared soil-cation concentrations between each sample and its replicate, and between the 0–4-cm depth and the 15–19-cm depth. For each depth, we compared via 2-factor ANOVA soil-cation concentrations at sites above vs. below the road for all sites. We also compared the 2 highest-chloride transects to the 2 lowestchloride transects by single-factor ANOVA and examined cation concentrations for correlations with soil sodium or with chloride in the stream on either side of which we had taken the soil samples. Concentrations were untransformed. For ANOVA, we assumed that any deviations from normality were acceptable due to a balanced design. For pairwise comparisons, we tested unsigned differences between pairs for normality and employed either the t-test or the Wilcoxon test, as appropriate. We defined statistical significance at the α = 0.05 level. Results Water quality Water testing confirmed 2 expected patterns in chloride concentrations, but failed to confirm 2 others. Streams crossed by state roads had higher mean chloride loads than streams crossed by county roads (Fig. 2). The means of upstream and downstream sample pairs from streams crossed by state versus county roads in 2008 were significantly different (state mean = 3.90 ± 0.77, county mean = 3.28 ± 0.19; F1,46 = 4.05, P ˂ 0.001), as were those in 2009 (state mean = 28.90 = ± 45.22, county mean = 3.14 ± 0.59; F1,49 = 4.04, P ˂ 0.01). Chloride loads were higher downstream from a state road as opposed to downstream from a county road. Downstream sites on streams crossed by a state road in 2008 had a mean chloride concentration of 5.04 mg/L ± 1.56, while downstream sites on streams crossed by a county road were significantly different at 3.77 mg/L ± 1.75 (F1,46 = 4.05, P = 0.01). In 2009, downstream sites on streams crossed by a state road had a mean chloride concentration of 34.52 mg/L ± 50.45 (Fig. 3). Downstream sites on streams crossed by a county road in that year had a significantly Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 467 different mean chloride concentration of 3.32 mg/L ± 1.02 (F1,49 = 4.04, P = 0.001) (Fig. 4). Chloride loads in streams were not consistently significantly higher downstream from a road as opposed to upstream from either a state or county road. One upstream–downstream comparison was significant, but a test with additional streams another year was not significant, even though in both cases the streams had been crossed by state roads. The mean Cl- concentrations from upstream sites crossed by a state road in 2008 (2.76 mg/L ± 0.24) were significantly different from those at downstream sites on the same streams (5.04 mg/L ± 1.56; F1,46 = 4.05, P ˂ 0.001). However, in 2009 the mean Cl- concentrations from upstream sites on streams crossed by state roads (23.28 mg/L ± 40.25) were not significantly different from those at downstream sites at the same streams (34.52 mg/L ± 50.46; F1,34 = 4.13, P ˃ 0.05). Similarly, the mean Cl- concentrations from upstream sites on streams crossed by a county road in 2007 and 2008 (2.80 mg/L ± 0.34) were significantly different from those at downstream sites on the same streams (4.39 mg/L ± 3.17; F1,46 = 4.05, P ˂ 0.001). In 2009, the mean Cl- concentration for the upstream sites of streams Figure 2. Chloride concentrations in stream sites upstream (_U) and downstream (_D) from state and county roads, 2008. (A) Stream 3-1 and (B) stream 3-2 crossed by State Route 3. (C) Stream P-1 and (D) stream P-2 crossed by County Route 45, Panther Mountain Road. Northeastern Naturalist 468 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 crossed by county roads was 2.96 mg/L ± 0.41. Downstream sections on those same streams had a mean Cl- concentration of 3.32 mg/L ± 1.02. Upstream sections on these streams were not significantly different from downstream sections (F1,64 = 3.99, P ˃ 0.05). The chloride load was not higher in streams crossed by a county road that received a sand and salt mixture as opposed to a stream crossed by a county road that received only sand. In 2009 the downstream sites of 4 stream sites receiving Figure 3. Chloride concentrations in streams crossed by state roads, 2009. Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 469 no salt had a mean chloride-ion concentration of 3.44 mg/L ± 1.17, whereas the downstream sites of 6 streams crossed by a county road receiving the traditional sand/salt mixture had a mean chloride-ion concentration of 2.96 mg/L ± 0.36. There was no significant difference between these sites (P ˃ 0.05). Ephemeroptera, Plecoptera, Trichoptera We collected 1259 EPT specimens at study sites, comprising 292 individual Ephemeroptera in 14 genera, 495 Plecoptera in 9 genera, and 469 Trichoptera in 28 genera. Linear regression did not reveal any negative relationship between the Figure 4. Chloride concentrations in streams crossed by county roads, 2009. Northeastern Naturalist 470 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 number of individual Ephemeroptera, Plecoptera, or Trichoptera collected, or number of genera collected in any sampling month across all sampling years. When we employed the Hutcheson diversity t-test to compare Shannon–Weiner diversities for upstream vs. downstream Ephemeroptera, Plecoptera, or Trichoptera, neither the number of individuals nor the number of genera differed (P ˃ 0.05). We did not observe Ephemeroptera if streamwater chloride concentrations were ≥154 mg/L. We collected 3 individuals of the Ephemeropteran genus Habrophlebiodes (mayflies) at a stream site on Route 28 in August 2009 where the chloride level was 153.4 mg/L; no Ephemeroptera were detected at higher chloride concentrations or at the next 2 lowest chloride concentrations, 115.8 mg/L and 76 mg/L. The next lowest chloride concentration at which Ephemeroptera we detected was 33.04 mg/L, also on Route 28, at which we collected 4 genera. At Ampersand Figure 5. Chloride concentrations at Ampersand Slough wetland 2008 and 2009. Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 471 Slough (Fig. 5), chloride concentrations of our samples varied from 215 mg/L to 398 mg/L Cl- on days a collection was made. We did not observe Ephemeroptera at Ampersand Slough, but we collected 135 Plecoptera individuals in 16 genera and 65 Trichoptera individuals in 22 genera at that site. Tree-species assemblage Point–centered–quarter analysis of forest quadrats centered on study streams yielded importance values for the tree species encountered. The importance values for the quartile of smallest trees by circumference represents tree-species recruitment. Trees species showing the highest importance value per transect, either for the whole database of that transect or for its lowest quartile by circumference, indicate which species dominate that forest and whether that dominance is in the process of changing over time. For both whole databases and lowest quartiles, Fagus grandifolia Ehrh. (American Beech), Tsuga canadensis (L.) Carrière (Eastern Hemlock), and Picea rubens Sarg. (Red Spruce) were well-represented in transects centered on streams with lower chloride content. Abies balsamea (L.) Mill. (Balsam Fir) was also present in lower-chloride transects, notably in a transect that could be characterized as wetland. The proportion of Balsam Fir in transects rose along with the Clcontent of study streams. This effect was most marked in the lowest quartile, as shown in Figure 6. Stream chloride content is a convenient measure of the roadsalt runoff along a transect. Figure 6 shows that for the recruiting tree species, road-salt runoff onto forest soil is associated with increasing importance value of Balsam Fir and decreasing Shannon–Weiner diversity. Figure 6. Shannon–Weiner diversity index (multiplied by 100) plus Importance Values of Abies balsamea (Balsam Fir) among young trees in 10 transects exposed to road-salt runoff and chloride concentration (mg/l) of local streams. Northeastern Naturalist 472 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 Soil cations Plant-nutrient cation concentrations of replicate sample pairs did not differ (n = 160, Wilcoxon test for the 0–4-cm depth: P ˃ 0.05). Cations at the 0–4-cm depth were present at greater concentrations than at the 15–19-cm depth (n = 360; Wilcoxon test P ˂ 0.001). Sodium was present at significantly higher concentrations below the road than above the road at both soil depths (mean above the road at 0–4 cm = 134.70 mg/Kg ± 53.44, mean below the road = 188.81 mg/Kg ± 64.90, Wilcoxon test: P ˂ 0.01; mean above the road at 15–19 cm = 49.83 mg/Kg ± 23.06, mean below the road = 90.96 mg/Kg ± 75.99, Wilcoxon test: P ˂ 0.05). The presence of plant-nutrient cation concentrations at significantly lower levels below the road as opposed to above the road, would suggest that excess sodium from road salt has depleted these nutrient cations. However, ANOVAs showed that nutrient concentrations were not lower at sites below a road at either the 0–4-cm or 15–19-cm depth. Another indication of road salt affecting cation concentrations would be a negative correlation with chloride or sodium. There were no negative correlations between stream chloride or soil sodium and any soil cation when we compared all samples for any particular cation against local stream chloride or sodium in that sample. Similarly, when we tested samples from the 0–4-cm depth only for each cation, there was no negative correlation between any cation and chloride or sodium. Finally, single-factor ANOVAs to compare cation concentrations at the 2 highest-chloride transects with cations in the 2 lowest-chloride transects (all 4 transects from Route 28) showed no difference for any cation besides sodium (F1,14 = 4.60, P < 0.05). Discussion Water quality Road salt will continue to be of interest in coming decades, as salt-ion inputs to groundwater are increasingly seen in surface waters. Groundwater initially acts as a sink for both sodium and Cl-, but then releases these ions as their concentration in groundwater increases (Howard and Haynes 1993). This process creates a lag effect, in which Cl- deposited years earlier appears in increasing amounts in springs and in streams by baseflow recharge (Howard and Haynes 1993, Kelly et al. 2008). It is important, therefore, to determine whether differences in road de-icing strategy, such as those employed in the Adirondacks, make a measurable difference in road-salt runoff. As we expected, our results showed that the local practice of driving on snow-packed roads, with sand and a little salt, delivers significantly less salt to local streams than the bare-roads policy of salting state roads. In general, the Cl- concentration in streams crossed by county roads was very low, and we found no statistically significant differences whether the streams were crossed by roads treated with the traditional sand/salt mixture or sand alone. An August spike in chloride concentration in one of the unsalted streams occurred when the stream had Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 473 almost dried up, thus concentrating any ions present, which could have come from groundwater-chloride levels. Salt levels in a stream no longer receiving road salt inputs would likely reflect past road salting. One site on a county road receiving the traditional sand/salt mixture had higher than expected chloride spikes, possibly as a result of a clearing along a powerline right-of-way that crosses the creek below this downstream site, which may receive additional salting. Salt that is applied below a test site cannot be assumed to be carried away in one direction, but rather should be seen as an input to the complex system of local groundwater. Similarly, for state roads, the expected pattern of higher levels of chloride downstream was not reliable—high chloride levels at upstream sites on 2 streams crossed by State Route 28 produced large standard deviations and lack of a significant difference in 2009 analyses comparing upstream vs. downstream sites on Route 28. These high upstream levels were likely due to baseflow recharge of chloridecontaminated groundwater. On Route 28, streams 1313 and 1333 both had high upstream as well as downstream chloride concentrations. The streams are not adjacent; stream 1324 is between them and does not show this pattern. It is likely that the local hydraulic gradient directs salt-bearing groundwater to the upstream portions of streams 1313 and 1324. Groundwater is known to carry salt contamination from road de-icing (Rosenberry et al. 1999). Snow plows push salt off the road on both sides all along the road length. The snow melts in the spring, becomes part of the local groundwater, and moves along the local hydraulic gradient. It is probable that on most roads, the higher-gradient side directs groundwater underneath the road to the lower gradient side, but that in some locations, more compacted soil beneath the road or another aspect of local hydraulic gradient keeps the water on the upgradient side long enough to join a stream on the upstream side of the road. Howard and Hayes (1993) showed that 55% of the chloride input per year in a metropolitan Toronto basin was being stored in groundwater. Thus, the expectation that stream chloride concentrations will always be higher downstream from a road cannot be sustained because it arises from visual bias. Such an assumption is based on the surface orientation of stream and road, whereas ground water has its own chloride load and its own flow patterns not perceived from the surface view. When considering the effect of road salting on streams or lakes, it is important to think of surface waters as part of a connected system, the major part of which is not seen. Similarly, the expectation that streams crossed by a county road that receives only sand as a winter treatment would contain less road salt than streams crossed by county roads receiving the traditional sand/salt mixture could not be supported. That hypothesis does not take into account the groundwater burden of Cl-, built up in past years. Ephemeroptera, Plecoptera, Trichoptera Just as stream-chloride levels are not reliably lower upstream from a road as opposed to downstream, Shannon–Weiner diversity of EPT taxa is not greater upstream versus downstream. We suggest that the upstream versus downstream Northeastern Naturalist 474 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 dichotomy is a gross oversimplification of the interconnected above- and belowground water system of a stream crossed by a road. Our finding that EPT Shannon–Weiner diversity is the same upstream or downstream agrees with those of Demers (1992), although in that study EPT collections in each of 4 streams were analyzed separately. It is unfortunate that Demers stated that “this was significantly different at P = 0.08” (Demers 1992). The manner of reporting apparently caused confusion because this study (erroneously referred to as Demers and Sage 1990) was cited as having “found severe impacts to macroinvertebrate species attributed to chlorides” (NYSDEC 2015). We feel that it is preferable not to strain or overstate results, and that results showing no EPT effect at relatively low chloride levels are useful data, which can be contrasted with studies on more heavily salt-polluted waterways. First-order streams (Strahler 1957) crossed by roads can be expected to have lower chloride loads than higher-order streams which may carry pollutant loads from multiple road crossings, and EPT numbers in higher-order streams may be significantly affected at those higher chloride levels. Future research could be discouraged by overstating results from lower-order streams. The absence of Ephemeroptera at Cl- concentrations above 154 mg/L suggests that research should focus on areas of higher Cl- concentration, such as higher-order streams, or low-drainage areas that parallel roadways, such as Ampersand Slough. In that human-made wetland, we collected abundant Plecoptera and Trichoptera at road-salt levels that excluded Ephemeroptera. Wetlands, including human-made features such as roadside ditches, typically drain slowly, and are considered by some to be the most sensitive environments to road-salt pollution (Environment Canada 2001, Tiner 2005). The Maryland Biological Stream Survey found “almost no” Ephemeroptera at chloride concentrations above 500 mg/L Cl- and concluded that Ephemeroptera are the most pollution-sensitive macroinvertebrates (Stranko 2013). It may be that Ephemeroptera are being increasingly excluded from areas where they previously occurred. Ephemeropteran populations provide food sources for fish and birds (Gray 1993, Weidel et al. 2000), but these taxa may be absent in road-salt impacted areas. Forest-tree assemblage Figure 6 shows that when road salt runoff was low (inferred from stream chloride), there was high Shannon–Weiner diversity of young trees. One exception is the first transect, 28-4-U, where there were many young American Beech trees and low diversity. American Beech thickets are common in the Adirondacks, and often follow the production of large numbers of suckers from older fungus-infected beech (Jenkins 1997). Another exception to the pattern of high diversity in low road-salt runoff areas is transect 3-1-2-D, which differed from the other 9 transects because it was in a wetland. The site had moist soil and contained Chamaecyparis thyoides (L.) Britton, Sterns & Poggenb.) (Atlantic White Cedar), an obligate wetland tree (Tiner 2005). Shannon–Weiner diversity was low in this transect because it was mostly comprised of Balsam Fir, a facultative wetland tree known to sometimes dominate wetlands Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 475 (Tiner 2005). In our study, with the exception of these transects, Shannon–Weiner diversity generally declined as the chloride content of the local streams increased. As chloride (road salt) increased, there was an increasing importance of Balsam Fir among recruiting trees (Fig. 6). As noted earlier, the wetland transect 3-1-2-D stands out as having a high Balsam Fir importance value at low stream-chloride levels. The last 4 transects in our figures are those that were exposed to considerable road-salt runoff. Balsam Fir had the highest importance value in both the whole databases and lowest quartiles for each of these transects. Balsam Fir was one of very few species that showed recruitment in these high road-salt transects. This occurrence pattern suggests that Balsam Fir is able to out-compete other tree species and that a tolerance for road salt might produce this effect. However, there are other reasons that Balsam Fir may be a superior competitor. In the Adirondacks, the species quickly colonizes gaps resulting from blow-downs and the deaths of large trees (Ketchledge 1996), and increases in areas where Red Spruce has declined (Bedison et al. 2007). Soil cations The contention that road salt caused the abundance of Balsam Fir would be better supported if plant-nutrient cations were depleted in areas with both high levels of road-salt and high abundance of Balsam Fir. However, this was not the case; we observed no pattern of soil-nutrient depletion in areas of high levels of chloride or sodium. It is possible that the high abundance of Balsam Fir at some transects was related to the regeneration pattern of Balsam Fir. Wave regeneration comes about when Balsam Fir trees at the edge of a forest gap that are exposed to the prevailing wind die due to a variety of exposure-related causes (Sprugel 1976). In time, the Balsam Fir trees behind these trees also die, creating a spreading band (or wave) of dead trees, and bands of young, regenerating trees as well. While we observed no bands of dead Balsam Fir in the vicinity of the transects, it is still possible that the areas with high density of Balsam Fir are artifacts of natural regeneration patterns of Balsam Fir. Locally high road-salt runoff may be unconnected to their appearance. Sampling along more transects, especially in areas with very high levels of road-salt runoff, could resolve this question. In summary, effects of road-salt runoff were, for the most part, undetectable in 1st-order streams, and road-salt runoff from state roads may not cause detectable changes in forest composition. Further research should concentrate on the effects of road-salt runoff on the diversity and abundance of Ephemeroptera nymphs. Forest research in road-salt runoff areas must be accompanied by soil-cation analysis. It would be very desirable to see which tree species are recruiting in areas of high levels of road-salt runoff where there is also soil nutrient depletion. Acknowledgments We thank Barabara Warkentine of SUNY Maritime College, Craig Milewski of Paul Smith’s College in the Adirondacks, Dwight Kincaid, and Amy Berkov of the Graduate Center of the City University of New York, and Richard Stalter of St. John’s University in New York City for their guidance and assistance. We also express gratitude to Adirondack Northeastern Naturalist 476 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 residents who allowed sampling on their private property. This manuscript was submitted by the senior author in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Biological Sciences at the Graduate Center of the City University of New York. Partial funding for this project was provided by a 2006 Research Grant for Doctoral Students from the Graduate Center of the City University of New York. Literature Cited Anning, D.W., and M.E. Flynn. 2014. Dissolved-solids sources, loads, yields, and concentrations in streams of the conterminous United States. US Geological Survey Scientific Investigations Report 2014-5012. 101 pp. Available online at https://pubs.usgs.gov/ sir/2014/5012/. Accessed 28 July 2018. Bedison, J.E., Johnson, S.A., Willig, S.L., Richter, and A. Moyer. 2007. Two decades of change in vegetation in Adirondack spruce–fir, northern hardwood, and pine-dominated forests. Journal of the Torrey Botanical Society 134:238–252. Brokaw, N., and J. Thompson. 2000. The H for DBH. Forest Ecology and Management 129:89–91. Cadwell, D.H., and D.L. Pair. 1991. Surficial geologic map of New York: Adirondack sheet, Map and Chart Series No. 40, scale 1:250,000. New York State Museum-Geological Survey, Albany, NY. Cornell Nutrient-analysis Laboratory (CNAL). 2015. Soil sample submission form. Available online at http://css.cornell.edu/cnal-forms/CNAL_Form_S.pdf. Accessed 28 July 2018. Cottam, G., and J.T Curtis. 1956. The use of distance measures in phytosociological sampling. Ecology 37:451–460. Demers, C.L. 1992. Effects of road-deicing salt on aquatic invertebrates in four Adirondack streams. Pp. 245–251, In F.M. D’Itri (Ed.). Chemical Deicers and the Environment. Lewis Publishers, Ann Arbor, MI. 585 pp. Demers, C.L., and R.W. Sage Jr. 1990. Effects of road-deicing salt on chloride levels in four Adirondack streams. Water, Air, and Soil Pollution 49:369–373. Environment Canada. 2001. Canadian Environmental Protection Act, 1999: Priority substance list assessment report–road salts. Environment Canada, Hull, QC, Canada. Granato, G.E., P.E. Church, and V.J. Stone. 1995. Mobilization of major and trace constituents of highway runoff in groundwater potentially caused by deicing chemical migration. Transportation Research Record 1483:92–104. Gray, L.J. 1993. Response of insectivorous birds to emerging aquatic insects in riparian habitats of a tallgrass prairie. American Midland Naturalist 129:288–300. Hammer, Ø., D.A.T. Harper, and P.D. Ryan. 2001. PAST: Paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4(1):1–9. Available online at http://palaeo-electronica.org/2001_1/past/issue1_01.htm. Accessed 17 January 2013. Howard, K.W.F., and J. Haynes. 1993. Groundwater contamination due to road de-icing chemicals: Salt-balance implications. Urban Geology 3, Geoscience Canada 20:1–8. Hutcheson, K. 1970. A test for comparing diversities based on the Shannon formula. Journal of Theoretical Biology 29:151–154. Isachsen, Y.W., and D.W. Fisher (Eds.). 1970. Geologic map of New York: Adirondack Sheet, 1:250,000, geologist James F. Davis. New York State Museum and Science Service Map and Chart Series, the University of the State of New York, Albany, NY. Northeastern Naturalist Vol. 25, No. 3 A. Tiwari and J.W. Rachlin 2018 477 Jenkins, J. 1997. Hardwood-regeneration failure in the Adirondacks: Preliminary studies of incidence and severity. Wildlife Conservation Society Working Paper Number 9, Wildlife Conservation Society, New York, NY. 60 pp. Kaushal, S.S., P.M. Groffman, G.E. Likens, K.T. Belt, W.P. Stack, V.R. Kelly, L.E. Band, and G.T. Fisher. 2005. Increased salinization of fresh water in the northeastern United States. Proceedings of the National Academy of Sciences 102:13517–13520. Kelly, V.R., G.M. Lovett, K.C. Weathers, S.E.G. Findlay, D.L. Strayer, D.J. Burns, and G. Likens. 2008. Long-term sodium-chloride retention in a rural watershed: Legacy effects of road salt on streamwater concentration. Environmental Science and Technology 42:410–415. Kelting, D.L., and C.L. Laxson. 2010. Review of effects and costs of road de-icing with recommendations for winter road management in the Adirondack Park. Adirondack Watershed Institute Report # AW12010-10, Adirondack Watershed Institute, Paul Smith’s College, Paul Smiths New York, NY. 40 pp. Ketchledge, E.H. 1996. Forests and trees of the Adirondack High Peaks Region. Adirondack Mountain Club, Lake George, NY. 171 pp. Lenat, D.R., and D.L. Penrose. 1996. History of the EPT taxa richness metric. Bulletin of the North American Benthological Society 13:305–307. Merritt, R.W., and K.W. Cummins. 1978. An Introduction to the Aquatic Insects of North America. Kendall-Hunt Publishing Company, Dubuque, IA. 439 pp. Mitchell, K. 2007. Quantitative analysis by the point-centered–quarter method. Available online at http://people.hws.edu/Mitchell/PCQM.pdf people.hws.edu/M. Accessed 17 January 2013. New York State Department of Environmental Protection (NYSDEC). 2015. Impacts of new development: Declining water quality. Chapter 2, section 2.1, In New York State Stormwater Management Design Manual. Available online at http://www.dec.ny.gov/ docs/water_pdf/swdm2015chptr02.pdf. Accessed 1 August 2016. Norris, R.H., and A. Georges. 1993. Analysis and interpretation of benthic macroinvertebrate surveys. Pp. 234–286, In D.M. Rosenberg and V.H. Resh (Eds.). Freshwater Biomonitoring and Benthic Macroinvertebrates. Chapman and Hall, New York, NY. 488 pp. Nystrom, E.A. 2012. Groundwater quality in the Delaware and St. Lawrence River basins, New York, 2010. Open-file Report 2011-1320, US Geological Survey, Reston, VA. Available online at https://pubs.er.usgs.gov/publication/ofr20111320. Accessed 21 February 2018. Peckarsky, B.L., P.R. Fraissinet, M.A. Penton, and D.J. Conklin Jr. 1990. Freshwater Macroinvertebrates of Northeastern North America. Cornell University Press, Ithaca, NY. 442 pp. Pest Management Regulatory Agency (PMRA). 2006. Proposed regulatory-decision document PRDD2006-01: Sodium chloride. Publications, Health Canada, Ottawa, ON, Canada. 37 pp. Available online at http://publications.gc.ca/site/archivee-archived. html?url=http://publications.gc.ca/collections/Collection/H113-9-2006-1E.pdf. Accessed 9 January 2013. Rosenberry, D.O., P.A. Bukaveckas, D.C. Buso, G.E. Likens, A.M. Shapiro, and T.C. Winter. 1999. Movement of road salt to a small New Hampshire lake. Water, Air, and Soil Pollution 109:179–206. Schweibert, E.G. 2007. Nymphs: The Mayflies, Major Species. The Lyons Press, Guilford, CT. 628 pp. Northeastern Naturalist 478 A. Tiwari and J.W. Rachlin 2018 Vol. 25, No. 3 Scott, T.M., and E.A. Nystrom. 2014. Groundwater quality in the Upper Hudson River Basin, New York, 2012. Open-file Report 2014-1084. US Geological Survey, Reston, VA. Vailable online at https://pubs.er.usgs.gov/publication/ofr20141084. Accessed 21 February 2018. Sprugel, D.G. 1976. Dynamic structure of wave-regenerated Abies balsamea forests in the northeastern United States. Journal of Ecology 64:889–911. Stewart, K.W., and B.P. Stark. 1988. Nymphs of North American Stonefly Genera (Plecoptera). Entomological Society of America, Lanham, MD. 460 pp. Strahler, A.N. 1957. Quantitative analysis of watershed geomorphology. Transactions, American Geophysical Union 35:913–920. Stranko, S., R. Bourquin, J. Zimmerman, M. Kashiwagi, M. McGinty, and R. Klauda. 2013. Do road salts cause environmental impacts? Maryland Department of Natural Resources, Monitoring and Non-Tidal Assessment Division. Available online at http:// dnr.maryland.gov/streams/Publications/RoadSalt2013.pdf. Accessed 20 August 2017. Tiner, R.W. 2005. In Search of Swampland: A Wetland Sourcebook and Field Guide, 2nd Edition. Rutgers University Press, New Brunswick, NJ. 330 pp. US Geological Survey (USGS). 2011. USGS Mineral Yearbook 2010: Salt. States Department of the Interior, Washington DC. 23 pp. Wallace, J.B, J.W. Grubach, and M.R. Whiles. 1996. Biotic indices and stream ecosystem processes: Results from an experimental study. Ecological Applications 6:140–151. Weidel B.C., D.C. Josephson, and C.C. Kruger. 2000. Diet and prey selection of naturalized Smallmouth Bass in an oligotrophic Adirondack lake. Journal of Freshwater Ecology 15:411–420. Wiggins, G.B. 1996. Larvae of the North American Caddisfly Genera. University of Toronto Press, Toronto, ON, Canada. 457 pp. Williams, D.D., N.E. Williams, and Y. Cao. 1999. Road-salt contamination of groundwater in a major metropolitan area and development of a biological index to monitor its impact. Water Research 34:127–138. Yoder, L. 1919. Adaptation of the Mohr volumetric method to general determinations of chlorine. Industrial and Engineering Chemistry 11:755.