Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 We incorporate high-resolution time-series data to calculate the total amount of dissolved organic carbon (DOC) and dissolved organic nitrogen (DON) transported during Hurricane Irene in Esopus Creek in New York (August 2011). During this 200-yr event the Esopus Creek experienced a 330-fold discharge increase and a 4-fold increase in concentration, resulting in the export of roughly 43% and 31% of its average annual DOC and DON fluxes, respectively, in just 5 days. The source of this large dissolved organic matter (DOM) flux also shifted during its course and showed an increased contribution of aromatic organic matter. We conclude that more frequent large events due to climate change will increase the export of terrigenous dissolved organic matter, and potentially impact the water quality and biogeochemistry of lakes and coastal systems. In addition, we show that the use of conventional models for extreme events lead to erroneous flux calculations, unless supported by high resolution data collected during the events.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Carbon transport via inland waters has been receiving increasing attention as its role in linking land, sea, and atmosphere is significantly more complex and influential than previously believed [Battin et al., 2009]. Long regarded as a passive conduit between land and ocean, inland water systems not only transport carbon from land to sea, but also degas carbon into the atmosphere [Butman and Raymond, 2011], store carbon internally [Cole et al., 2007], and support the food web of aquatic environment by fueling microbial metabolism [Fisher and Likens, 1973]. Export of dissolved organic carbon is also a special concern for drinking water treatment since it can influence water quality [Kaplan et al., 2006].
 The quantification of inland water carbon fluxes often relies on simple periodic sampling and modeling approaches that do not always account for biogeochemical variation during precipitation events. Precipitation events serve as a major driver for export of labile organic carbon and organic-bound nutrients [McClain et al., 2003], and a number of studies confirm this strong link across different systems [Hinton et al., 1997, 1998; Inamdar et al., 2004; Clark et al., 2007; Eimers et al., 2008]. A meta-study of forested watersheds in the Northeast, for example, concluded that 86% of annual DOC is exported during precipitation events, and large precipitation events (>1.38 cm day−1) are responsible for 57% of annual DOC flux [Raymond and Saiers, 2010]. Despite this importance of intense precipitation events in controlling carbon fluxes, only a few studies examined carbon fluxes during very large precipitation events [Avery et al., 2004].
 In the United States, studies have demonstrated both an increase in precipitation [Groisman et al., 2004] and an increase in intense precipitation events [Groisman et al., 2005], particularly in the Northeast. For the eastern United States, the scientific community acknowledges that hurricane intensity has increased over the past few decades [Emanuel, 2005] and the frequency of intense hurricanes may double in the future [Bender et al., 2010]. Such frequent occurrence of intense events coupled with the importance of events in exporting DOC could translate to increased transport of terrigenous organic matter to lakes, reservoirs, and coastal waters, potentially altering their biogeochemical cycles. In light of this critical link between climate change, intense precipitation events, and organic matter transport, this paper quantifies the amount of dissolved organic carbon and dissolved organic nitrogen transported during Hurricane Irene in Catskill Mountain, NY, and elucidates the importance of extreme events in lateral transport dynamics. In addition, we also test the accuracy of conventional models in estimating DOC and DON fluxes during the hurricane to assess how data availability and model selection influence the resulting outcomes.
2. Materials and Methods
2.1. Study Area and Hurricane Irene
 The watershed examined for this study is situated in Allaben, NY where Esopus Creek drains 16,500 ha of the Catskill Mountain (Figure 1). The watershed is predominantly composed of forest cover [Fry et al., 2011] with deciduous tree species such as sugar maple (Acer saccharum), beech (Fagus grandifolia), and red oak (Quercus borealis) [McIntosh, 1972]. The general geology of the watershed can be described as layers of glacial till overlaid on top of bedrock composed of sandstones and shales [Stoddard and Murdoch, 1991]. The climate of the region can be characterized as humid continental with cool summers and cold winters, and relatively uniform precipitation throughout the year (NRCC, available at http://www.nrcc.cornell.edu). Data from Slide Mountain weather station, located within the watershed boundary, indicate that the region receives an average annual precipitation of 160 cm with an average temperature of 5.5°C (NOAA SIS, available at http://www.ncdc.noaa.gov). As a primary source of drinking water for New York City, much of the Catskill Mountain area is designated as forest preserve, and roughly 62.5% of the study watershed is protected by the state [NY DEC, available at www.dec.ny.gov]. The precipitation pattern in this watershed is a major concern for New York City's water supply and is closely monitored since high stream flows caused by large events are directly associated with increased nutrient, sediment, and pollutant transport to the reservoir [New York City Department of Environmental Protection (NYC DEP), 2011a].
 The sampling point for this study is located on Esopus Creek where a USGS gauging station has been monitoring its discharge since 1963 (USGS Esopus Creek at Allaben, data available at http://waterdata.usgs.gov). Average annual total discharge recorded by the gauging station is 137 million m3 yr−1 (SD = ±38.9) over the recording period (1964–2010). Discharge of the stream shows strong seasonality with high flows during spring snowmelt and significantly lower values during the summer months (Figure 2, grey dashed line). The hydrograph for year 2011 reflects the large snowfall of the previous winter and the effect of Hurricane Irene (Figure 2, bold line).
 Hurricane Irene was the first hurricane of the 2011 Atlantic hurricane season that affected various areas in the Caribbean, nearly all states of eastern U.S., and the eastern part of Canada. It was downgraded to a tropical storm on 28 August 2011 as it entered the state of New Jersey. During three days of its passing through the Catskill Mountains (27–29 August 2011), Irene produced massive precipitation of at least 29.3 cm (http://www.ncdc.noaa.gov) and increased the discharge of Esopus Creek by roughly 330-fold at the sampling location (2.5 m3 s−1 to 829 m3 s−1, Figure 3). Both the daily precipitation for 28 August (21.9 cm d−1) and the peak discharge recorded on that morning are the highest values recorded in this watershed since their respective recording programs began in 1954 and 1963 (http://waterdata.usgs.gov; http://www.ncdc.noaa.gov). Another stream discharge measurement record from a downstream location also indicates that Hurricane Irene was the largest event for the Catskills since the record started there in 1931 (Coldbrook, NY, data available at http://waterdata.usgs.gov). Statistical inference using the recorded precipitation and discharge indicates Hurricane Irene was at least a 200 year storm (NRCC & NRCS, available at http://precip.eas.cornell.edu, also see auxiliary material).
2.2. Event Sampling
 Hurricane Irene was one of six storm events that were analyzed for this watershed from the period of April 2011 to November 2011 (see auxiliary materialfor sampling conditions). For each event, a high-resolution sample series was collected in sterilized glass bottles using an auto-sampler (Sigma 900 MAX, Hach) that was programmed to collect samples at 2 to 8 hour intervals. The sampler was filled with ice prior and throughout each event, and the samples were generally retrieved within 48 hours. To minimize contamination of the samples, the intake tube was programmed to be rinsed multiple times before an actual sample is collected. For Hurricane Irene, the sampling program was initiated at 5 PM on 27 August and ended at 10 PM on 30 August, collecting 350 ml of the stream water at 3.5 hours intervals. An additional base flow sample was collected in the afternoon of 31 Aug, and resulted in a total sample number of 23 for the duration of 95 hours (Figure 3). In addition to the 23 hurricane samples, we collected 91 samples during the other 5 events between April and November of 2011. All samples were collected in sterilized glass bottles and kept in ice until filtered at 0.2 μm, and the filtered samples were kept at 4°C until further analysis. All measurements were conducted within a week of collection time. In addition to the samples collected by this study, water chemistry data (25 [DOC] and 43 [DON]) from USGS for the period 1993–2002 were incorporated into our analysis (USGS NWIS, available at http://waterdata.usgs.gov).
2.3. DOC, DON, and SUVA Analysis
 All 114 samples collected throughout the year 2011 were analyzed for dissolved organic carbon concentration ([DOC]) and total dissolved nitrogen concentration ([TDN]) using a Shimadzu TOC-V analyzer. To calculate the concentration of dissolved organic nitrogen ([DON]) for samples collected during Hurricane Irene, inorganic nitrogen concentrations (NO2−, NO3−, NH3) were measured with a flow analyzer (Astoria-Pacific 2-channel Flow Analyzer, EPA Method 353.2) and a spectrophotometer (Hach DR2700, Hach method 8155), then subtracted from [TDN] of each sample. All measurements were done in duplicate.
 Dissolved organic matter in streams undergoes both quantitative and qualitative change during precipitation events. As rising water level transports more terrestrial organic matter from soil surface, DOM concentration rises and its composition shifts to reflect its terrestrial origin [Hood et al., 2006; Vidon et al., 2008]. Ultraviolet absorbance (UVA) at 254 nm normalized by [DOC], also known as SUVA, has a strong linear correlation with aromatic carbon content in water, and can be efficiently used as an index for relative amount of humic substances [Weishaar et al., 2003]. All samples collected from Esopus Creek during the study period were analyzed for SUVA using a Beckman DU-530.
2.4. Flux Calculation
 To estimate average annual fluxes of DOC and DON in the study watershed for the 18 year period of water chemistry record, we combined our 114 samples with the existing USGS water chemistry data (43 samples). Annual fluxes were calculated with daily average [DOC], [DON], and daily average discharge rate using LoadRunner program. LoadRunner is designed to automate LoadEST [Runkel et al., 2004] runs to find a best fit model that expresses daily flux as a function of daily discharge, and extrapolate this model to all available daily discharge data. The resulting daily fluxes were then summed to yield annual total fluxes for the period of 1993 to 2011.
 The magnitude of DOC and DON exported by Hurricane Irene was calculated using multiple approaches to reflect the importance of sampling during events for accurate flux estimation. All three models examined the period of 27 to 31 August (Table 1). Model 1 interpolated between the 23 [DOC] and [DON] measurements made during Irene to assign a concentration for each fifteen minute interval flow measurement (Figure 3, solid black lines). Although simplistic, the availability of high resolution data allows this model to provide the most accurate estimate and can serve as a reference to other models. Model 2 and 3 utilized all available [DOC] and [DON] values except the samples collected during Hurricane Irene to examine the accuracy of using historic data for flux estimation. Model 2 utilized LoadRunner to calculate daily fluxes based on average daily discharge, and Model 3 was based on a log-linear regression that expresses flux as a function of discharge and incorporated all sub-daily measurements. (Model 3: dashed lines inFigure 3; see auxiliary material for the detailed explanation of all models.)
Table 1. DOC, DON Flux (in t C and t N) During Hurricane Irene (27–31 August, 2011)a
DOC Flux (t C)
Percent Annual Average (t C)
DON Flux (t N)
Percent Annual Average (t N)
See auxiliary material for detailed description of the models. Average annual DOC flux (1993–2010): 221 t C yr−1 (SD: ±67.0); average annual DON flux (1993–2010): 17.8 t N yr−1(SD: ±7.86).
Interpolation between measured concentrations
Daily flux from LoadRunner using non-Irene concentrations (n = 51, 115; R2 = 0.94, 0.98 over 19 year period for DOC and DON, respectively)
Ln(Flux) = f(Ln(discharge)), utilized all sub-daily measurements (n = 69, 133; R2 = 0.97, 0.88 for DOC and DON over 19 year, respectively)
3. Results and Discussion
 During the five-day period, the watershed exported 33.1 million m3 of water or 20.0 cm of discharge (68% of the measured precipitation), which translates to 24.2% of the historic annual average water flux. Hurricane Irene also was the largest precipitation and discharge event recorded, and significantly contributed to making year 2011 the wettest year for the study watershed (auxiliary material, Figure 3).
 Surprisingly, during this historic hydrologic event, concentrations of both DOC and DON did not dilute (Figure 3, solid black lines). Interpolating between our high resolution DOM measurements provides DOC flux of 102.5 t C and DON flux of 5.59 t N for Hurricane Irene (Model 1; Table 1). These fluxes imply that during the 5 days of the event (1.4% of the year) 43.2% and 30.9% of the 18 year average annual fluxes were exported for DOC and DON, respectively. For its receiving water body the Ashokan Reservoir the estimated DOC flux from the study watershed alone is equivalent to 13% of its average standing DOC stock [NYC DEP, 2011b]. Furthermore, assuming similar loading from its entire catchment reveals that the reservoir could have received as much as 52% of its average standing DOC stock in just five days (see auxiliary material). We also estimate that Hurricane Irene resulted in the total DON loss of 33.9 mg N m−2 from the watershed (Model 1). Nitrogen is a limiting nutrient for the terrestrial landscape, and the organic nitrogen serves as a reservoir for production of inorganic nitrogen. This loss translates to only 0.5% of DON available in the organic soil layer of the watershed [Templer et al., 2005], but even such a small loss from the terrestrial system is expected to have a disproportionately larger impact on the downstream aquatic systems. Both the percent annual discharge and DOM fluxes presented in this study make Hurricane Irene one of the largest events studied for DOM export with the high temporal resolution of sample collection (also see Avery et al.  for extreme events).
 Possible sources of this DOM include DOM in the rainwater and terrestrial DOM. Using DOC concentrations of rainwater available from the literature for hurricanes and tropical storms (∼40uM) [Willey et al., 2000; Avery et al., 2004; Raymond, 2005], rainwater DOC is ∼15 t C or 14.6% of the 5-day flux. The variation in SUVA values also provides evidence for compositional change of dissolved organic matter and indicates that the rising water level was accompanied with an increase in aromatic content, most likely contributed by terrigenous humic substances from surface soil and litter (Figure 3c). Using the SUVA model by Weishaar et al. , we estimate that the percent aromatic content of DOC increased from 19.8% to 32.1% within the first 11 hours of the event. Thus this work is consistent with the mechanism of increased water table [Boyer et al., 2000; Hornberger et al., 1994] and change in the flow path of water through the soil [McDowell and Likens, 1988] resulting in higher DOC fluxes during events. The peak [DOC], [DON], and SUVA values seen from the third sample in the time series were the highest values in their respective records (Figure 3).
 This study is the first to demonstrate that the increase in DOC concentration extends even to the largest events and further underscores the importance of events to lateral DOC transfers. The impact of large DOC transfers and composition changes during such large events are important to aquatic metabolism, UV processes, and water quality of downstream ecosystems. The metabolism of inland waters and coastal oceans rely on terrestrial DOC transfers [Duarte and Prairie, 2005]. The impact of DOC transferred during these largest events, however, has received little attention. It is expected that the large DOC fluxes and high velocity associated with these extreme events will both amplify the degree of impact and shift where within the aquatic continuum impacts occur. The metabolic stability of small lakes, for instance, may be completely overcome by extreme events shifting them from autotrophic to heterotrophic systems [Tsai et al., 2011]. Downstream systems that do not receive metabolic DOC subsidies with smaller events due to upstream utilization may also receive important loading with these extreme events [Buffam et al., 2001].
 The high SUVA values of the collected samples also indicate that the observed DOC flux contains photoreactive DOC that is important to ecosystem function. A portion of the DOC can provide labile organic compounds for microbial uptake [Moran and Zepp, 1997]. Absorption of sunlight by photoreactive DOC can also influence rates of primary production through shading [Carpenter et al., 1998], and impact the heat budget and thermal stratification of receiving water bodies [Caplanne and Laurion, 2008]. For water quality managers, the raised [DOC] coupled with high SUVA values directly translates to higher formation potential of disinfection by-products [Kitis et al., 2001]. Sunlight is also effective at inactivating parasites and pathogens such as Cryptosporidium in water [Connelly et al., 2007] and the UV-absorption of DOC can interfere with this critical ecosystem service.
 Another important contribution of this study is that the conventional models without high resolution data leads to significant underestimation of DOM fluxes during extreme events. Most watershed DOM data sets do not have representative sampling from events and tend to underestimate DOM fluxes during events [Raymond and Saiers, 2010]. If we had missed sampling this event, and used existing data and models (Model 2 and 3) to estimate the fluxes during Hurricane Irene, we would have underestimated it by 18–62% (except for Model 3 for DON, see Table 1). Furthermore, the fact that Model 2 and Model 3 utilized data sets with representative sampling for five other events shows that accurate load estimation of extreme events must be accompanied by sampling during the extreme events themselves. In regard to all of the aforementioned ecological and biogeochemical processes affected by large DOM fluxes, the underestimation seen from Model 2 and Model 3 is a clear sign that our conventional modeling methods fail to capture or predict the magnitude of their effects. Therefore, sampling during extreme events is of critical importance for refining our carbon budget, predicting change in aquatic metabolism, and managing water quality.
 Expanding further to the regional scale, this study provides evidence that current and future climate change may be increasing the amount of terrestrial organic matter exported to receiving water bodies. Large events have high concentrations of DOC, and the changing hydrologic variability forecasted and documented for the Northeast [Groisman et al., 2004, 2005] may lead to a greater percentage of terrestrial DON and DOC to be transferred to downstream lakes and rivers [Lepistö et al., 2008]. As stated, a greater number of larger events may lead to a greater overall travel distance of terrigenous DOM and strengthen the land-sea linkage under future climate scenarios, impacting both microbial processes and light attenuation in the coastal ocean. While these potential biological and biogeochemical outcomes remain to be studied, the increasing trend of large event frequency suggests that the importance of event-based DOM transport should be given more attention and monitoring in the future.
 This study demonstrates that the increase in DOC concentration during events holds for even the largest measured events. Hurricane Irene, the biggest event documented for the Catskill Mountain region, exported 43% and 31% of average annual DOC and DON fluxes, respectively. Conventional models for estimating DOM fluxes failed to capture the true magnitude of organic matters exported during this extreme event, if not supported by high-resolution data collected during the event. We argue that sampling during extreme events is of critical importance for refining our carbon budget and maintaining ecosystem services of the aquatic environment. Furthermore, the extreme fluxes seen through this study emphasize the critical role that extreme events play in exporting terrigenous organic matter through inland waters, and potential strengthening of land-sea linkage under future climate change scenarios.
 This study was supported by Hixon Center for Urban Ecology at Yale University. The authors would like to thank Brooks Avery and two anonymous reviewers for their constructive comments, USGS for loaning the necessary equipment, and Michael McHale for providing valuable guidance.
 The Editor thanks the anonymous reviewers for assisting in the evaluation of this paper.