Quantifying uncertainty in annual runoff due to missing data.

Long-term streamflow datasets inevitably include gaps, which must be filled to allow estimates of runoff and ultimately catchment water budgets. Uncertainty introduced by filling gaps in discharge records is rarely, if ever, reported. We characterized the uncertainty due to streamflow gaps in a reference watershed at the Hubbard Brook Experimental Forest (HBEF) from 1996 to 2009 by simulating artificial gaps of varying duration and flow rate, with the objective of quantifying their contribution to uncertainty in annual streamflow. Gaps were filled using an ensemble of regressions relating discharge from nearby streams, and the predicted flow was compared to the actual flow. Differences between the predicted and actual runoff increased with both gap length and flow rate, averaging 2.8% of the runoff during the gap. At the HBEF, the sum of gaps averaged 22 days per year, with the lowest and highest annual uncertainties due to gaps ranging from 1.5 mm (95% confidence interval surrounding mean runoff) to 21.1 mm. As a percentage of annual runoff, uncertainty due to gap filling ranged from 0.2-2.1%, depending on the year. Uncertainty in annual runoff due to gaps was small at the HBEF, where infilling models are based on multiple similar catchments in close proximity to the catchment of interest. The method demonstrated here can be used to quantify uncertainty due to gaps in any long-term streamflow data set, regardless of the gap-filling model applied.

Trace metals in Northern New England streams: Evaluating the role of road salt across broad spatial scales with synoptic snapshots.

Mobilization of trace metals from soils to surface waters can impact both human and ecosystem health. This study resamples a water sample archive to explore the spatial pattern of streamwater total concentrations of arsenic, cadmium, copper, lead, and zinc and their associations with biogeochemical controls in northern New England. Road deicing appears to result in elevated trace metal concentrations, as trace metal concentrations are strongly related to sodium concentrations and are most elevated when the sodium: chloride ratio is near 1.0 (~halite). Our results are consistent with previous laboratory and field studies that indicate cation exchange as a metal mobilization mechanism when road salt is applied to soils containing metals. This study also documents associations among sodium, chloride, dissolved organic carbon, iron, and metal concentrations, suggesting cation exchange mechanisms related to road deicing are not the only mechanisms that increase trace metal concentrations in surface waters. In addition to cation exchange, this study considers dissolved organic carbon complexation and oxidation-reduction conditions affecting metal mobility from soils in a salt-rich environment. These observations demonstrate that road deicing has the potential to increase streamwater trace metal concentrations across broad spatial scales and increase risks to human and ecosystem health.

Estimating uncertainty in the volume and carbon storage of downed coarse woody debris.

Downed coarse woody debris, also known as coarse woody detritus or downed dead wood, is challenging to estimate for many reasons, including irregular shapes, multiple stages of decay, and the difficulty of identifying species. In addition, some properties are commonly not measured, such as wood density and carbon concentration. As a result, there have been few previous evaluations of uncertainty in estimates of downed coarse woody debris, which are necessary for analysis and interpretation of the data. To address this shortcoming, we quantified uncertainties in estimates of downed coarse woody debris volume and carbon storage using data collected from permanent forest inventory plots in the northeastern United States by the Forest Inventory and Analysis program of the USDA Forest Service. Quality assurance data collected from blind remeasurement audits were used to quantify error in diameter measurements, hollowness of logs, species identification, and decay class determination. Uncertainty estimates for density, collapse ratio, and carbon concentration were taken from the literature. Estimates of individual sources of uncertainty were combined using Monte Carlo methods. Volume estimates were more reliable than carbon storage, with an average 95% confidence interval of 15.9 m3 /ha across the 79 plots evaluated, which was less than the mean of 31.2 m3 /ha. Estimates of carbon storage (and mass) were more uncertain, due to poorly constrained estimates of the density of wood. For carbon storage, the average 95% confidence interval was 11.1 Mg C/ha, which was larger than the mean of 4.6 Mg C/ha. Accounting for the collapse of dead wood as it decomposes would improve estimates of both volume and carbon storage. On the other hand, our analyses suggest that consideration of the hollowness of downed coarse woody debris pieces could be eliminated in this region, with little effect. This study demonstrates how uncertainty analysis can be used to quantify confidence in estimates and to help identify where best to allocate resources to improve monitoring designs.

Controls of Chloride Loading and Impairment at the River Network Scale in New England.

Chloride contamination of rivers due to nonpoint sources is increasing throughout developed temperate regions due to road salt application in winter. We developed a river-network model of chloride loading to watersheds to estimate road salt application rates and investigated the meteorological factors that control riverine impairment by chloride at concentrations above thresholds protective of aquatic organisms. Chloride loading from road salt was simulated in the Merrimack River watershed in New Hampshire, which has gradients in development density. After calibration to a regional network of stream chloride data, the model captured the distribution of regional discharge and chloride observations with efficiencies of 93 and 75%, respectively. The estimate of road salt application is within uncertainties of inventoried estimates of road salt loading and is 122 to 214% greater than recommended targets. Model predictions of chloride showed seasonal variation in chloride concentrations despite a large groundwater storage pool. Interannual variation of mean summer chloride concentration near the outlet varied up to 18%, and the total river length exceeding impairment thresholds varied 12%. Annual snowfall, which drives road salt loading, correlated with chloride impairment only in headwater streams, whereas concentration variability at the outlet was driven primarily by dilution from clean runoff-draining undeveloped forested areas of the watershed. The role of summer meteorology complicates the protection of freshwater systems from chloride contamination.

Downsizing a long-term precipitation network: Using a quantitative approach to inform difficult decisions.

The design of a precipitation monitoring network must balance the demand for accurate estimates with the resources needed to build and maintain the network. If there are changes in the objectives of the monitoring or the availability of resources, network designs should be adjusted. At the Hubbard Brook Experimental Forest in New Hampshire, USA, precipitation has been monitored with a network established in 1955 that has grown to 23 gauges distributed across nine small catchments. This high sampling intensity allowed us to simulate reduced sampling schemes and thereby evaluate the effect of decommissioning gauges on the quality of precipitation estimates. We considered all possible scenarios of sampling intensity for the catchments on the south-facing slope (2047 combinations) and the north-facing slope (4095 combinations), from the current scenario with 11 or 12 gauges to only 1 gauge remaining. Gauge scenarios differed by as much as 6.0% from the best estimate (based on all the gauges), depending on the catchment, but 95% of the scenarios gave estimates within 2% of the long-term average annual precipitation. The insensitivity of precipitation estimates and the catchment fluxes that depend on them under many reduced monitoring scenarios allowed us to base our reduction decision on other factors such as technician safety, the time required for monitoring, and co-location with other hydrometeorological measurements (snow, air temperature). At Hubbard Brook, precipitation gauges could be reduced from 23 to 10 with a change of <2% in the long-term precipitation estimates. The decision-making approach illustrated in this case study is applicable to the redesign of monitoring networks when reduction of effort seems warranted.

The promise and peril of intensive-site-based ecological research: insights from the Hubbard Brook ecosystem study.

Ecological research is increasingly concentrated at particular locations or sites. This trend reflects a variety of advantages of intensive, site-based research, but also raises important questions about the nature of such spatially delimited research: how well does site based research represent broader areas, and does it constrain scientific discovery? We provide an overview of these issues with a particular focus on one prominent intensive research site: the Hubbard Brook Experimental Forest (HBEF), New Hampshire, USA. Among the key features of intensive sites are: long-term, archived data sets that provide a context for new discoveries and the elucidation of ecological mechanisms; the capacity to constrain inputs and parameters, and to validate models of complex ecological processes; and the intellectual cross-fertilization among disciplines in ecological and environmental sciences. The feasibility of scaling up ecological observations from intensive sites depends upon both the phenomenon of interest and the characteristics of the site. An evaluation of deviation metrics for the HBEF illustrates that, in some respects, including sensitivity and recovery of streams and trees from acid deposition, this site is representative of the Northern Forest region, of which HBEF is a part. However, the mountainous terrain and lack of significant agricultural legacy make the HBEF among the least disturbed sites in the Northern Forest region. Its relatively cool, wet climate contributes to high stream flow compared to other sites. These similarities and differences between the HBEF and the region can profoundly influence ecological patterns and processes and potentially limit the generality of observations at this and other intensive sites. Indeed, the difficulty of scaling up may be greatest for ecological phenomena that are sensitive to historical disturbance and that exhibit the greatest spatiotemporal variation, such as denitrification in soils and the dynamics of bird communities. Our research shows that end member sites for some processes often provide important insights into the behavior of inherently heterogeneous ecological processes. In the current era of rapid environmental and biological change, key ecological responses at intensive sites will reflect both specific local drivers and regional trends.

Changing forest water yields in response to climate warming: results from long-term experimental watershed sites across North America.

Climate warming is projected to affect forest water yields but the effects are expected to vary. We investigated how forest type and age affect water yield resilience to climate warming. To answer this question, we examined the variability in historical water yields at long-term experimental catchments across Canada and the United States over 5-year cool and warm periods. Using the theoretical framework of the Budyko curve, we calculated the effects of climate warming on the annual partitioning of precipitation (P) into evapotranspiration (ET) and water yield. Deviation (d) was defined as a catchment's change in actual ET divided by P [AET/P; evaporative index (EI)] coincident with a shift from a cool to a warm period - a positive d indicates an upward shift in EI and smaller than expected water yields, and a negative d indicates a downward shift in EI and larger than expected water yields. Elasticity was defined as the ratio of interannual variation in potential ET divided by P (PET/P; dryness index) to interannual variation in the EI - high elasticity indicates low d despite large range in drying index (i.e., resilient water yields), low elasticity indicates high d despite small range in drying index (i.e., nonresilient water yields). Although the data needed to fully evaluate ecosystems based on these metrics are limited, we were able to identify some characteristics of response among forest types. Alpine sites showed the greatest sensitivity to climate warming with any warming leading to increased water yields. Conifer forests included catchments with lowest elasticity and stable to larger water yields. Deciduous forests included catchments with intermediate elasticity and stable to smaller water yields. Mixed coniferous/deciduous forests included catchments with highest elasticity and stable water yields. Forest type appeared to influence the resilience of catchment water yields to climate warming, with conifer and deciduous catchments more susceptible to climate warming than the more diverse mixed forest catchments.

Decreased water flowing from a forest amended with calcium silicate.

Acid deposition during the 20th century caused widespread depletion of available soil calcium (Ca) throughout much of the industrialized world. To better understand how forest ecosystems respond to changes in a component of acidification stress, an 11.8-ha watershed was amended with wollastonite, a calcium silicate mineral, to restore available soil Ca to preindustrial levels through natural weathering. An unexpected outcome of the Ca amendment was a change in watershed hydrology; annual evapotranspiration increased by 25%, 18%, and 19%, respectively, for the 3 y following treatment before returning to pretreatment levels. During this period, the watershed retained Ca from the wollastonite, indicating a watershed-scale fertilization effect on transpiration. That response is unique in being a measured manipulation of watershed runoff attributable to fertilization, a response of similar magnitude to effects of deforestation. Our results suggest that past and future changes in available soil Ca concentrations have important and previously unrecognized implications for the water cycle.