Toward characterizing heterogeneity of biogeochemical solutes in wetlands of central Pennsylvania

Our research focuses on investigating the impacts of anthropogenic disturbance on ecosystem processes, including nutrient biogeochemistry, within wetland and headwater stream aquatic ecosystems. More specifically, we are interested in characterizing the response of a number of physical, chemical, and biological parameters across this disturbance gradient, including statistical descriptors of heterogeneity such as value ranges, variance, and autocorrelation. Here, we present results from a case study aimed at an initial characterization of the variability of biogeochemical parameters in groundwater in wetland environments (specifically headwater riverine wetlands, i.e., those associated with first and second order streams) of central Pennsylvania. Water samples were collected in saturated pore waters of two naturally occurring, relatively un-impacted headwater riparian wetland sites. Each site contained 20 shallow (about 40 cm below surface) groundwater monitoring wells placed in a random arrangement within a 400 m2 total sampling area. Water samples from each well were analyzed for water quality parameters related to the biogeochemical status of the wetland, including pH, temperature, conductivity, dissolved oxygen (DO), dissolved organic carbon (DOC), dissolved organic nitrogen (DON), ammonia, nitrate, total nitrogen (TN), and more. The various descriptors of heterogeneity exhibit changes across the disturbance gradient, and we were thus interested in the implications of these differences on estimates of biogeochemical cycling. In order to make first comparisons to other published studies and to utilize nitrogen cycling models to provide first cut estimates, we were interested in how our spatially-intensive data set affected the representation of average site conditions. To assess the minimum number of samples that are necessary to characterize average site-wide conditions in shallow ground water quality across each wetland site, the mean of all 20 samples for each chemical constituent was taken to represent the actual site mean. Then, between 1 and 17 samples were randomly selected 100 times for each chemical constituent and the mean of the selected samples was calculated. The coefficient of variation for the resulting 100 means was then determined. If the resulting coefficient of variation was under 10%, then the mean was considered to be representative of the site mean. The number of samples required to obtain a mean within 10% of the site mean varied depending on both the chemical constituent and the site. For characterizing pH and temperature, fewer than 5 samples were needed to be within 10% of the site mean. However, for DOC, TN, and nitrate more than 10 samples were required to be within 10% of the site mean. For some reactive constituents such as ammonia and DON, more than 15 samples were needed to be within 10% of the site mean. In many published studies of nutrient cycling performed in wetlands, water sample numbers are very sparse due to monetary, design, and time constraints, and it is common to find studies where single or few observations are taken to represent descriptive statistics of the wetland and to parameterize models of ecosystem response. Our results highlight that a sparse number of samples is inadequate to accurately represent the biogeochemical status of a typical wetland environment.