Seasonal and Long-Term Nutrient Trend Decomposition along a Spatial Gradient in the Neuse River Watershed SONG S. QIAN* Environm ental Sciences and Resources, Portland State University, Portland, Oregon 97207-0751 MARK E. BORSUK AND CRAIG A. STOW Nicholas School of the Environm ent, Duke University, Durham , North Carolina 27708-0328 The Neuse River Estuary in North Carolina has recently received considerable public attention for severe algal blooms, large fish kills, and outbreaks of toxic microorganisms. To investigate the belief that nutrient enrichment has worsened in recent years, we analyzed long-term and seasonal trends in nutrient concentrations along the river and estuary employing seasonal trend decomposition using local regression analysis (STL). The nonparametric nature of the STL approach makes it possible to identify nonlinear trends and seasonal interactions that would be missed by traditional trend detection methods. The results indicate that while there may have been minor increases in nitrogen concentrations at upstream locations over the past twenty years, those changes are not reflected in the lower river and estuary.However,the pronounced decreases in phosphorus concentrations that occurred upstream, corresponding to a phosphorus detergent ban in 1988, do persist downstream. The net result is that the ratio of nitrogen to phosphorus concentrations in the estuary has increased considerably in the last 10 years. When compared with the Redfield value, ambient nutrient ratios suggest that phytoplankton growth in the estuary may be experiencing a shift from nitrogen to phosphorus limitation during much of the year. This shift may be inducing a change in the biotic community that would help explain the perception of worsening eutrophication, despite an overall reduction in nutrient concentrations. Introduction Nutrient enrichment is a serious water quality problem in many coastal rivers and estuaries worldwide (1). Excessive nutrient loading may result in algal blooms, bottom water hypoxia, massive fish kills, and outbreaks of toxic microor- ganisms. Management of coastal eutrophication can be enhanced byinformation on the spatialand temporaltrends in historicalnutrient concentration data.Clearlydocumented historical trends reveal the response of a natural system to either the unintentional consequences of human activity or the deliberate results of past management (2, 3). Many statistical methods have been proposed for the detection of environmental trends (4, 5). The choice of a trend detection method depends on the type of trend expected (continuous versus step function), the adherence of the data to various assumptions (normality, independence, homoscedasticity), and the occurrence of censored data (6, 4). Traditionally, environmental trend analyses have used linear regression analysisor nonparametricmethodsbased on order statistics, including Kendall’s tau test for correlation (7)and variations (8, 9). However, these methods are constrained to linear or monotonic trends and cannot detect intermediate reversals in direction. Short-term departures from a long-term trend maybe important in assessing the impacts ofpreliminaryor spatially limited management actions. Additionally, tradi- tionaltrend detection methods decompose a time series into a long-term component and an additive seasonalcomponent. However,this approach assumes that long-term patterns for allseasonsdifferonlyin magnitude.Thisadditiveassumption isoften not appropriate forenvironmentaldata since different seasons often involve different forcingfunctions.Discerning systematic changes in the seasonal pattern is essential for accurately assessing ecosystem response to environmental changes. One solution is to use locally weighted regression (loess) methods to describe nonlinear trends that do not have an obvious functional form (10, 11). By using the generalized additive modeling approach (12), seasonal trend decompo- sition using loess (STL) (13) provides a nonparametric graphical method for describing nonlinear trends with seasonal interaction. We present an analysis of spatial and temporal trends in nutrient concentrations in the Neuse River and Estuary,North Carolina, U.S.A. using the STL methodology. The Neuse estuary has recently received considerable public attention due to characteristic symptoms ofexcessive eutrophication. The general perception is that these problems are due to a recent increase in watershed nutrient inputs (14, 15, 1), and management options,includingriparian buffers and source reductions,are planned.While conventionaltrend detection methodsbased on monotonictrendsand an additive seasonal component have provided useful information about other estuaries in the past, a more flexible method, such as STL, is desirable for the Neuse. Sudden changes have occurred in the watershed, including the impoundment of a large reservoir in the headwaters of the river in 1983 and a phosphate detergent ban in 1988,which would be improperly described by a monotonic trend analysis. Additionally, discerning changes in seasonal patterns is essential given the seasonalnature ofboth proposed nutrient management actionsand bioticresponse.Finally,an assessment ofspatial patterns provides information about source changes, sub- watershed specificmanagement actions,and bioticresponse that would not be obtained from trends described at one location only. Therefore, we will present our analysis in a format that facilitates the assessment of both temporal and spatial pattern. 2. Materials and Methods 2.1. Study Site and Data Description. The Neuse River originates north of Durham, NC, at the confluence of the Flat and Eno rivers. Shortly below this confluence, a 35 km reach was impounded in 1983 to create Falls Lake, a multipurpose recreation and drinking water reservoir (16). Belowthe dam,the riverflowsapproximately320km through the central piedmont to the coastal plain and comprises a watershed area of16,108 km 2 (Figure 1). The major land uses in the basin are agriculture (35%) and forestry(34%) with the remainder consisting primarily of urban areas, wetlands, *Correspondingauthor phone: (503)725-8190;fax: (503)725-3888; e-mail: qians@pdx.edu. Environ. Sci. Technol. 2000, 34, 4474-4482 4474 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 21, 2000 10.1021/es000989p CCC: $19.00  2000 American Chemical Society Published on Web 09/22/2000