EFFECTS OF MULTI‐SCALE ENVIRONMENTAL CHARACTERISTICS ON AGRICULTURAL STREAM BIOTA IN EASTERN WISCONSIN1

ABSTRACT: The U.S. Geological Survey examined 25 agricultural streams in eastern Wisconsin the determine relations between fish, invertebrate, and algal metrics and multiple spatial scales of land cover, geologic setting, hydrologic, aquatic habitat, and water chemistry data. Spearman correlation and redundancy analyses were used to examine relations among biotic metrics and environmental characteristics. Riparian vegetation, geologic, and hydrologic conditions affected the response of biotic metrics to watershed agricultural land cover but the relations were aquatic assemblage dependent. It was difficult to separate the interrelated effects of geologic setting, watershed and buffer land cover, and base flow. Watershed and buffer land cover, geologic setting, reach riparian vegetation width, and stream size affected the fish IBI, invertebrate diversity, diatom IBI, and number of algal taxa; however, the invertebrate FBI, percentage of EPT, and the diatom pollution index were more influenced by nutrient concentrations and flow variability. Fish IBI scores seemed most sensitive to land cover in the entire stream network buffer, more so than watershed‐scale land cover and segment or reach riparian vegetation width. All but one stream with more than approximately 10 percent buffer agriculture had fish IBI scores of fair or poor. In general, the invertebrate and algal metrics used in this study were not as sensitive to land cover effects as fish metrics. Some of the reach‐scale characteristics, such as width/depth ratios, velocity, and bank stability, could be related to watershed influences of both land cover and geologic setting. The Wisconsin habitat index was related to watershed geologic setting, watershed and buffer land cover, riparian vegetation width, and base flow, and appeared to be a good indicator of stream quality Results from this study emphasize the value of using more than one or two biotic metrics to assess water quality and the importance of environmental characteristics at multiple scales.     

REMOTE SENSING OF RIPARIAN BUFFERS: PAST PROGRESS AND FUTURE PROSPECTS1

Riparian buffer zone management is an area of increasing relevance as human modification of the landscape continues unabated. Land and water resource managers are continually challenged to maintain stream ecosystem integrity and water quality in the context of rapidly changing land use, which often offsets management gains. Approaches are needed not only to map vegetation cover in riparian zones, but also to monitor the changes taking place, target restoration activities, and assess the success of previous management actions. To date, these objectives have been difficult to meet using traditional techniques based on aerial photos and field visits, particularly over large areas. Recent advances in remote sensing have the potential to substantially aid buffer zone management. Very high resolution imagery is now available that allows detailed mapping and monitoring of buffer zone vegetation and provides a basis for consistent assessments using moderately high resolution remote sensing (e.g., Landsat). Laser‐based remote sensing is another advance that permits even more detailed information on buffer zone properties, such as refined topographic derivatives and multidimensional vegetation structure. These sources of image data and map information are reviewed in this paper, examples of their application to riparian buffer mapping and stream health assessment are provided, and future prospects for improved buffer monitoring are discussed.     

Implications of Land Use/Land Cover Change in the Buffer Zone of a National Park in the Tropical Andes

The impacts of land use and land cover (LULC) change in buffer zones surrounding protected ecological reserves have important implications for the management and conservation of these protected areas. This study examines the spatial and temporal patterns of LULC change along the boundary of Rio Abiseo National Park in the Northern Peruvian Andes. Landscape change within four ecological zones was evaluated based on trends expected to occur between 1987 and 2001. Landsat TM and ETM imagery were used to produce LULC classification maps for both years using a hybrid supervised/unsupervised approach. LULC changes were measured using landscape metrics and from-to change maps created by post-classification change detection. Contrary to expectations, tropical upper wet montane forest increased despite being threatened by human-induced fires and cattle grazing of the highland grasslands inside the park. Within the park’s buffer zone, tropical moist forest remnants were fragmented into more numerous and smaller patches between 1987 and 2001; this was in part due to conversion into agricultural land. The methods used in this study provide an effective way to monitor LULC change detection and support the management of protected areas and their surrounding environments.     

Estimating the Variability and Confidence of Land Use and Imperviousness Relationships at a Regional Scale1

Abstract: Impervious cover is a commonly used metric to help explain or predict anthropogenic impacts on aquatic resources; often it is used as a surrogate for intensity of human impacts when evaluating effects on aquatic resources. The most common way to estimate imperviousness is based on relationships with land use. Few studies have evaluated how the relationship between impervious surface and land use varies among geographies with different levels of development and between types of imagery used to assign land use type. In this study, we assess variability in estimates of imperviousness based on two locally available land use datasets: one based on aerial imagery (2‐m resolution) and another based on satellite imagery (30‐m resolution). The ranges and variability in imperviousness within land use categories were assessed at several spatial scales, including within counties, between counties, and between watersheds. Results indicate that there was considerable variability for all developed land use types. Estimated impervious cover often varied over a range of 20‐40% points within a land use category. Furthermore, there were clear spatial patterns both between and within counties, with impervious cover for a given land use type being higher near the urban centers and lower at the margins of development. Estimates of imperviousness for 12 study watersheds indicated that variability increased with increasing watershed development, making it difficult to confidently set management or regulatory targets based on impervious cover. This study suggests that locally derived, high resolution satellite or aerial imagery should be used to estimate imperviousness when a high level of accuracy and precision is required for regulatory or management decisions. Furthermore, the error associated with impervious land use relationships should be accounted for when using impervious cover in runoff or water quality models, or when making management decisions regarding stream health.     

Relating landscape characteristics to non-point source pollution in mine waste-located watersheds using geospatial techniques

Landscape characteristics of a watershed are important variables that influence surface water quality. Understanding the relationship between these variables and surface water quality is critical in predicting pollution potential and developing watershed management practices to eliminate or reduce pollution risk. To understand the impacts of landscape characteristics on water quality in mine waste-located watersheds, we conducted a case study in the Tri-State Mining District which is located in the conjunction of three states (Missouri, Kansas and Oklahoma). Severe heavy metal pollution exists in that area resulting from historical mining activities. We characterized land use/land cover over the last three decades by classifying historical multi-temporal Landsat imagery. Landscape metrics such as proportion, edge density and contagion were calculated based on the classified imagery. In-stream water quality data over three decades were collected, including lead, zinc, iron, cadmium, aluminum and conductivity which were used as key water quality indicators. Statistical analyses were performed to quantify the relationship between landscape metrics and surface water quality. Results showed that landscape characteristics in mine waste-located watersheds could account for as much as 77% of the variation of water quality indicators. A single landscape metric alone, such as proportion of mine waste area, could be used to predict surface water quality; but its predicting power is limited, usually accounting for less than 60% of the variance of water quality indicators.     

Preliminary comparison of landscape pattern-normalized difference vegetation index (NDVI) relationships to Central Plains stream conditions

We explored relationships of water quality parameters with landscape pattern metrics (LPMs), land use-land cover (LULC) proportions, and the advanced very high resolution radiometer (AVHRR) normalized difference vegetation index (NDVI) or NDVI-derived metrics. Stream sites (271) in Nebraska, Kansas, and Missouri were sampled for water quality parameters, the index of biotic integrity, and a habitat index in either 1994 or 1995. Although a combination of LPMs (interspersion and juxtaposition index, patch density, and percent forest) within Ozark Highlands watersheds explained >60% of the variation in levels of nitrite-nitrate nitrogen and conductivity, in most cases the LPMs were not significantly correlated with the stream data. Several problems using landscape pattern metrics were noted: small watersheds having only one or two patches, collinearity with LULC data, and counterintuitive or inconsistent results that resulted from basic differences in land use-land cover patterns among ecoregions or from other factors determining water quality. The amount of variation explained in water quality parameters using multiple regression models that combined LULC and LPMs was generally lower than that from NDVI or vegetation phenology metrics derived from time-series NDVI data. A comparison of LPMs and NDVI indicated that NDVI had greater promise for monitoring landscapes for stream conditions within the study area.     

EFFECTS OF ADJACENT AGRICULTURAL ACTIVITIES AND WATERSHED CHARACTERISTICS ON STREAM MACROINVERTEBRATE COMMUNITIES1

Benthic macroinvertebrate communities in streams adjacent to cornfields, streams where cows had unrestricted access, and reference locations without agriculture were compared to examine the effects of local land use and land use/land cover in the watershed. At each local site, macroinvertebrates and a variety of habitat parameters were measured upstream, adjacent, downstream, and farther downstream of the local land use. A geographic information system (GIS) was used to calculate drainage basin area, land use/land cover percentages in each basin, and the distance from sample sites to the stream source. Three‐way analysis of covariance (ANCOVA) tests with date, site type, and sampling location as main effects were used to explore differences in macroinvertebrate metrics using median substrate size, percent hay/pasture area, and stream depth as covariates. The covariates significantly improved model fit and showed that multiple contributing factors influence community composition. Local impacts were greatest at sites where cows had access, probably because of sedimentation and embeddedness in the substrate. Differences between the upstream and the adjacent and downstream locations were not as great as expected, perhaps because upstream recolonization was reduced by agricultural impacts or because of differences in the intensity or proximity of agriculture to riparian areas in the watershed. The results underscore the importance of both local and watershed factors in controlling stream community composition.     

IDW‐Plus: An ArcGIS Toolset for Calculating Spatially Explicit Watershed Attributes for Survey Sites

Watershed characteristics such as land‐use and land‐cover affect stream condition at multiple scales, but it is widely accepted that conditions in close proximity to the stream or survey site tend to have a stronger influence. Although spatially weighted watershed metrics have existed for years, nonspatial lumped landscape metrics (i.e., areal mean or percentage) are still widely used because relatively few technical skills are needed to implement them. The Inverse Distance Weighted Percent Land Use for Streams (IDW‐Plus) custom ArcGIS toolset provides the functionality to efficiently calculate six spatially explicit watershed metrics which account for the Euclidean or flow length distance to the stream or outlet, as well as the probability for overland runoff. These include four distance‐weighted metrics, those being inverse Euclidean distance to the stream or outlet, and the inverse flow length to the stream or outlet. Two tools are also included to generate hydrologically active (i.e., runoff potential), inverse flow length to the stream or outlet metrics. We demonstrate the tools using real data from Southeast Queensland, Australia. We also provide detailed instructions, so readers can recreate the examples themselves before applying the tools to their own data.     

Examining Landscape Dynamics at a Watershed Scale Using Landsat TM Imagery for Detection of Wintering Hooded Crane Decline in Yashiro,Japan

It is usually inappropriate to define rectangular land areas or administrative units as the extent for quantifying landscapes that possess hierarchical structure. As a functional unit established by geophysical relationships, the watershed is one of many natural scales in the hierarchical landscape. We examined the dynamics of the Yashiro watershed of Japan at the landscape level using pattern metrics based on Landsat thematic mapper (TM) imagery from 1985 to 1998. This watershed provides important habitats for the hooded crane (Grus monachus), a vulnerable species. While its world population has remained stable, the number wintering at Yashiro has declined in recent years. Changes in landscape metrics reveal that the spatial pattern within the watershed underwent homogenization due to depopulation of local people and shifts in local energy requirements and forest management policy at Yashiro. Specific changes include: a decrease in bare land area from 6.2% to 1.0% of the landscape, increased forest cover from 69.2% to 76.1%, reduction in patch number from 1194 to 616 and enlarged mean patch size, and a decrease in total edge from 223,740 m to 158,040 m. The rate of change in landscape metrics indicates a rapid change towards homogeneity in the landscape since 1990. The temporal changes in hooded crane populations corresponded to the changes in landscape. An alternative explanation has been proposed that decline of the species is influenced by landscape dynamics affecting both habitat selection and food resources. Conservation at the watershed scale is suggested to be complementary to the current conservation measures of the species.     

INFLUENCES OF WATERSHED,RIPARIAN‐CORRIDOR,AND REACH‐SCALE CHARACTERISTICS ON AQUATIC BIOTA IN AGRICULTURAL WATERSHEDS1

ABSTRACT: Multivariate analyses and correlations revealed strong relations between watershed and riparian‐corridor land cover, and reach‐scale habitat versus fish and macroinvertebrate assemblages in 38 warmwater streams in eastern Wisconsin. Watersheds were dominated by agricultural use, and ranged in size from 9 to 71 km² Watershed land cover was summarized from satellite‐derived data for the area outside a 30‐m buffer. Riparian land cover was interpreted from digital orthophotos within 10‐, 10‐to 20‐, and 20‐to 30‐m buffers. Reach‐scale habitat, fish, and macroinvertebrates were collected in 1998 and biotic indices calculated. Correlations between land cover, habitat, and stream‐quality indicators revealed significant relations at the watershed, riparian‐corridor, and reach scales. At the watershed scale, fish diversity, intolerant fish and EPT species increased, and Hilsenhoff biotic index (HBI) decreased as percent forest increased. At the riparian‐corridor scale, EPT species decreased and HBI increased as riparian vegetation became more fragmented. For the reach, EPT species decreased with embeddedness. Multivariate analyses further indicated that riparian (percent agriculture, grassland, urban and forest, and fragmentation of vegetation), watershed (percent forest) and reach‐scale characteristics (embeddedness) were the most important variables influencing fish (IBI, density, diversity, number, and percent tolerant and insectivorous species) and macroinvertebrate (HBI and EPT) communities.     

Scale Effects on Spatially Varying Relationships Between Urban Landscape Patterns and Water Quality

Scientific interpretation of the relationships between urban landscape patterns and water quality is important for sustainable urban planning and watershed environmental protection. This study applied the ordinary least squares regression model and the geographically weighted regression model to examine the spatially varying relationships between 12 explanatory variables (including three topographical factors, four land use parameters, and five landscape metrics) and 15 water quality indicators in watersheds of Yundang Lake, Maluan Bay, and Xinglin Bay with varying levels of urbanization in Xiamen City, China. A local and global investigation was carried out at the watershed-level, with 50 and 200 m riparian buffer scales. This study found that topographical features and landscape metrics are the dominant factors of water quality, while land uses are too weak to be considered as a strong influential factor on water quality. Such statistical results may be related with the characteristics of land use compositions in our study area. Water quality variations in the 50 m buffer were dominated by topographical variables. The impact of landscape metrics on water quality gradually strengthen with expanding buffer zones. The strongest relationships are obtained in entire watersheds, rather than in 50 and 200 m buffer zones. Spatially varying relationships and effective buffer zones were verified in this study. Spatially varying relationships between explanatory variables and water quality parameters are more diversified and complex in less urbanized areas than in highly urbanized areas. This study hypothesizes that all these varying relationships may be attributed to the heterogeneity of landscape patterns in different urban regions. Adjustment of landscape patterns in an entire watershed should be the key measure to successfully improving urban lake water quality.     

Classifying the Health of Connecticut Streams Using Benthic Macroinvertebrates with Implications for Water Management

Bioassessments have formed the foundation of many water quality monitoring programs throughout the United States. Like many state water quality programs, Connecticut has developed a relational database containing information about species richness, species composition, relative abundance, and feeding relationships among macroinvertebrates present in stream and river systems. Geographic Information Systems can provide estimates of landscape condition and watershed characteristics and when combined with measurements of stream biology, provide a useful visual display of information that is useful in a management context. The objective of our study was to estimate the stream health for all wadeable stream kilometers in Connecticut using a combination of macroinvertebrate metrics and landscape variables. We developed and evaluated models using an information theoretic approach to predict stream health as measured by macroinvertebrate multimetric index (MMI) and identified the best fitting model as a three variable model, including percent impervious land cover, a wetlands metric, and catchment slope that best fit the MMI scores (adj-R ² = 0.56, SE = 11.73). We then provide examples of how modeling can augment existing programs to support water management policies under the Federal Clean Water Act such as stream assessments and anti-degradation.     

Linking Landscape Characteristics and High Stream Nitrogen in the Oregon Coast Range: Red Alder Complicates Use of Nutrient Criteria

Red alder (Alnus rubra), a nitrogen(N)‐fixing deciduous broadleaf tree, can strongly influence N concentrations in western Oregon and Washington. We compiled a database of stream N and GIS‐derived landscape characteristics in order to examine geographic variation in N across the Oregon Coast Range. Basal area of alder, expressed as a percent of watershed area, accounted for 37% and 38% of the variation in summer nitrate and total N (TN) concentrations, respectively. Relationships between alder and nitrate were strongest in winter when streamflow and landscape connections are highest. Distance to the coast and latitude, potential surrogates for sea salt inputs, and watershed area were also related to nitrate concentrations in an all‐subsets regression analysis, which accounted for 46% of the variation in summer nitrate concentrations. The model with the lowest Akaike's Information Criterion did not include developed or agricultural land cover, probably because few watersheds in our database had substantial levels of these land cover classes. Our results provide evidence, at a regional scale, that background sources and processes cause many Coast Range streams to exceed proposed nutrient criteria, and that the prevalence of a single tree species (N‐fixing red alder) exerts a dominant control over stream N concentrations across this region.     

Stream Communities Along a Catchment Land-Use Gradient: Subsidy-Stress Responses to Pastoral Development

When native grassland catchments are converted to pasture, the main effects on stream physicochemistry are usually related to increased nutrient concentrations and fine-sediment input. We predicted that increasing nutrient concentrations would produce a subsidy-stress response (where several ecological metrics first increase and then decrease at higher concentrations) and that increasing sediment cover of the streambed would produce a linear decline in stream health. We predicted that the net effect of agricultural development, estimated as percentage pastoral land cover, would have a nonlinear subsidy-stress or threshold pattern. In our suite of 21 New Zealand streams, epilithic algal biomass and invertebrate density and biomass were higher in catchments with a higher proportion of pastoral land cover, responding mainly to increased nutrient concentration. Invertebrate species richness had a linear, negative relationship with fine-sediment cover but was unrelated to nutrients or pastoral land cover. In accord with our predictions, several invertebrate stream health metrics (Ephemeroptera–Plecoptera–Trichoptera density and richness, New Zealand Macroinvertebrate Community Index, and percent abundance of noninsect taxa) had nonlinear relationships with pastoral land cover and nutrients. Most invertebrate health metrics usually had linear negative relationships with fine-sediment cover. In this region, stream health, as indicated by macroinvertebrates, primarily followed a subsidy-stress pattern with increasing pastoral development; management of these streams should focus on limiting development beyond the point where negative effects are seen.     

IMPACTS OF URBAN LANDUSE ON MACROINVERTEBRATE COMMUNITIES IN SOUTHEASTERN WISCONSIN STREAMS1

ABSTRACT: Macroinvertebrates were used to assess the impact of urbanization on stream quality across a gradient of watershed imperviousness in 43 southeastern Wisconsin streams. The percentage of watershed connected imperviousness was chosen as the urbanization indicator to examine impact of urban land uses on macroinvertebrate communities. Most urban land uses were negatively correlated with the Shannon diversity index, percent of pollution intolerant Ephemeroptera, Plecoptera, and Trichoptera individuals, and generic richness. Nonurban land uses were positively correlated with these same metrics. The Hilsenhoff biotic index indicated that stream quality declined with increased urbanization. Functional feeding group metrics varied across a gradient of urbanization, suggesting changes in stream quality. Proportions of collectors and gatherers increased, while proportions of filterers, scrapers, and shredders decreased with increased watershed imperviousness. This study demonstrated that urbanization severely degraded stream macroinvertebrate communities, hence stream quality. Good stream quality existed where imperviousness was less than 8 percent, but less favorable assessments were inevitable where imperviousness exceeded 12 to 20 percent. Levels of imperviousness between 8 and 12 percent represented a threshold where minor increases in urbanization were associated with sharp declines in stream quality.     

An assessment of landscape characteristics affecting estuarine nitrogen loading in an urban watershed

Exploring the quantitative association between landscape characteristics and the ecological conditions of receiving waters has recently become an emerging area for eco-environmental research. While the landscape-water relationship research has largely targeted on inland aquatic systems, there has been an increasing need to develop methods and techniques that can better work with coastal and estuarine ecosystems. In this paper, we present a geospatial approach to examine the quantitative relationship between landscape characteristics and estuarine nitrogen loading in an urban watershed. The case study site is in the Pensacola estuarine drainage area, home of the city of Pensacola, Florida, USA, where vigorous urban sprawling has prompted growing concerns on the estuarine ecological health. Central to this research is a remote sensor image that has been used to extract land use/cover information and derive landscape metrics. Several significant landscape metrics are selected and spatially linked with the nitrogen loading data for the Pensacola bay area. Landscape metrics and nitrogen loading are summarized by equal overland flow-length rings, and their association is examined by using multivariate statistical analysis. And a stepwise model-building protocol is used for regression designs to help identify significant variables that can explain much of the variance in the nitrogen loading dataset. It is found that using landscape composition or spatial configuration alone can explain most of the nitrogen loading variability. Of all the regression models using metrics derived from a single land use/cover class as the independent variables, the one from the low density urban gives the highest adjusted R-square score, suggesting the impact of the watershed-wide urban sprawl upon this sensitive estuarine ecosystem. Measures towards the reduction of non-point source pollution from urban development are necessary in the area to protect the Pensacola bay ecosystem and its ecosystem services.     

EXTRACTION AND UTILIZATION OF SPACE ACQUIRED PHYSIOGRAPHIC DATA FOR WATER RESOURCES DEVELOPMENT1

ABSTRACT: ERTS-1 satellite imagery has been evaluated as a means of providing useful watershed physiography information. From these data physiographic parameters such as drainage basin area and shape, drainage density, stream length and sinuosity, and the percentage of a watershed occupied by major land use types were obtained in three study areas. The study areas were: (1) Southwestern Wisconsin; (2) Eastern Colorado; and (3) portions of the Middle Atlantic States Using ERTS-1 imagery at 1:250,000 and 1:100,000 scales it was found that drainage basin area and shape and stream sinuosity were comparable (within 10%) in all study areas to physiographic measurements derived from conventional topographic maps at the same scales Land use information can be usefully extracted for watersheds as small as 30 mi²(78 km²) in area. Improved drainage network and density information is obtained from ERTS-1 imagery in dissected areas such as Southwestern Wisconsin, but in heavily vegetated areas (Middle Atlantic States) or areas with little physical relief (Eastern Colorado) low order streams are difficult to detect and the derived drainage densities are significantly smaller than those obtained from standard maps. It is concluded that ERTS-1 imagery can be employed to advantage in mean annual runoff prediction techniques and in providing or maintaining land use information used in the calibration and operation of watershed models.     

Statistical and Spatial Analysis of Land Cover Impact on Selected Metro Vancouver, British Columbia Watersheds

The Greater Vancouver area has undergone significant land use and land cover (LULC) change over the past several decades, often adversely affecting stream health and water quality, particularly in those areas that have undergone the most urbanization. In this study 30 years of historical LULC and water quality data were examined using GIS and statistical analysis to better understand these impacts and to help build a broader understanding of cause and effect relationships of changing LULC, especially since urbanization is increasingly occurring within sensitive watersheds at greater distances from the City of Vancouver. Urban, agriculture, and disturbed LULC data from 1976, 1986, and 2000 were examined within a number of watersheds and related to historical water quality data sampled from streams during similar time frames. Additional higher resolution 2006 LULC data from a smaller number of watersheds were then examined and compared to stream health data to investigate the sensitivity of LULC data resolution on monitoring watershed impact. While LULC impact can be clearly seen at both high and lower resolutions, issues of ambiguous land cover and land use designations can potentially affect the magnitude of the relationship.     

Development of a Bird Integrity Index: Measuring Avian Response to Disturbance in the Blue Mountains of Oregon, USA

The Bird Integrity Index (BII) presented here uses bird assemblage information to assess human impacts to 28 stream reaches in the Blue Mountains of eastern Oregon. Eighty-one candidate metrics were extracted from bird survey data for testing. The metrics represented aspects of bird taxonomic richness, tolerance or intolerance to human disturbance, dietary preferences, foraging techniques, and nesting strategies that were expected to be positively or negatively affected by human activities in the region. To evaluate the responsiveness of each metric, it was plotted against an index of reach and watershed disturbance that included attributes of land use/land cover, road density, riparian cover, mining impacts, and percent area in clearcut and partial-cut logging. Nine of the 81 candidate bird metrics remained after eliminating unresponsive and highly correlated metrics. Individual metric scores ranged from 0 to 10, and BII scores varied between 0 and 100. BII scores varied from 78.6 for a minimally disturbed, reference stream reach to 30.4 for the most highly disturbed stream reach. The BII responded clearly to varying riparian conditions and to the cumulative effects of disturbances, such as logging, grazing, and mining, which are common in the mountains of eastern Oregon. This BII for eastern Oregon was compared to an earlier BII developed for the agricultural and urban disturbance regime of the Willamette Valley in western Oregon. The BII presented here was sensitive enough to distinguish differences in condition among stream riparian zones with disturbances that were not as obvious or irreversible as those in the agricultural/urban conditions of western Oregon.     

Spatial analysis of land use impact on ground water nitrate concentrations

In spatial analyses of causes or health effects of environmental pollutants, small units of analyses are usually preferred for internal environmental homogeneity reasons but can only be done when fine resolution data are available for most units. Objectives of this study were to determine which land use practices were spatially associated with ground water nitrate concentrations across Prince Edward Island (PEI), Canada, and which spatial aggregation is the preferred unit of analyses. Nitrate concentrations were determined for 4855 samples from private wells. Validated field-by-field land use data were available. Average nitrate concentration and percentage of area for the 14 major land use categories in PEI were determined for each of three spatial aggregations: watersheds based on topography and hydrology; freeform polygon boundaries based on similar neighboring nitrate concentrations; and 500-m buffer zones around each well. Results showed that the percentages of potato, grain, and hay coverage were positive predictors of ground water nitrate concentrations. Percentage of blueberry was a marginally significant negative predictor in the watershed and freeform polygon models, and percentage of residential coverage was a positive predictor in the freeform polygon and buffer zone models. Spatial autocorrelation was present in the freeform polygon and buffer zone models even after land use was taken into account. In conclusion, analyses based on watersheds produced the best predictive model with the percentages of land cover of potato, hay, and grain being significantly associated with ground water nitrate concentrations, and the percentages of blueberry, clear-cut woodland, and other agriculture being marginally significant.