PREDICTING NONPOINT SOURCE POLLUTION IN AGRICULTURAL WATERSHEDS USING STATISTICAL ANALYSIS AND GIS TECHNIQUES IN LOWER GRAND RIVER WATERSHED, MISSOURI, USA

The water quality in many Midwestern streams and lakes are negatively impacted by agricultural activities. To better understand how a range of land use, geologic, and topographic parameters affect water quality in Midwestern watersheds, we sampled surface water quality parameters, including nitrate, phosphate, dissolved oxygen, turbidity, bacteria, pH, specific conductance, temperature, and biotic index in 35 independent sub-watersheds within the Lower Grand River Watershed in northern Missouri. For each sub-watershed, land use/land cover, soil texture, depth to bedrock, depth to the water table, and recent precipitation were determined. A digital elevation model was used to calculate the sub-watershed area, total stream length, watershed shape/relief ratio, topographic complexity, mean elevation, and slope. Water quality sampling was conducted at two times; one sampling campaign occurred in the spring, early in the growing season, and one campaign occurred in the late summer/early fall, near the end of the growing season. A pair-wise comparison of water quality parameters acquired in the fall and spring showed that each of these parameters was statistically different with season, suggesting that timing is very important in comparing water quality indicators. Correlations were calculated between each of the water quality indicators and the watershed characteristics, and predictive relationships were determined for each water quality indicator. Correlation analysis showed that both geologic and land use parameters were significantly correlated to water quality parameters. The percentage of land used for pasture/hay was important for several water quality parameters in both the fall and spring, while the percentage of land used for cultivated crops was only significant during the spring. Predictive relationships of water quality based on land use or geologic variables typically could explain less than half the variability of the water quality parameters, but may still be useful for indicating watersheds that would benefit from more sophisticated modeling or sampling. Principal component analysis was used to identify how correlations changed spatially and with season and to determine which variables were most beneficial for prediction of the biotic index.