Outlier detection in phosphorus dry deposition rates measured in South Florida

Dry atmospheric deposition contributes a significant amount of phosphorus to the Everglades of South Florida. Measurement of this deposition is problematic, because samples often are contaminated to varying degrees by bird droppings and other foreign materials. This study attempted to detect and remove the outliers in phosphorus (P) flux rates measured from dry deposition samples. Visual inspection of the samples, recorded in field notes, found that 30.1% of the samples contained animal droppings and frogs. Some of the samples with droppings and frogs (2.3%) had P values greater than 884 μg P m−2 d−1 (a value twice the standard deviation of the raw data mean), and were removed from further analysis. Outlier detection statistics based on a linear regression were then used for additional data screening. Eight stations in the network of 19 were removed because high contamination precluded the use of the regression model. Of the remaining samples, 15.7% were identified through the regression procedure as contaminated and were removed. The 11 station mean for P dry deposition was 85.8±79.0 μg P m−2 d−1, prior to the regression analysis, and 74.8±75.1 μg P m−2 d−1 after removal.