Statistical Modeling of Phosphorus Dry Deposition Rates Measured in South Florida     

Ahn  Hosung  James  R. Thomas 《Water, air, and soil pollution》2000,121(1-4):43-57

Dry deposition may be a substantial source ofphosphorus (P) to the Florida Everglades. Drydeposition has been measured on a weekly basis in theregion since 1987, but a significant amount of thisdata is missing (about 34%) due to instrumentalfailures and sample contamination. This study developsa statistical model of the P dry depositiontime-series to estimate missing data. The model isbased on a multivariate stochastic time-series theory.Model parameters are calibrated using theexpectation-maximization algorithm which is efficientfor data sets with many gaps. The pooled mean andstandard deviation of the data before estimating themissing values was 88.4±85.7 μg P m^-2d^-1 and after estimating the missing values was87.8±82.4 μg P m^-2 d^-1. Modelverification demonstrates that the calibrated modelsprovide unbiased data estimates while preserving thestatistics of the raw data. For each sampling site themean and standard deviation before and after werequite similar. No trend with time was detected. The Pdeposition fluctuates seasonally (highest in Octoberand lowest in June), but this fluctuation does notfollow the seasonal pattern of Florida's rainfall. Random noise in the data, however, is significant andcauses long-term fluctuations of the data. The datawith gaps filled in are useful for computing theweekly P load distribution.  相似文献   

Variability,Uncertainty, and Sensitivity of Phosphorus Deposition Load Estimates in South Florida     

Ahn  Hosung  James  R. Thomas 《Water, air, and soil pollution》2001,126(1-2):37-51

Atmospheric deposition, a substantial source of phosphorus (P) to the Florida Everglades, has been measured on a weekly basis in South Florida since 1974, but P measurements are highly variable due to random noise in the data. This study applies statistical approaches that calculatethe variability and uncertainty of the P load estimation model based on wet and dry P concentrations and rainfall volume.The average mean and standard deviation of the estimated P deposition rates for 13 sites in South Florida are 41±33 mg P m^-2 yr^-1. First order analysis of the random and measurement errors in the input variables produces a propagation error estimate in P load calculation. The atmospheric P deposition load shows high spatial and temporal variability with no consistent long-term trends. Because of the random noisy nature of P deposition, estimated P deposition loads have a significant amount of uncertainty no matter what type of collection instrument is used. Thus, duplicate sampling is highly recommended to increase the amount of uncontaminated data.  相似文献