ABSTRACT: Dynamic linear models (DLM) and seasonal trend decomposition (STL) using local regression, or LOESS, were used to analyze the 50‐year time series of suspended sediment concentrations for the Yadkin River, measured at the U.S. Geological Survey station at Yadkin College, North Carolina. A DLM with constant trend, seasonality, and a log10 streamflow regressor provided the best model to predict monthly mean log10 suspended sediment concentrations, based on the forecast log likelihood. Using DLM, there was evidence (odds approximately 69:1) that the log10 streamflow versus log10 suspended sediment concentration relationship has changed, with an approximate 20 percent increase in the log10 streamflow coefficient over the period 1981 to 1996. However, sediment concentrations in the Yadkin River have decreased during the decade of the 1990s, which has been accompanied by a concomitant increase in streamflow variability. Although STL has been shown to be a versatile trend analysis technique, DLM is shown to be more suitable for discovery and inference of structural changes (trends) in the model coefficient describing the relationship between flow and sediment concentration. 相似文献
The applications of chlorine have been broadly used in many industrial products, such as bleaching agents, synthetic rubbers, plastics, disinfectants, iron chlorides, fire refractory materials, insecticides, and anti-freezers, etc. According to the Taiwan Environmental Protection Administration (TEPA), more than 30 thousand tons were used in the year 2000. In addition, there were more than 12 reported incidents from 2000 to 2003—mostly on using chlorine as disinfectants (five) and as process agents (four).
This study investigated 15 chlorine operation plants in central Taiwan. These chlorine usages included bleaching agents, disinfectants, iron chloride, synthesizing rubber plastics, and others. Thirteen plants were located in the industrial parks and two were in or near residential zones. The consequence analysis were used three different methods to analyze the worst-case scenarios (WCSs) and alternative release case scenarios (ACSs) in order to compare impact zones for applying various active and passive mitigation systems, such as confined space, scrubber, water-spray, and so no. For two plants in or near residential zones, multi-layers mitigation systems and operation limits should be implemented in order to enforce more stringent protection measures. However, there was no specific regulation for chlorine plants operated at different locations, such as industrial parks or residential zones. In order to reduce chemical accidents and their impacts on public safety, our results suggest that source mitigation/management and warning systems should be adopted simultaneously. 相似文献
A discriminate analysis method for probability forecast of dust storms in Mongolia has been developed. The prediction method
uses data recorded at 23 meteorological stations in the Gobi and steppe regions of Mongolia, including surface air pressure
and geo-potential height at the 500-hPa level on grid points, and weather maps from 1975 to 1990.
Weather elements such as air temperature, pressure, geo-potential height etc, which influence the formation of dust storms,
are prepared as predictors. To select the most informative/important predictors (variables), we used a mean correlation matrix
of variables together with the Mahalonobis distance, and correlation coefficients between dust storms and predictors with
an orthogonalization for removing correlated predictors. The most informative predictors for dust storm prediction are intensities
of surface cyclones and migratory anticyclones, passage of cold fronts, the horizontal gradients of the surface air pressure
in the cold frontal zone, cyclonic circulations from the ground surface up to the 500-hPa level, the geo-potential height
at 500-hPa level and its temporal changes.
Selected predictors are used in discriminate analysis for formulating dust storm prediction equations. Sandstorm data have
been classified into three classes, viz., strong, moderate and weak dust storms, depending on their intensities, durations
and areas covered. Predictions of the probabilities of dust storm occurrence use the prediction equations for each class.
The prediction is made from 12 hours to 36 hours.
Verification of the probability forecasts of dust storms is also shown. The accuracy of forecasts is 72.2–79.9% with the data
used for developing equations (dependent variables), in contrast to 67.1–72.0% with unrelated data for deriving equations
(independent variables). 相似文献
The recreational-use value of hiking in the Bellenden Ker National Park, Australia has been estimated using a zonal travel cost model. Multiple destination visitors have been accounted for by converting visitors’ own ordinal ranking of the various sites visited to numerical weights, using an expected-value approach. The value of hiking and camping in this national park was found to be $AUS 250,825 per year, or $AUS 144,45 per visitor per year, which is similar to findings from other studies valuing recreational benefits. The management of the park can use these estimates when considering the introduction of a system of user pays fees. In addition, they might be important when decisions need to be made about the allocation of resources for maintenance or upgrade of tracks and facilities. 相似文献