The integration of measured, modelled and remotely sensed air quality data and its' impacts on the Highveld.

[摘要] Although a vast number of air quality investigations have been conducted on theMpumalanga Highveld previously, there has been limited attempt to integrateavailable datasets from the different methods of air quality monitoring (satellite, insituand ground-based observations) and modelling. This study compares modelled,satellite and measured data to determine the most accurate estimate of ground levelsulphur dioxide (SO2) and nitrogen oxide (NOx) concentrations.The main value of the project comes from the 'improvement’ of modelledconcentration fields using measurements. Measurements only provide information onair quality at isolated places (for example monitoring stations) or at isolated times(aircraft measurements and satellite observations). Dispersion models predictconcentrations continually over a wide area. However, models have inherentinaccuracies based on the assumptions made in developing the model and thevariability in the input parameters supplied. These can be accounted for or are part ofthe inherent variability of the model results. This study assists in the refinement ofmodelled outputs as well as the verification of satellite data using ground-basedmeasured data as a point of reference.In the wake of increasing governance on air pollution, industry has been compelled toaccount for their impacts on the environment. This study aids industry by proposing amethod to quantify their impacts on the environment and possibly on human health.Three datasets from 2003 (modelled, measured and satellite) were integrated using ageographic information system in order to analyse and interrogate the data andproduce an integrated set of data, maps of potentially sensitive ecosystems and mapsof potential exposure to poor air quality of sensitive population groupings.The results of the study have shown that although the concentration value for the NO2ivtropospheric column is greater than the values observed on the ground there is a goodcorrelation between measured observations and SCIAMACHY retrievals. The samplesize was too small to indicate a statistically significant bias.The results from the validation of the CALPUFF model indicate with respect to SO2predictions that themodel is only reliable for 62% of the time within the United StatesEnvironmental Protection Agency’s model performance guideline of acceptance i.e.predictions within a factor of 2, and for NO2 the predictions are reliable for only 50%of the time. There is also no constant value by which the model under or overpredicts. The cumulative distribution function graphs illustrate that the CALPUFFmodel predicts the highest values from the bulk of the distribution rather that the tailof the distribution where the extremes lie. This could possibly account for the largevariance between measured and modelled outputs.The results of identifying areas of potential harm from SO2 emissions reveal thathotspots for high to very high risk to human health occur around all power stations.Generally the category of high risk around power stations seems to be located in areaswith population agglomerations between 0-1 000 per km2 and 2001- 5000 per km2.Several high risk areas for potential harm to ecosystems from SO2 emissions can beseen on the Highveld with a large spatial extent around Kendal, Matla and Krielpower station. Approximately 871 wetlands fall within the high risk areas. Thevegetation risk map indicates a high risk to several grassland and bushveld types.Model results for this study indicate no potential risk to human health from emissionsof NO and subsequent conversion to NO2 in the atmosphere.

[发布机构] University of the Witwatersrand