Air Pollution Modeling With Geostatistical Analysis
Faculty of Surveying engineering
K.N.Toosi university of Technology
Faculty of Surveying engineering
K.N.Toosi university of Technology
Dept. of environmental health
Shahid Beheshti medical science university
The air pollution of metropolitan areas is one of the major problems of the world at present. Expanded spatial dispersion of the pollutants such as CO, NO2, SO2 and PM10 which are emitted from stationary and mobile source or natural disasters has
In this research we consider on GIS design concept and application of GIS in Air Quality Management. Further more interpolation method selected according to capability of GIS among various air pollution modeling.
Then the researcher presents the optimum interpolation method for each pollutant. Cross validation method is suitable for evaluation of interpolation method because there are few monitoring stations in Tehran. Based on theoretical studies, Kriging and IDW methods always include the perfect result.
The results of GIS as environmental decision support system (EDSS) are being effective on environmental data management, spatial analysis of different pollutants and alternative procedure of pollution amount.
Tehran is the capital city of Iran and located at 36?N, 51?E. Mountains area which surround north and east and only have a slight slope on the south. Air pollution has created by increasing of population and industrial development as well [Rahmatizadeh, 2004]. Air pollution may be defined as the accumulation in the air of any natural or artificially composed material, solid particles, liquid droplets or gas, that are either directly or indirectly harmful or dangerous to humans in particular, or to natural environment at large. It has been demonstrated that air pollution causes serious health problems and negative environmental impacts.
The air pollution reach to a level which can significantly influence human’s health. For each city clean air is one of the most valuable factors. Air quality management system (AQMS) can be defined as a regulation of the amount, location and time of pollutant emissions to achieve some clearly defined ambient air quality standards or goals. For an efficient AQMS definition a decision support system is needed [Hussain, 2003]. AQM should be detected the pattern of each pollutant in space and time dimensions. In this regard we monitor air quality in ten stations and next preparation of air pollution modeling.
A geospatial information system (GIS) is a computer-based information system which enables to capture, model, manipulate, retrieve, analyze and present the geographically referenced data [Aronoff, 1991].
It is quite suitable to utilize geostatistical analysis which is efficient for modeling ambient air quality.
2. Air pollution Modeling
Several air quality modeling techniques have been developed to tackle different aspects of air pollution, ranging from metrological conditions to the economic costs of abatement. There are five categories such as: Diffusion-based, economic optimization, spatio-temporal, neural networks and acid rain models.
Through resent years spatio-temporal models is practical. Most Spatio-temporal models are regression model basically. Spatio-temporal model attempts to organize air pollution data in both spatial and temporal domain.
Interpolation is the procedure of predicting the value of attribute at unsampled site from measurements made at location within the same area or region.
Interpolation is necessary when the data we have do not cover the domain of interest completely. There isn't any action method for interpolation. And in order to data, sampling method, target of interpolation, we have several methods: Exact and inexact interpolation, global and local interpolation, deterministic and geostatistic interpolation.
Air pollution dispersion is local phenomena and we prefer that the model shows real measurements in sample points. In this regard we should use exact and local method of interpolation. Therefore Inverse distance and kriging interpolation is used for modeling between the monitoring stations.
When data are abundant, most interpolation techniques give similar results. When data are sparse, however, the assumption made about the underlying variation that has been sampled and the choice of method and its parameters can be critical if one is to avoid misleading results.
3.1. Inverse distance weight interpolation
The assumption of this method is that the value of an attribute at some unvisited point is a distance-weighted average of data points occurring with in a neighborhood or window surrounding the unvisited point. This model computes these formulas [Barrough,1998].
3.2. Kriging interpolation
Geostatistical methods of interpolation, popularly known as kriging attempt to optimize interpolation by dividing spatial variation into three components: deterministic variation, spatially autocorrelation, and uncorrelated noise. Kriging fulfils the aims of finding better ways to estimate interpolation weights and of providing information about errors. This model computes these formulas [Barrough,1998]:
We calculate weights with two constraints. Sum of all weight should be one and estimation variance should be at least.
Kriging is an unbiased with minimum variance estimator. Error estimation in kriging interpolation related to distribution of sampling points.
3.3. Accuracy Evaluation
It is very important to understand the nature of errors in spatial data. In this case we have a lot of source of errors in data such as: measurement error, positioning error, modeling error…
In this paper we assume that we don't have any error in measurements and positioning of stations and concentrate on modeling error.
Accuracy of interpolation can be calculated by check point or cross validation method. If we have a few sampling point, we use cross validation. In this method we omit a sample point each time and then we interpolate with other point. It is necessary to mention we estimate omitted point as well. Fore there more during this presses we get real and estimated value for each sampling point. We can evaluate accuracy of interpolation with computing RMSE.
4. Case study
Air quality changes continually and we can monitor different values over the time for each critical pollutant such as CO, NO2, SO2 and TSM. In Tehran, ten monitoring stations, monitor air quality by Automatic Analyzer methods in half an hour time period, so we have large amount of changed data. In this paper we concentrate on CO pollutant because CO is very critical pollutant In Tehran. Table 1 shows properties and statistical analysis of our data.
Table 1:Properties and satatistical parameter of air quality data
Obtained data from monitoring stations are points wise, so we have a lot of spatial gap between monitoring stations. For modeling air pollution data we use ArcGIS software and customized it by Visual Basic. We model data with kriging and IDW methods and after that, we use cross validation method for evaluation of model. Mean prediction error and root mean square prediction error was calculated for both of them. The numeric results of two different method of interpolation are shown in table 2.
Table 2:The numeric result of accuracy evaluation
The results show Kriging interpolation is much better for CO pollutant. When data are sufficient to compute variograms kriging provides good interpolator for sparse data but it is time consuming.
Because predictions are not true values, the uncertainty associated with predictions should be provided. Figures 1a is kriging interpolation and 1b is examples of prediction standard error maps, quantified by the minimized prediction root mean squared error that makes kriging optimum.
Figure 1: a)Kriging interpolation of CO pollutant b) Prediction error map of CO pollutant
5. Conclusions and recommendations
The air pollution problem originating from the various sources can be controlled by the development of air quality management system. TGIS is a powerful environment for AQM because the air pollution is related to location and time.
The air pollution data are point wise so we need an optimum model for other place. Obtained data from monitoring stations are points wise, so we have a lot of spatial gap between monitoring stations. Therefore interpolation is used for modeling between the monitoring stations.
There are a lot of interpolation method therefore accuracy of each interpolation should be checked then we can choose the best one for this case. Kriging interpolation is suitable for CO pollutant.
For data interpolation of air pollution in Tehran, we have a limited number of measurements of several air pollution variables and detailed information on geographic variables, such as elevation, distance to the ocean, and distance to the road (this variable is useful because cars are major sources of air contamination in the most parts of Tehran). Among the possible approaches to interpolate multivariate data is cokriging, which combines spatial data on several variables to make a single map of one of the variables. It is appealing to use information from other variables to help make predictions, but it comes at a price. Cokriging requires more estimations than kriging, including estimating the autocorrelation for each variable as well as all the cross-correlations.
- Aronoff, S., 1989, "Geographic Information Systems: A management perspective", Ottawa, WDL pub.
- Burrogh, P.A., and McDonnell, R.A., 1998b, "Principles of Geographic Information Systems", Oxford University Press. 1
- Hussain, F. 2003. GIS in AIR QUALITY MANAGEMENT, www.iiitmk.ac.in/iiitmk/newsnevents/technoweek2003/paper
- Johnston, K., Ver Hoef, J.M., Krivoruchko, K., Lucas, N., 2001, "Using ArcGIS™ Geostatistical Analyst", GIS by ESRI™
- New Zealand Ministry for the Environment, 2000, "Good-practice guide for air quality monitoring and data management", Air Quality Technical Report No.10.
- Rahmatizadeh, Sh., Delavar, M. R., Motessadi, S. 2004, Design and Development of a Customised GIS for Air Quality Management, ISPRS Congress, Istanbul, Turkey.
- Rahmatizadeh sh., Mesgari, M. S., Motesaddi S., 2005, "Monitoring of Air Quality (in Tehran) by Analyzing Spatio-Temporal Data", Map Middle East Congress, GIS Development, Dubai.