NARSTO Annual Executive Assembly Air Quality Forecasting Activities in Mexico Rafael Ramos-Villegas rramos@sma.df.gob.mx Director of the Mexico City Ambient Air Monitoring System GOBIERNO DEL DISTRITO FEDERAL México, La Ciudad de la Esperanza
Air Quality Monitoring in Mexico City 1967 1973 TSP samplers 4 sites Impringers 14 sites
Air Quality Monitoring in Mexico City 2006 UAM Monitoring Station
Air Quality Monitoring in Mexico City The ultimate result of these performance audits indicates that the GDF monitoring system is functioning well. * * Report from the Performance Audit of the Mexico City Ambient Air Monitoring Network, US EPA, 2004. p. 8. http://www.sma.df.gob.mx/simat/pdf/auditoria_ig03.pdf
Software by: Ozone Map October 15, 1999 Episode
Air Quality Forecasting Milestones 1986: Forecasts based on daily local radiosonde (at 12:00 Z). 1988: Incorporation of Synoptic Charts and Satelite Imagery in the forecasts. 1990: Incorporation of weather forecasts. 1992: Temporary use of a statistical model based on linear regressions and another based on neural networks Not transferred from developer. 1993: Temporary use of an Expert System based on expert meteorologist knowledge rules Discarded, no results. 1993: Incorporation of the Mixing Layer Height provided by a sodar. 1994: Refining the traditional procedure by the incorporation of real-time Internet data and information.
Empirical Air Quality Forecasting System Radiosonde Isobaric Maps Thermodynamic Diagram Satellite Imagery Early morning pollutants concentrations Weather forecasts Other considerations
Empirical Air Quality Forecasting System * Forecast
Empirical Air Quality Forecasting System Difference Average = 33 IMECA * Forecast Registered - Forecasted O 3 IMECA
Improve Air Quality Forecasting
US EPA s proposed Steps! Step 1. Understanding Users s Needs! Step 2. Understanding Air Quality Processes " Step 3. Choosing Forecasting Methods " Step 4. Data Types, Soures and Issues " Step 5. Forecasting Protocol " Step 6. Forecasting Verification
Forecasting Methods Chosen Multiple-Variant Linear Regression Analysis Classification And Regression Trees - CART
Advances in the Implementation # Complete time series by interpolation and other techniques. # Comparison of software packages (SAS, SPSS and Eviews) # Identify variables (precursors and meteorology) for the linear regression models # Anaysis of Principal Components among the variables to identify correlations with ozone levels. # Ajust time series models to begin forecast precursos and ozone.
Chosen Variables " Day of the week " Month of the year " Years (2001 2003) " Previous maximum of O 3 " Precursors: NOx, NO 2 and CO " Surface temperature, wind speed and wind direction " Mixing layer (max.) " Temperature (700 and 500 mb) at 00:00 Z and12:00 Z " Dew Point (700 and 500 mb) at 00:00 Z and12:00 Z " Wind Speed and Direction (700 and 500 mb) at 00:00 Z and12:00 Z
Future steps $ Complete missing meteorological data. $ Conduct a cluster analysis to identify associations among precursors, meteorological variables and ozone. $ Evaluate the output of the MM5 model for the climate forecast. $ Adjust the time series of meteorological variables for a best estimate model. $ Compare the MM5 climate forecast and time series with the real observations. $ For future reference, define a typical wind vertical profile to compensate for lack of data.
Advances in the Implementation Temperature WD Ln[O 3 ] = f WS Day of the week Month O 3 day before CO, NOx and NO 2
Conclusions The Mexico City Ambient Air Monitoring System is developing a stochastic ozone forecasting system in order to estimate the maximum values of ozone for the next day in 5 zones of the City. The methodology used is that proposed by the US EPA. The advance in the implementation is around 60%. A full operative forecast system is expected to be ready in the second semester of 2006. This forecasting system will complement the deterministic and the empirical ones in use. Developing our own forecasting system provides us with a in-depth understanding of the processes involved and will allow to undertake more complex systems in the future. Our capacity building will be shared with other air quality managers in Mexico and Latin America.
Muchas Gracias por su Atención