A simplified model for the study of smoke plume dispersion from grassfires and a methodology for forecasts validation with satellite images

Forest and grassland fires are responsible for the deterioration of ecosystems, often the recovery may require several decades, and they also have a major impact on population and social activities. Every year in Argentina a variable number of wildfires take place especially during summer over dry regions, and in occasions they are due to prescribed burnings in rural zones, which become uncontrolled and remain so for several days.

This study focuses on the smoke produced by pasture burnings that took place during April and May 2008 in the Paraná River Delta, some 70 km to the northwest of the city of Buenos Aires. The smoke however, propagated over wide regions according to the prevailing atmospheric conditions for each day. Smoke spread hundreds of kilometers to the north across the La Plata River to Uruguay and southern Brazil, and as far south as the extreme south of Buenos Aires province. The grassfires burnt out about 70,000 hectares in the provinces of Buenos Aires and Entre Ríos.

Within that period, an extreme event without historical precedent occurred during 16 to 20 April 2008 affecting the city of Buenos Aires and its suburbs, with an approximate population of 13 million people. The episode resulted in an increase of health problems among the population (respiratory problems, eye irritation, etc.) and, due to visibility reduction; there were hazardous driving conditions and accidents that forced the intermittent closure of highways, as well as the inoperability of airports. During those days, the persistence of anomalous northwesterly winds contributed to such situation, which under normal conditions for that time of the year would have not occurred.

The purpose of this work is to present the smoke plume dispersion model HIRHYLTAD (HIgh-Resolution HYbrid Lagrangian Trajectory and Atmospheric Dispersion), and a methodology for the objective validation of smoke plume forecasts.

In order to do that, high resolution imagery from AQUA and TERRA satellites are used, which are equipped with the Moderate-Resolution Imaging Spectroradiometer (MODIS) instrument. True-color images (one per day for each satellite) are available from the MODIS Rapid Response System at NASA/GSFC. The AERONET_CEILAP-BA subset is used (centered in the La Plata River basin), as it best covers the affected region for the analysis. This dataset of images is the most appropriate for the study of local scale atmospheric phenomena such as smoke dispersion in the boundary layer.

Two meteorological models are used as input for HIRHYLTAD, the operational regional Eta/SMN model from the Argentine Servicio Meteorológico Nacional (SMN), and the Mesoscale Boundary Layer Model (MBLM) especially designed for the La Plata River region.

The MBLM is a primitive equation, dry and hydrostatic model, with a high resolution definition of the water-land temperature gradient. The upper boundary condition includes the geostrophic wind temperature at the top of the model. These data can either be taken from the local radiosonde observation, or from the Eta/SMN model outputs.

The smoke plume forecasts are performed with the HIRHYLTAD puff model, which has two calculation modules. The first one determines the smoke lines using lagrangian trajectories, while the second one performs the dispersion calculations and determines concentrations. The model uses the Pasquill stability classes, the Pasquill-Gifford dispersion coefficients according to the stability class, and the Briggs scheme for buoyant plume rise. HIRHYLTAD is based on a modified Gaussian dispersion equation in order to resolve a discrete problem in time and pace, that requires the use of a non-homogeneous and non-stationary wind fields. The advantage of HIRHYLTAD in comparison with other dispersion models widely used such as HYSPLIT, is that it requires minimum data, i.e. wind, temperature and cloudiness.

The methodology for this work involves adjustments of MODIS images, digitalization of smoke plumes, definition of smoke events, selection of events, determination of amount of smoke sources for each event, and definition of initial and final time of emission for each source.

Different indices of simple and direct interpretation are employed for the validation: the plume matching area hit index (PMAH), the plume overestimation area error index (POAE), and the plume mean orientation error index (PMOE). The reason for including these three indices is that each one focuses on a particular aspect.

Results for the 59 analyzed events (with 165 smoke sources associated) for the 45-day period considered, show a better overall performance of forecasts obtained with the Eta/SMN model over both versions of the MBLM model. There was a minor tendency towards underestimation of plume areas. In the best experiment (diagnose mode for Eta/SMN model) the PMAH index has a median of 56.5% and an average of 53.4% [0% = no match, 100% = perfect match], and the results for the PMOE index are less than 10% for both mean and median values [0% = observed and forecasted plume with same orientation; 100% = with opposite direction (180º difference)].