Nowadays, humanity is facing natural disasters caused by climate changes, especially changes in precipitations patterns and global warming due to additional green house effect. These increases in temperature are for the most cases caused by industrial activities [1] which release radiative gases called green house effect gases in the atmosphere. These gases absorb the infrared radiation from the earth, reflect it back to the earth's surface, contributing thus to an increase in the mean air temperature on the surface. The climate prediction and the potential impacts of these warming on human being are the needs of the public and thus constitute a major challenge to scientists research.

The release of pollutants from multiple sources, most frequently by punctual sources, as stacks, affects considerably the air quality. Air becomes stagnant, trouble, thick, foggy and misty [2]. In order to estimate the concentration of atmospheric contaminants, two ways are possible: direct measure through monitoring sites or a use of an air dispersion model.

Unfortunately, in some areas, such as in African continent, installing a reliable air quality network is a challenge due to the economic conditions of the continent; accordingly air quality monitoring sites are not available. Thus the use of a model appears mandatory.

Numerical models have proved to be very useful tools in studying atmospheric mesoscale phenomena including air pollution transport [3] and atmospheric dispersion. Industrial releases, propagation of odor and automobile pollution make air quality improper to human being and environment. Air dispersion modeling has been recognized as a promising approach to predicting outdoor spatial and temporal variations of pollutants and the behaviors of these pollutants through mathematical algorithms that take into account atmospheric dispersion, chemical, and physical processes in an attempt to approximate concentrations of pollutants [4].

Researchers working in environmental exposure assessment are often interested in measuring concentrations of air pollutants at different time scales in large geographic areas. For example, environmental epidemiologists who are interested in studying the relationship between the proximity of maternal residences to air pollution sources and the health outcome of offsprings would require data about the concentration of a given pollutant at the daily or at least the monthly time scale. Air pollution data at these time scales are necessary for epidemiologists to determine the concentration of the pollutant at a location during a given time period such as the first three months of pregnancy which is considered to be the most critical time period of a healthy pregnancy [4].

There has been recently increased emphasis on air dispersion model. Before 1990, the most used dispersion model was the ISCST3 (see [5–8]). This model has shown some weakness and was judged inadequate for some regulatory uses. It is now replaced by AERMOD which is a new advanced air quality model and takes into account the characteristics of the ground cover of the site as input. As a state-of-the-art dispersion model for regulatory applications, AERMOD aims at modeling short-range (up to 50 km) dispersion from a variety of polluting sources (e.g., point, area, and volume sources) using a number of model configurations. These configurations include different sets of urban or rural dispersion coefficients as well as simple and complex topography. The model has the capacity to employ hourly sequential preprocessed meteorological data to estimate concentrations of pollutants at receptor locations at different time scales ranging from 1 h to 12 months [4]. AERMOD is an advanced plume model that incorporates updated treatments of the boundary layer theory, understanding of turbulence and dispersion, and includes handling of terrain interactions. The model was formally proposed by EPA in April 2000 as a replacement for the ISCST3 model. Several model enhancements were made as a result of public comment, including the installation of the PRIME downwash algorithm [9].

Several reasons account for the focus of this study. In Cameroon, precisely in Douala which is the most industrialized town in the country, the local circulation is strongly influenced by diurnal cycle of sea breeze [10] due to its location by sea. People are then exposed to many damages due to the transport and propagation of pollutants by wind. Generally in Douala city, industrial sites are implanted without any environmental impact and risk assessment. This study comes accordingly to specify some restrictions and to suggest some areas appropriated to industrial sites implantation so that pollution risk can be reduced. In addition, during the five past years, many people have been suffering from breathing diseases, specially students at school and pregnancy women in Douala city. Some investigations show that these health problems are due to short-term exposure to pollutants released from industrial sites around the schools and the hospitals [11].

Hight impact refers to conditions that affect or alter behavior by general public, government agency, and private sector activities, including health and the environment, aviation, surface transportation, electric power, public events, broadcasting, and emergency management [12]. Also, the atmospheric boundary layer is that region between the earth's surface and the overlying atmosphere where many local phenomena occur, including turbulence processes and free flowing (geostrophic) atmosphere. Heat fluxes and momentum drive the growth and structure of this boundary layer. The depth of the layer and the dispersion of pollutants within it, are influenced on a local scale by surface characteristics, such as the roughness of the underlying surface, the reflectivity of the surface (or albedo), and the amount of moisture available at the surface (Bowen ratio) [13]. It is important to know how sensitive the model results are to these parameters, so that their input values are characterized with sufficient accuracy for modeling purposes. This study evaluates the effects of meteorological fields and land use parameters on concentration patterns predictions from AERMOD in Douala city . We aim to provide recommendations to policies makers, precisely the Ministry of Environment of Cameroon and Non Government Organization (NGO), who are concerned about the acute effect of an air pollution in short and long-term exposure (3 h, 24 h and a months to years) assessment.

The U.S Environmental Protection Agency (EPA) in conjunction with the American Meteorological Society (AMS) has developed a new air quality dispersion model called AMS/EPA Regulatory Model (AERMOD). AERMOD is a gaussian model which uses three components: AERMET, AERMAP and AERMOD. AERMAP is used to compute the heights of the receptors and the sources. AERMET computes the parameters which are not measured, estimates necessary boundary layer parameters for dispersion calculation in AERMOD [6,14], which is the dispersion model. The run stream files used by AERMET require, in addition to the surface and upper air observation data, the characteristics of the terrain being modelled such as roughness length, albedo and Bowen ratio [14].

From a theoretical formulation, the forecast of pollutants concentration in the atmosphere may be done by solving a system of non-linear equation grouping advection-diffusion, Navier-Stokes, energy and continuity equations [15]. In the case of turbulent flow and using the K theory, we obtain the Equation (1): (1) ∂ ϕ ∂ t = − V → . ∇ → ϕ + K H ( ∇ → ϕ ) 2 + ∂ ∂ z ( K ∂ ϕ ∂ z ) + F → + R → + S →where ϕ is the mean pollutants concentration, V⃗ is the mean wind vector, F⃗ and S⃗ are the sink and source strength rate respectively, R⃗ the chemical reaction rate, K_H the horizontal eddy diffusivity and K is the vertical eddy diffusivity.

Equation (1) has many solutions depending on the nature of the source [16]. Thus, there are some parameters which have a direct effect on the evolution of the pollutants concentration in the air, such as the albedo, the Bowen ratio and the surface roughness length.

Physically, the albedo is the fraction of total incident solar radiation reflected by the surface back to space without absorbtion. Typical values range from 0.1 for thick deciduous forests to 0.90 for fresh snow. The Bowen ratio, an indicator of surface moisture, is the ratio of the sensible heat flux to the latent heat flux. Although the Bowen ratio can have significant diurnal variation, it is used to determine the planetary boundary layer parameters for convective conditions and can be computed experimentally [17]. During the daytime, the Bowen ratio usually attains a fairly constant positive value, which range from about 0.1 over water to 10.0 over desert at midday. The surface roughness length is related to the height of obstacles to the wind flow and is, in principle, the height at which the mean horizontal wind speed is zero. Values range from less than 0.001 meter over a calm water surface to 1 meter or more over a forest or urban area.

From these input parameters and observed atmospheric variables, AERMET calculates several boundary layer parameters that are important in the evolution of the boundary layer, which, in turn, influences the dispersion of pollutants. These parameters include the surface friction velocity u_*, which is a measure of the vertical transport of horizontal momentum; the sensible heat flux H, which is the vertical transport of heat to/from the surface; the Monin–Obukhov length L, a stability parameter relating to u_* and H; the daytime mixed layer height zi and the nocturnal surface layer height h; and w_*, the convective velocity scale that combines zi and H. All these parameters depend on the characteristics of the underlying surface [13]. To evaluate those parameters, AERMET considers two layers: the convective and the stability layers. From the energetic budget at the Planatary Boundary Layer (PBL), considering an ideal and a homogeneous surface [18], we can write: (2) R n = H + H l + H gwhere R_n is the net radiation, H is the sensible heat flux, H_l is the latent heat flux and H_g is the flux emitted or received by the soil.

In the Convective Boundary Layer, using Monin–Obukhov length [19] and [6], the friction velocity is given by: (3) u ∗ = k U ref l n ( Z ref Z o ) − Ψ m ( Z ref L ) + Ψ m ( Z o L )where k is the Von Karman constant, U_ref the wind speed at the reference height, Z_ref and Z_o the roughness length. Ψ m ( Z ref L ) and Ψ m ( Z ref L ) are functions. The turbulent convective speed in the Convective Boundary Layer (CBL) is: (4) W ∗ = [ g H Z i c ρ C p T ref ] 1 3where g is the gravity, C_p the specific heat at constant pressure and ρ the air density.

In the Stability Boundary Layer (SBL), due to important daytime and night variation in energy flux [20], parameters are experimentally estimated. [14,21–23] describe in details these equations.

Basic data required for dispersion medeling include meteorological data (surface and upper air data), pollutants characteristics and data describing the site being modelled [24]. Meteorological data constitute the main input data. Their use requires a good choice of the observation station, a convenable treatment and a limited time for the collection of data [25].

Hourly and statistical meteorological data used to run the model must be from a station whose meteorological characteristics are similar to that of the site being modelled. To show this similarity, the modeler must describe the design meteorological factors such as topography so that local meteorology should be known. In the case of a uniform terrain, the nearest meteorological station is the most adequate. At a distance more than 30 km, data will be acceptable if the site is homogenous and possesses a uniform topography. In this case, a whole description of wind field is required. All the meteorological parameters have to be taken at the same station. If meteorological data are not available, the modeler can use in situ data, on condition that he has at least one year of data. Ideally, we need 5 consecutive years of meteorological data to carry out a dispersion study. Raw data are passed through preprocessor such as AERMET to be quality assessed. To do so, data are classified into the Card Deck 144 (CD144) and Forecast Systems Laboratory (FSL) formats so that they can be read by the preprocessor. A Fortran utility program is used to convert surface data into the CD144 format. The output of AERMET is used as input to AERMOD which calculates the concentration of pollutants.

Daily in situ observations data have been collected from the Cameroonian's Ministry of territorial administration and decentralisation, direction of civilian protection. Since AERMOD just produces highest mean concentrations for the a given period (3 h for example) and indicates the dates at which they occur, we have collected data corresponding to the first 20 maximum average concentration values for a period of 24 h. The outputs discussed here is the base case presented in the Table 2. Figure 15 represents the observed concentration of NO_x against simulated. The model underestimates the simulated NO_x concentration. For the twenty data we recorded, the Root Mean Square Error (RMSE) is about 2.91. It comes out that the model does not reproduce satisfactory the real case pollution over Douala. This is due to lack of meteorological data.

The sensitivity of the AERMOD dispersion model to atmospheric fields and land use parameters has been investigated. We have shown that calm winds and anticyclonic conditions are not favorable to atmospheric dispersion. Air temperature and solar radiation play an important role: Cool air conditions decrease the volatility of some gases. Air concentration of primary pollutants are highest near the sources, and tend to decrease far from the source by dilution with ambient air [29]. Also, pollutants concentration decreases exponentially with respect to the inverse of distance from the source. The diurnal variation of albedo, Bowen ratio and roughness length are modulated by the solar radiation and energy budget at the top of atmosphere. An increase in albedo or roughness length leads to a decrease in pollutants concentration, but more significantly for the roughness length. Whereas an increase in the Bowen ratio increases pollutants concentration. Globally, albedo and Bowen ratio have less effects on concentration patterns than the roughness length, this is to be expected since albedo and Bowen ratio only affect the retention of incoming solar radiation, and therefore have no effect at night or just during convective conditions. Moving from 3 h to 24 h and to 5 years, pollutants concentration has significantly decreased. Thus, we recommend Cameroonian authority to move industrial sites from West site of Douala (Bonaberi industrial area, a popular quater of Douala city) to Nyala area (East site of Douala and not yet so more populated). Minimization (reduction) of environmental impacts are prior to a choice of land with a high roughness length and albedo. Thus, preliminary studies, meaning that an analysing of the land properties should be done before implanting industrial sites.