by Bret A Schichtel and Rudolph B Husar
Center for Air Pollution Impact and Trend and Trend Analysis (CAPITA)
During the past two years a previous Monte Carlo model was reimplemented onto the IBM-PC platform. This model was designed in a modular framework, separating the emissions, transport and kinetics calculations. The transport module employs a Monte Carlo technique for the simulation of atmospheric boundary layer physics. Kinetic processes are simulated using first order rate equations where the kinetic rate coefficients vary in space and time. The rate coefficients are determined via a tuning process comparing simulated and actual measurements. This paper presents the methodology used to implement the transport and kinetics processes; the architecture of the model addressing data flow and software implementation; and various applications including an initial tuning process for estimation of rate coefficients for the transformation and removal of SO2 and SO4 over the western U.S. during 1992.
The source receptor relationship is the impact of specific sources and/or source types upon a receptor concentration. This is accomplished by determining the role of meteorology and physical/chemical effects linking source emission to receptor concentrations. The understanding of the source-receptor relationship has become a central issue in air quality management. Only when the identification of relevant sources and their impact upon the receptor are known can a rational control policy be developed that will minimize costs of implementation and attain desired results.
Due to the importance of the source-receptor relationship it has been an active field of research for many years, and continues to be developed. Today, the primary processes involved in the relationship are fairly well understood. Based upon this understanding, a number of models and techniques have been developed for the quantification and study of this relationship. One such model, the CAPITA Monte Carlo Model, developed in the 1980's allowed for the full quantification of regional atmospheric transport, transformation, and removal processes governing the source receptor relationship. (Patterson et al., 1981 , Husar 1986).
Recently, the CAPITA Monte Carlo Model was reimplemented onto the IBM-PC platform. The implementation was conducted with special attention paid to "opening" the model up. That is, as opposed to treating the model as a black box, it was designed in such a manner as to allow for the viewing and analysis of the transport, transformation, and removal processes, to better understand their influence upon the source receptor relationship.
This paper will present the methods by which the transport and kinetic mechanism are implemented. However, the full physical justification and quantification of the impact of the simulated processes will be presented in subsequent papers. Also, the architecture and flow of data through the modeling system will be presented along with model applications and preliminary results.
The Monte Carlo approach to aerosol simulation differs from the more traditional Eulerian approach. The Eulerian approach is a numerical simulation of pollutants using the conservation equations to model transport, transformation, and removal processes. It is deterministic in that given a set of initial conditions, the model can estimate future concentrations fields. Alternatively, the Monte Carlo approach subjects individual quanta or "particles" to simulated physical and chemical processes as defined by a modeled environment. This approach is frequently referred to as a direct simulation, rather than indirect through the use of pertinent equations.
The Monte Carlo simulation of atmospheric pollutants has two key characteristics. First, each emitted quanta contains a fixed quantity of mass for various pollutants based on the source's emission rate. The quanta is then subjected to transport, transformation, and removal. Mass conservation is maintained at the quanta level by accounting for the mass that has been chemically transformed, and physically removed by dry and wet deposition as the quanta travels through three dimensional space. The pathway of the quanta arises from the integration of the equation of motion. This integration can be accomplished independently from the nature of the sources, and physical/chemical processes affecting the pollutant. The role of diffusion is simulated by releasing multiple pollutant quanta representing an airmass. The dispersion of the airmass is then represented by the spread of the quanta driven by transition probabilities based on horizontal and vertical mixing.
The Monte Carlo simulation can be viewed as Lagrangian in that the trajectories of individual particles are followed. However, in the limit of infinite particles, the solutions to the conservation equations of motion and mass can be obtained, thereby providing an Eulerian viewpoint. The Monte Carlo approach to the dispersions simulation over the mesoscale is well documented in the literature(e.g. Hall 1975 , Alsmiller et al., 1979, Patterson et al., 1981).
The following discussion presents the physical and chemical processes employed by the Monte Carlo model. The full physical justification and quantification of the impact of these simulated processes will not be discussed here.
The simulation of transport of pollutants is accomplished by transforming three dimensional wind fields from the Eulerian to Lagrangian view point. This is a conversion processes whereby, wind vectors representing the mean atmospheric flow on a fixed grid are used to advect the particle in space. Diffusion of an airmass in the transformation process is also directly simulated by employing a Monte Carlo process simulating the atmospheric behavior below and above the mixing layer.
Vertical Transport. The role of vertical transport in the fate of a pollutant is at least fourfold. Vertical exchange dilutes the pollutant throughout the mixing layer; it exchanges matter among atmospheric layers, each having its own velocity field, thereby causing regional scale dispersion of pollutants; it delivers matter to the top of the planetary boundary layer (PBL), where interaction with clouds may occur, creating catastrophic chemical changes for short lived pollutants; and vertical eddy exchange delivers matter to the surfaces for dry deposition.
Within the mixing layer, the vertical velocities are large enough that pollutants become roughly uniformly mixed throughout this layer in a matter of hours (Husar et al., 1978). This mixing is more intense during the day than at night. This is due to convective heating of the surface by the sun which produces buoyant thermals that establish the mixing layer to a height up to 3 km.
During the evening, with the reduction of the surface heating, the rise of buoyant thermals diminish, and the bulk of the former daytime boundary layer becomes stable. The lack of vertical exchange confines the movement of air parcels to shallow layers roughly parallel to the ground surface. The pollutants that have been part of the daytime mixing layer are now transported in horizontal layers with the wind direction and wind speed of their own layers. Unlike the daytime conditions, the nocturnal transport within the former mixed layer will exhibit substantial veer and shear: the top layers heading in directions and at speeds that may be markedly different from the movement of the bottom layer. Figure 1 presents a graphical depiction of this interaction between pollutant quanta (dots) released from a surface source and the diurnal cycle of the mixing layer in the time - height plane.
Figure 1. The dynamics of the planetary boundary layer, and
pollutant being emitting into this layer as simulated by the CAPITA Monte
Carlo Model.
Pollutants within the surface layer are trapped in that shallow layer, while new emission into this layer will accumulate resulting in generally high concentrations for a given emission rate. On the other hand, the matter that was either left from previous days or emitted at night with an effective stack height that is above the surface layer tends to float above the nocturnal mixing layer. Pollutants confined to the surface layer are mixed to and exposed to the absorbing ground surfaces and to the chemical mix of that surface layer. Thus, dry deposition and potential chemical reactions will significantly influence the fate of the pollutants. Matter remaining above the surface layer overnight will not be depleted by dry deposition, and its chemical reactions will only include species mixed during the previous day. This dichotomy in the physical-chemical fate provides another impetus for the separate simulation of pollutants within and outside the nocturnal boundary layer.
Simulation of Vertical Transport. For the simulation of pollutant transport within the planetary boundary layer, the above discussion provides two key inputs: first, an upper boundary or "lid" exists up to which convective mixing occurs, and second, the vertical daytime mixing occurs rather rapidly. The time scale associated with vertical mixing is t=H2/Kz (Gillani, 1978), which ranges from 30 minutes in the morning to several hours during afternoon conditions. During this time scale, pollutants emitted within the mixing layer tend to acquire a roughly uniform vertical concentration. In the Monte Carlo simulations, the rapid daytime vertical mixing is implemented by randomly changing the height of each particle per time step between the surface and the top of the mixing layer. This is done in accordance with a uniform probability distribution. All particles above the local mixing height are subjected to only the mean vertical motion.
A graphical illustration of how the above processes are implemented is shown by three trajectories in Figure 1. As shown, the stable layers during nocturnal hours are reflected in the lines of near constant height for each trajectory of the ensemble overnight. During the daylight hours the crossing lines indicate the repeated random height changes imposed upon each quanta during the convective daylight hours.
Horizontal Transport. The horizontal movement of pollutants from a source to a receptor is facilitated by the horizontal wind field provided by the meteorological input data. The horizontal transport of pollutants is then perturbed by random turbulent motion, particularly during daytime hours. This is simulated by a horizontal eddy diffusion coefficient, K, with magnitude from about 10m2/s in stable conditions to order of 1000 m2/s during intense convective activity. We use a value of 100 m2/s during nighttime hours and 2000 m2/s during midday hours. This diffusion process is implemented as a random walk displacement of radius for each time step in the model (Figure 2).
Figure 2. The methodology used to advect and diffuse a particle
in the horizontal direction.
Transport Simulation. The affects of the simulated transport can be seen in Figure 3. Figure 3A, displays the dispersion process on a single "puff" of pollution released from a source. The puff is simulated by multiple particles released at an initial time in Northern AZ, and the positions of the particles are plotted after each day of travel. As shown, the puff follows the mean flow to the North. However, due to the vertical and horizontal diffusion processes the puff continually spreads out in a Guassian like pattern.
Figure 3. A) The positions of quanta comprising a puff of pollution released
from a source after each day of travel. B) Back trajectories from a receptor
at the Grand Canyon showing flow coming from two distinct vertical and
spatial regions then converging in Northern New Mexico.
Often the diffusion of a plume is not a "well behaving" guassian process. Due to the vertical mixing during the day, different parts of the puff can be subjected to very different wind velocities that cause the puff to separate into distinctly different paths . This is demonstrated in Figure 3B where multiple back trajectories are used to simulate a receptor airmass. As shown, the airmass at the receptor is comprised of two airmasses that converged about a day before impacting the receptor. In the back trajectory simulation of this airmass, the particles were uniformly mixed throughout the mixing layer to a height of about 1 km, then at the convergence zone the airmass divided into a surface airmass (A) coming from the North and an upper air airmass (A) coming from the south. Consequently, within the column at the convergence zone the wind vectors were nearly 180 degree apart between the surface and about 500m above the surface.
Kinetic processes define the transformation and removal rates of pollutant species while in transit from the source to the receptor. In the Monte Carlo approach, this is accomplished by applying mass conservation principles to individual quanta. When each quanta is released, it is assigned species weights based upon the source's emission rate. The quanta is then subjected to transport, transformation, and removal. Mass conservation is maintained by accounting for the mass that has been chemically transformed, and physically removed by dry and wet deposition for each quanta (Figure 4). This is essentially an accounting procedure of the species weights as the quanta ages.
Figure 4. A depiction of the Monte Carlo approach simulating
the physical / chemical processes applied to a single quanta. The kinetic
processes are modeled using first order rate equations. However, the rate
coefficients defining these processes are dependent upon the chemical,
meteorological, and geological environment of the quanta. Consequently,
these coefficients vary with space and time creating non-linear transformation
and removal processes.
The derivation of the kinetic coefficient is accomplished via a semi-empirical approach (Husar et al., 1986). The main thrust of this approach is to make maximum use of existing meteorological, emission, receptor, etc. data in conjunction with a physical-chemical model to derive a set of rate constants. This is achieved by assuming fixed pollutant transport and emission rates, then "deriving" a set of rate coefficients that best fit simulated to measured concentrations.
It is hypothesized that given sufficient quality and quantity of data, it would be possible to identify a unique set of rate parameters, thereby simulating the physical/chemical processes. This approach is semi-empirical in that transformation and removal follow the deterministic rules of chemical and physical kinetic processes while mass is conserved. It is an empirical approach in the sense that to determine the set of rate constants, a fitting process to match simulation to actual concentrations is used. The resulting coefficients can then be used to quantify kinetic modulations within the source receptor relationship.
The resulting rate coefficients will vary due to changing meteorological, chemical, and quanta positions as the quanta travel from the source to the receptor. However, all coefficient must be within physical bonds. An illustrative example of the variation of the rate coefficients is demonstrated in Figures 5 and 6 where the transformation and removal of a unit mass of SO2 for a single quanta is shown as the quanta travels from the LA basin to Northern AZ. The minimum set of kinetic coefficients needed to describe the sulfur behavior of the quanta consists of five which are listed in Table 1. Also presented in this table, are some of variables that can be used to modulate each coefficient.
Table 1. Rate Coefficient and modulating meteorological variables.
Rate Coefficients Influencing Parameters SO2 oxidation rate;kt Clouds,Relative Humidity,Temperature, Solar Radiation, Location SO2 dry deposition rate;kd2 Scale height, Solar Radiation, Location SO42- dry deposition rate;kd4 Scale height, Solar Radiation, Location SO2 wet deposition rate;kw2 Precipitation rate SO42- wet deposition rate; kw4 Precipitation rate
Possible functional dependence of rate coefficients on these meteorological variables are:
kt = ktmin + ktmax*SR + kRH * RH kd = 1/H (Vdmin + Vdmax*SR) kw = W * Precipitation SR = (1 - % Sky Cover) Solar Radiation H = Scale Height Vd = Deposition velocity W = Washout Ratio The mass conservation equations for sulfur in an air column then becomes: 1) d(SO2) /dt = -(kt + kd2 + kw2) SO2 2) d(SO42-) /dt = kt SO2 -(kd4 + kw4) SO42-
Six of the modulating parameters are presented in Figure 5. An illustration of the affects of these parameters on the five kinetic coefficients, and the corresponding quanta sulfur budget is shown in Figure 6. The kinetic coefficients were calculated using the equations above. The varied affect of the meteorological conditions are clearly demonstrated in the first day of travel for the quanta. As shown, the quanta is outside of the mixing layer almost immediately after being released, and is not re-entrained for about 14 hours. Consequently, during the ensuing 14 hours, the quanta experiences no deposition and is subjected to only transformation. Also, during this period the quanta passes through a high humidity area with possible cloud interaction, enhancing the transformation rate of SO2 to SO42. After about a day of travel the quanta encounters a precipitating airmass which removes a substantial fraction of the SO2 and SO42.
Figure 5. The variation of six parameters along a quanta's
trajectory defining the quanta's meteorological state.
Figure 6. The kinetic rate coefficients describing the transformation
and removal of a SO2 quanta as it ages, and the corresponding sulfur budget.
The Monte Carlo model employs a systems approach to the problem of dispersion modeling. The systems approach allows the Monte Carlo model to be developed in a modular fashion employing data processors and browsers for the manipulation, creation, integrating, and viewing of multiple data sets.
This approach rests on the physical principle that the meteorological dispersion of pollutants proceeds independently from the nature and chemistry of the pollutant. This physical argument permits the formulation and computational execution of pollutant transport processes without the burden of pollutant chemical kinetics. Then the chemical kinetic computations can rely on pre-computed transport and diffusion computations.
Figure 7. presents a data flow chart demonstrating the flow of raw meteorological, emission, geographic, etc. data through the data processors for the creation of transport and chemical species databases. As shown, there are two distinct data flow paths. The first, transforms meteorological data into a Lagrangian database, simulating the transport of emissions of conservative species from sources to receptors. The second data flow path, combines the Lagrangian data with emission data and performs kinetic processes simulating the transformation and removal of atmospheric pollutants.
Figure 7. A data flow chart for the PC Monte Carlo Model.
The first step in the processing of the meteorological data involves converting the raw data from a random spatial distribution to a grid. Various institutions convert meteorological state variables collected by the National Weather Service to a grid format, and these data are publicly available. Currently, we use data derived from the National Meteorological Centers Nested Grid Model. The is mounting evidence that this data is biased in the Southwest during the summer of 1992 (Appendix A). The meteorological database grid is then converted into Lagrangian air mass histories by the Lagrangian/Eulerian transformer. Along each particle's trajectory path the meteorological variables as well as the particle positions are saved. These meteorological variables are selected such that they are relevant to aerosol formation and removal (temperature, dew point, solar radiation, time in clouds, precipitation, background chemistry, etc.).
Once the Lagrangian meteorological database has been prepared, it can be used as the driving database for the simulation of concentration fields. This is accomplished by integrating the Lagrangian database with chemical and emission fields. This creates new Lagrangian variables containing various species weights for each particle. Receptor concentrations fields can be created from the resulting database.
The main design paradigm of the software is the client-server architecture (Figure 8). The client is the part of the software that interfaces the user with the Monte Carlo model. It has two components, (1) the query input, which specifies the process to be computed, i.e. creation of airmass histories, simulated concentration fields, etc.; and (2) the data output and visualization that presents the resulting data files to the user. The client-server architecture follows the concepts and developments in modern relational database management systems (RDBMS).
Figure 8. Client-server architecture for trajectory software.
The client software uses multiple windows and forms to specify the query: For example the creation of an airmass history database can be accomplished by selecting the source or receptor location in a map view; a time view is used to select the time range; and the variables view for the specifications of the output variables. The output from the trajectory server would be a "hyperslab", i.e. a multi-dimensional array.
The server itself consists of a query processor and the Monte Carlo model processors. The query processors translates the input query into specific actions to be executed. The Monte Carlo model then performs the queried actions, such as calculating trajectories by the integrating the equation of motion using Eulerian wind fields. The output is a multidimensional array (hyperslab) which can then be display by rendering program in the client software.
At this time, scientific visualization is more of an art than a science. However, it is now well established that the data space to be visualized is composed of at least five dimensions: x, y, z, time, and variables. Standard computer displays , having only two dimensions, require facilities to slice and cut the data space into different cross-sections. The use of 3D-rendering, color, and animation can add further dimensions to the displays.
Eulerian Browser and Visualization. The current implementation of the browser for Eulerian meteorological data displayed in Figure 9. This figure was created using the data from the NMC model (Appendix A). As shown, the work space consists of a list of the variables in the database, as well as the three graphic data views: Map, Time, and Height, which characterize the dimensions of the data.
The Map View encompasses the left-hand side of the work space. In this view the spatial domain of the data is overlaid over a geographic map for a specified time and height. The spatial variation of one of the Eulerian variables for the specified time and height can be viewed by a contour. Also, wind vectors can be displayed for the same time and height.
Figure 9. The Eulerian Browser displaying the wind vectors
calculated from the NOAA meteorological data.
The Time View is located in the lower right hand side of the work space and displays the time series of a variable for a specified height and a geographic grid point. In Figure 9, the time series of the U component of the wind direction is displayed in two hour increments. The Height View displays the vertical profile of the selected variable for a specified time and a geographic grid point.
Lagrangian Query Processor. The specification of what domain trajectories are to be computed is set by the Lagrangian Query Process. Figure 10 displays the user interface for the Lagrangian Query Preprocessor. This interface is made up of multiple windows and forms to specify the trajectory query. The map view gives spatial context to the selection of receptors/sources. In this view an unlimited number of receptors/sources can be selected. Options exist to set the height of the receptors/sources. The time view is used to select the time range the trajectories will span, while the variables view specifies the output variables to be generated. A number of other inputs and options can also be conveniently adjusted and selected, i.e. the input Eulerian data to be used, trajectory length, diffusion parameters, etc.
Once the given inputs have been selected, the query processor uses this information to create a skeleton database in the form of a Hyperslab. This Hyperslab fully describes the trajectory database to be created.
Figure 10. The user interface for the Eulerian to Lagrangian
transformer.
Lagrangian Data Browser. The Lagrangian browser allows the viewing and exploration of Lagrangian data i.e. output from the Monte Carlo Model. The front end of the Lagrangian browser is displayed in Figure 11. This Figure was created from an airmass history database containing one year of back airmass histories at the Grand Canyon. The browser contains a Variable, Map, Time (release time), Time History, and a spatial slice view. The variable view contains a list of all of the variables in the airmass history database.
The Map view displays the location of the particles for a given release (back trajectories) or arrival time (forward trajectories). The particle positions can be displayed in either a Trajectory or a plume mode. The trajectory mode displays the spatial pathway of a ensemble of particles. The plume mode displays location of all of the particles which have been emitted from all sources at the currently selected Time. This viewing mode can be thought of as a direct simulation of a plume. User options exist to allow for coloring the particles based on the value of selected variables.
The Spatial Slice view (not shown) displays the height and longitudinal position of all of the particles for a horizontal cross section of the map. The position of the cross section in the map is determined by a query in the Map view.
The variation of a selected variable for all of the particles in a puff ensemble over the life time of the puff is displayed in the Time History view. The Time History displayed is for a single site, chosen in the Map view and variable. The Horizontal axis of this view is time. The time axis starts at the release or arrival time, and extends forward in time for a forward airmass history, and backwards in time for a back airmass history.
The Time view is located in the lower left hand side of the browser. It indicates the release or arrival time for which the data in the other view is to displayed. No data is displayed in this view, it is strictly for querying the database. The images generated by the Lagrangian browser are frequently arranged in a time sequence and assembled into an animated sequence.
Figure
11. The front end to the airmass history browser, displaying the multi-particle
trajectory at Hopi Point.
The systems approach to the implementation of the Monte Carlo model creates a flexible platform for the study of the source receptor relationship. By breaking the SRR into its primary processes: transport, kinetics, and emissions, influencing receptor concentrations, allows for the study of each one of these processes separately or in conjunction with each other. Below, several applications of the Monte Carlo model are presented which allow for the study of the SRR and the individual processes governing the relationship.
The Eulerian/Lagrangian transformer allows for the creation of airmass histories for individual locations at any 3-D point in the modeling domain. The airmass histories can be created from either a source or receptor view point. Thereby identifying the transit of pollutants from a source or an airmass to a receptor. The resulting data sets can then be used to characterize air flow and provide meteorological information of airmasses to the receptor. Such data is valuable in the interpretation and analysis of air quality data. Airmass history data have been previously used by multiple researchers (Ashbaugh et al., 1985. , Poirot and Wishinski, 1986. , White et al., 1994). Figure 12 present the airmass history created for a Grand Canyon receptor. The map view identifies the airmass pathway to the Grand Canyon while the time history views shows the variation in trajectory height, relative humidity, temperature, and precipitation.
Figure 12. Backward airmass history from a receptor at the
Grand Canyon.
The Eulerian to Lagrangian transformer can also be run in a mode where sources are uniformly placed over a geographic domain. Such a run is known as a uniform emission simulation. The resulting database can be accessed to animate and simulate atmospheric flow of potentially polluted airmasses over regional scales (Figure 13). The database can also provide a wealth of meteorological information characterizing the atmospheric transport, such as, spatial and temporal variations in atmospheric diffusion, residence times, wind directions, etc. This database is ideally suited for use as input into the chemical module of the Monte Carlo Model.
Figure 13. Map view and three vertical slices of the location
of particles released from a uniform emission source grid.
The Monte Carlo model can be run to solve the source receptor relationship for receptor concentrations over regional scales. This is accomplished by using the kinetic processor of the model coupled with an emission and Lagrangian transport database and a proper set of rate coefficients. The ideal transport databases to be used for concentration simulations are from uniform emission simulations. These transport databases contains all potential sources in a given region, so the transport data is not tied to any emission scenario. Consequently, as new emission scenarios arise the computer intensive task of the Eulerian to Lagrangian transformation does not need to be redone.
In addition to the uniform emission simulation transport database, specific point sources can be modeled, releasing particles from the source's location and stack height. This Lagrangian data set would be produced separately from the uniform emission data set. The two transport data sets would then be merged together in the process of calculating kinetics. The results of sulfate simulation over the western US is presented in Figure 14. This figure was created using constant rate coefficients of: 0.1%/hr transformation of SO2 to SO4, 1%/hr SO2 deposition and 0.1%/hr SO4 deposition. The time trend comparing modeled column concentrations to measure concentrations at Hopi Point AZ are also presented.
Figure 14. Simulated SO4 concentrations generated by the
CAPITA Monte Carlo Model.
Transfer matrices represent the potential mass transfer of unit emissions from a source to a receptor. Consequently, the impact a source has upon a receptor is determined by multiplying the transfer matrix element for that source by its emission rate. The receptor concentrations are then determined from the sum of the source contributions.
Once a Lagrangian database has been operated upon by the kinetic processes, it will contain variables of the concentrations for simulated pollutant species. From these new variables the transfer matrices can be computed for a number of receptor locations where a receptor location is defined by the meteorological grid used in the model. Sample transfer matrices for the Grand Canyon receptor during the winter and summer seasons are presented in Figure 15.
Figure 15. Transfer matrix patterns for a Grand Canyon receptor
in a)summer and b)winter.
The solution of the source receptor relationship in the forward mode produces receptor concentrations. Through the use of transfer matrices, the source receptor relationship can be inverted to produce emission fields. The inversion of the relationship is accomplished by using a numeric inversion scheme to invert the transfer matrices, then weight the receptor concentrations by the respective inverted matrix element.
Meteorological Data
The primary meteorological data set used to drive the Monte Carlo model were derived from the National Meteorological Center's Nested Grid Model (NGM) and obtained from NOAA. The meteorological data are in two hour increments, and the spatial domain covered is a 33 by 28 polar stereographic grid over North America. The data have a grid size of 182.9 km at 60 degrees latitudes and approximately 160 km at 35 degrees latitudes (Figure 1). This database contains meteorological variables for the surface and upper air heights (Table 1). The upper air data are positioned on ten sigma surfaces relative to the model terrain, from approximately 150 m to 7000 m. These data are available in 2 hour time increments from 1991 - 1995.
There is mounting evidence that the wind fields do not properly reproduce the flow in the Southwest during the summer of 1992. The reasons for this, and the extent of the wind biases are currently under investigation.
Figure 1. Location of each grid point for the NGM meteorological
database.
Table 1. List of surface and upper air meteorological variables for NGM meteorological database. Surface Variables
SURFACE VARIABLES UNITS
Ice Cover Water Flag 0/1 Snow Cover Flag 0/1 Terrain Height meters Mean Sea Level Pressure mbar Accumulated Convective Precipitation meters Accumulated Total Precipitation meters Exchange kg/m2/s Heat Flux W/m2 Water Flux kg/m2/s Surface Pressure mbar Mixed Layers Sigma
UPPER AIR VARIABLES UNITS
U component of Wind m/s V component of Wind m/s W component of Wind mbar/s Specific Humidity kg/kg Temperature Kelvin
| Submit your comments, feedback, questions, and ideas pertaining this page. Your input will be automatically added to the existing annotations. In order to add a new comment, you must be registered with the CAPITA People's Page. |
