Seasonal Average of Occurrence Probability

VFM Level 2 swath data is gridded into a three dimensional latitude, longitude, and height matrix with dimensions of 1˚ x 1˚ x 30m for 290 levels with a range of -0.5 km to 8.17 km.  The rectangular domain of interest covers 40S - 60N and 60E - 100W. This domain includes regions that are known for their aerosol sources and concentrations, such as Africa, the Atlantic Ocean, South America, Caribbean, and the eastern United States. There are 161 x 101 x 290 horizontal and vertical grid points in this 3-D domain. At each grid point, the occurrence of dust, polluted dust (dust/smoke mixture), and smoke are tallied.  The total occurrence number of each types of aerosol is then normalized by the total number of valid satellite passes at that grid point.  A valid satellite pass is defined here as a VFM indication of clear air, aerosol, or cloud. 

To minimize the effect of anomalous aerosol identifications, a grid point is considered invalid and treated as NAN if its total number of valid satellite passes is less than a threshold.  This threshold is defined to be 15% of the maximum number of valid satellite passes through the domain during a season over the observing period.  The number of valid satellite passes is highly variable horizontally and vertically, with highest values at upper altitudes.

The normalized data were then smoothed to highlight coherent large-scale structures.  First a 13-point zonal running mean finite impulse response (FIR) filter was applied.  For this, data outside the analysis domain were used for the grid points near the boundaries.  A second smoothing was applied using a 3-point meridional running mean FIR filter in a similar manner.

The thus processed data yielded a 3-dimensional (3D) structure of occurrence frequency or probability.  It is visualized by meridional-vertical cross-sections at every 10˚ longitude.  Averages in the zonal, meridional, and vertical directions are projected onto the right side, back, and base of the figure, respectively. Cross-sections are projected below the main figure to better observe the vertical structure.  Polluted dust contours are added to individual cross-sections upon selection.  Probabilities less than 0.15 are not plotted so a cohesive structure can be well observed. 

Seasonal Cycles of Occurrence Probability

In addition to seasonal averages, two types of 3D seasonal cycle in the occurrence probability were produced. One is a zonally integrated 3D structure as a function of latitude, height, and time. The other is meridionally integrated 3D structure as a function of longitude, height, and time.  The VFM data were first tallied into a daily 3D array at each latitude (1˚), longitude (1˚), and height (30 m) over the same analysis domain.  For the zonally averaged seasonal-cycle structure, occurrences in longitude (x) are summed and normalized by the total valid satellite passes in a given longitudinal range (not the entire longitudinal extent of the whole domain - more information is given later):

where p is the resulting occurrence probability at the grid point, s is the total number of valid satellite passes at the same grid point, and N is variable based upon the number of grid points in a specified longitudinal range.  The 3D dataset is then smoothed twice using a 29-day running mean FIR filter and once using a meridional three point running mean FIR filter.  The meridionally integrated seasonal-cycle structure was produced in a similar manner with the exception that the sum and normalizing is imposed over the meridional (y) direction yielding p = p(x,z,t).  There are small numbers of valid satellite passes at each grid points for the seasonal-cycle structures, so it is not necessary to impose a threshold for satellite minimum passes as for a seasonal averaged structure.  Probabilities less than 0.15 are still not shown in a 3D plot so the large-scale cohesive structure is better seen.

**CALIPSO data were obtained from the NASA Langley Research Center Atmospheric Science Data Center.

**TRMM data were obtained from the NASA Langley Research Center EOSDIS Distributed Active Archive Center.