6.4
Spatial and temporal resolution and accuracy of meteorological parameters
G. Louis Smith, Virginia Polytechnic Institute and State University, Hampton, VA
Satellite measurements are used to generate data products describing a variety of meteorological fields. These data products often take the form of maps with specified spatial and temporal resolutions. In addition, mission requirements contain accuracies which these products must attain. Outgoing longwave radiation (OLR) is used for the examples, though the considerations apply to a broad range of parameters. A meteorological parameter has a spatial variability which can be described in terms of its spatial spectrum. This spectrum is associated with a time averaging period. For an ÒinstantaneousÓ map, the spatial variability is quite high and the spectrum maintains a high level through large wave numbers. In order to describe this detail, a fine grid is required. When one computes the monthly mean map, the resulting map is much smoother because features moving across that map are Òsmeared outÓ in the average. The spectrum of the monthly-mean map decreases much more rapidly with wavenumber than does the Òinsta The variability of the parameter within a grid box is computed in a root-mean-square sense by use of the spatial spectrum. The RMS change of the parameter from one grid box to the next can be computed also. There are errors in the grid-box values due to the instrument, parameter retrieval, spatial averaging and temporal interpolation or averaging. If the random error in the grid-box value due to these sources exceeds the box-to-box variability, the resolution is greater than is required to describe the parameter field. For a monthly-mean map based on limited spatial or temporal sampling, the data product can in some cases be improved by combining grid boxes such that the random errors are reduced more than the box-to-box variability. In this case, the reduced resolution data product gives a better description of the field than does the higher resolution map. There is thus an optimal spatial resolution for a map of meteorological parameters, which depends on the spatial variability of the field for the time-av
Session 6, Probability and statistics in remote sensing
Friday, 12 May 2000, 8:00 AM-10:00 AM
Previous paper Next paper