Radar reflectivity data is usually converted into precipitation intensities in a fairly straightforward way using a single Z-R relation and therefore not taking into account different precipitating cloud types and their associated different Z-R relations. The proposed algorithm seeks to distinguish between convective and stratiform precipitation types from operational radar products and consequently improve precipitation estimates. The algorithm also offers the opportunity to investigate the relevance of using space-borne radar systems for measurement of precipitation at different spatial- and time scales. Another use would be its function as an operational now-casting product.
The algorithm is comprised of two elements. One is the use of dBZ values to differentiate between convective and stratiform precipitation within a CAPPI. The second is the tracking of clouds and their development through time to again make decisions on what type of precipitation is being remotely sensed by the radar. The algorithm may also include echo top height and lightning intensity products as input for added accuracy.
The first part of the algorithm uses the work of Steiner et al. (1995) and that of Yuter and Houze (1997) to make a first guess about convective and stratiform areas within the measured precipitation. A summary of the used algorithm is:
Reflectivity values above 16dBZ are identified as precipitation.
All values above 40dBZ are assumed to be convective.
For pixels between 16 and 40dBZ, depending on the difference between the average of the surrounding pixels and the current pixel value the pixel and a fixed area around it are identified either as stratiform or convective.
The second part tracks the development of precipitating areas through consecutive images. This is not based on other work, but is similar to other cloud tracking algorithms using correlation. Within this part of the algorithm:
Separate precipitating areas within the CAPPI are identified.
These areas are tracked by correlating previous and the current radar image.
The correlation is handled by taking a frame around a precipitating area in the previous image and correlating this entire frame with part of the new image. The larger the identified rain cell is, the smaller the surrounding frame has to be, as the correlation will be easier to compute.
A database is maintained that keeps track of the history of each separate cloud and includes max intensity, size, direction of travel, total and average precipitation.
Keep track of merging and separating or dissipation of clouds.
If the cloud is small (one to 5 pixels) use its history to determine its most likely location and put a weight function around this location for correlation.
Depending on the growth speed of a cloud a decision can be made whether it is a convective or stratiform system.
Combining the two algorithms into one coherent structure.
Inclusion into the algorithm of an option for echo top height and lightning data input and investigate their effectiveness.
Divide larger clouds into sections to take into account possible cyclonality. Using other methods than the current correlation method might be preferred (Handwerker, 2003 and Mukherjee and Acton, 2002).
Use raingauge data and the high resolution volume data near the radar location around the Cabauw site for detailed case studies.
Do a full analysis of the algorithm at different spatial- and time scales to assess its performance and obtain the scale-dependent parameter input.
Handwerker, J.: Cell tracking with TRACE3D- a new algorithm. Atmospheric Research 61: 15-34. 2001
Mukherjee, D.P. and S. T. Acton: Cloud Tracking by Scale Space Classification. IEEE Transactions on Geoscience and Remote Sensing 40(2): 405-415. 2002
Steiner, M., R.A. Houze and S.E. Yuter: Climatological Characterization of 3-Dimensional Strom Structure from Operational Radar and Rain-Gauge Data. Journal of Applied Meteorology 34(9): 1978-2007. 1995
Yuter, S.E. and R.A. Houze: Measurements of Raindrop Size Distributions over the Pacific Warm Pool and Implications for Z-R Relations. Journal of Applied Meteorology 36(7): 847-867. 1997