FLOOD05893 - Laserfiche WebLink

Home Browse Search

378 PJ. BRtMAUD AND Y.B. POINT1N 'I I I , 30 so .. 70 60 , RADAR 20 40 -10 30 -2 -30 .I' 10 o 10 30 60 40 so 2ll East-West distance in km. Fig. 3. eEL echoes detected on the image observed on II October 1988 al 10 h 42 min above the Cevcnnes (gray scale in dB). (quotient of the principal moments of inertia); inclination of the principal axis of inertia; previous detected motion. Tracked entity matching and detection of the different motions The matching procedure defined in the PARAPLUIE algorithm is based on our experience in trac,king rain cells using radar data. When looking at an ....iriini'ation of successive "t-adar images, we first detect the global motion of the radar echoes on the entire image, Then, for the detection of a particular cell motion, our attention focuses on the area where the radar echo detected in the previous picture is expected to be. In this area, we try to isolate the radar echo ..t FORECASTING HEAVY RAINfALL FROM RAIN CELL MOTION 379 which could correspond to the radar echo of the preceding image, One important criterion for this match is the constancy of reflectivity characteris- tics between the two pictures,' particularly the median reflectivity value. Finally, we look at other details such as pattern resemblance, We have introduced seven criteria for evaluating our confidence in the tested match between any two successive CEL echoes: two criteria related to the amplitude and direction difference between the tested individual motion determined from the above procedure and the mean motion established for all the CEL echoes; one criterion of reflectivity difference between the two tested eEL echoes; two criteria of pattern difference; one criterion of elongation difference; one criterion of inclination difference, For each tested match, these seven criteria are used to define the unique value of a confidence level which characterizes the likelihood of this match, In this confidence level evaluation, the seven matching criteria are given different weights according to their position in the list above (with the first criterion being the'most important) and according to their natural variability. The most important criteria are those based on the difference between the tested motion and the mean motion, followed by the criterion of reflectivity difference and then the two cnteria of pattern difference and the two 'inertia' criteria. . Therefore, to obtain realistic trajectories, the test of each matching CEL echo pair is stopped whenever the' displacement of the centroid is very different from the mean displacement of all CEL echoes, When the test proceeds, the new position of each CEL echo is precisely obtained by a local cross-correlation technique and then all the above criteria are evaluated so that the tested match can be accepted or rejected, , The CEL echoes tested in this matching procedure also include the union of CEL echoes (Einfalt et aI., 1990) to account for mergers and splits of such echoes, although these are less frequent than in the case ofT echoes, Indeed, the CEL echoes are usually associated with a particular air motion regime within the cloudy air and are more stable than T echoes, which can artificially separate because a saddle point reflectivity value may be close to the chosen threshold value. Furthermore, when a merger of CEL echoes does occur, it is particularly interesting for hydrological purposes because this can lead to meteorological structures such as a supercell which can produce heavy-rainfall and have a lifetime of more than I h, At the end of this matching procedure, the motion vector assigned to each CEL echo is equal to the arithmetic average of all instantaneous speeds of the same rain cell detected in the prevjous pictures, This time average value is less dependent upon the small pattern differences that result from the coarse spatial resolution of the radar data than the instantaneous speed would be,