EFFECTS OF THE RADAR OBSERVATION PROCESS ON INFERRED RAINFALL STATISTICS

Contrary to the popular notion that radar can measure rainfall, we dem onstrate that both the physics of the radar measurement process as wel l as the processing of radar data can have significant effects on infe rred statistics of rainfall. Only the following effects were considere d: (1) the radar beam averaging scheme, (2) the nonlinear transformati ons inherent in estimation of rainfall based on radar observations, an d (3) the coordinate transformation from a polar to a Cartesian grid. The statistics included the following: mean, variance, spatial covaria nce, and a certain scaling parameter. We limited our considerations to spatial aspects of the problem. The methodology used in this paper is based on analysis of two radar rainfall estimation algorithms and a n umerical Monte Carlo simulation. A series of simple experiments allows us to consider the above effects individually and collectively. In th e experiments, the true rainfall process is simulated using a variety of statistical models. Its statistics are compared to those obtained a fter the radarlike processing has been applied. The models used to sim ulate the rainfall field include point process models, which conceptua lize mesoscale rainfall organization according to marked Poisson proce sses, a random cascade model, and an iterated random pulse model, whic h is a combination of the above two models.