Electrical Engineering-Systems Dept.

סמינר

Jonatan Ostrometzky

(M.Sc. student under the supervision of
Prof. Hagit Messer - Yaron)

on the subject:

Robust Precipitation Estimation From Commercial Wireless Microwave Links

In recent years, the field of accurate precipitations monitoring has become a major issue with global importance. While the common practice for precipitations estimation involves the usage of rain gauges and terrestrial radars, a revolutionary new method of using commercial Microwave Communication Networks (MCN) for precipitation estimation has been studied. It has already been shown that measuring the Received Signal Level (RSL) from commercial cellular networks backhaul Microwave Links (ML) can be used in order to estimate the amount of rainfall (Messer et. al., 2006). Both rain gauges and ML – RSL data have been proven as good and trustworthy methods for liquid precipitations (i.e. rainfall) monitoring. Radar systems, on the other hand, can detect other types of precipitations as well. However, radar systems are known for their inability to measure near-ground precipitations. In addition, studies for dry-snow precipitation and ML attenuation relationship have been carried out. But, once sleet (i.e. a mixture of rain and wet-snow particles) is introduced into the medium, the near ground precipitation monitoring technique is lacking.

            Since the redundancy requirement from MCN usually dictates the need for numerous ML roughly in the same area (especially in populated regions), multiple data entries from the same event could be gathered. We have considered this characteristic and developed a new estimation model, which takes advantage of the redundant MCN design. Doing so, we were able to establish a method for robust precipitations estimation, regardless of the specific water phase (liquid, solid or a mixture of both). The suggested model is actually a parametric estimation problem, which needs a minimum of four data sources (I.e., four ML – RSL Attenuation data). This new approach of robustness opens the door for new precipitations measuring techniques, in which the most important size – the total amount of fallen precipitations could be monitored directly, even when multiple types of different precipitations fall simultaneously.

            The model was tested on multiple ML – RSL data, during 3 major weather storms spanning from December 2010 to January 2012. The results show a major accuracy improvement in the overall precipitation estimation. When sleet is introduced into the medium, the estimation error (MMSE) drops by 20%.