Characteristics of TRMM/PR System Noise and Their Application to the Rain Detection Algorithm

Abstract

This study investigates the characteristics of system noise sampled by the Precipitation Radar (PR) onboard the Tropical Rainfall Measuring Mission satellite. First, we examine the long-term trend in system noise. The result shows that the system noise level was quite stable for a long time period except for a few sudden change events. The averaged system noise shows a periodic change that relates to the temperature of the PR. This change corresponds to the changes in the solar beta angle, and the amplitude of the fluctuation is about 0.15 dB. The contribution of surface emission from the ocean to the system noise is also examined by averaging the system noise data under a no-rain condition over a long period (e.g., one month) to remove the effect of the temperature fluctuation related to the solar beta angle. The resulting amplitude is less than 0.1 dB, and the spatial distribution almost corresponds to the pattern of the sea surface temperature. These results confirm that the system noise of the PR is quite stable, indicating that the fluctuation of the sampled system noise is caused mainly by the fading effect. The characteristics of the system noise of three rain categories - ldquorain certain,rdquo ldquorain possible,rdquo and ldquono-rainrdquo - are examined by taking a histogram. Because the PRpsilas system noise is supposed to be stable, broadening the system noise level of about 2.5 dB almost corresponds to the fading effect. The histograms of rain-certain and of no-rain show a similar shape with a difference of a few tenths of decibels. The higher noise levels in rain-certain conditions are caused by the emission from raindrops. In contrast, the rain-possible histogram shows a widespread and skewed shape. It can be explained by saying that the rain-possible pixels are obtained when the sampled noise level is made accidentally small by the fading effect. Therefore, a fixed noise level threshold is introduced to the rain/no-rain classification for a more reliable rain-possible classification. The result shows that the rain area increases by about 15%, and the amount of rainfall increases by about 1% by adding reliable rain-possible pixels to the current standard product.