Infrared and Visible Satellite Rain Estimation. Part II: A Cloud Definition Approach

View More View Less
  • 1 Laboratory for Atmospheres, NASA/Goddard Space Flight Center, Greenbelt, MD 20771
© Get Permissions
Restricted access

Abstract

This study examines the relationships between satellite infrared clouds and rainfall, and infrared-threshold visible clouds and rainfall. Clouds are defined by the outline of the 253 K isotherm. Cloud infrared area was highly correlated with rain area (ρ = 0.85) and with volume rainrate (ρ = 0.81). It was poorly correlated with mean cloud rainrate (ρ = −0.28). One-parameter models were as effective in explaining the variance of cloud volume rainrate as multiparameter methods, due to the high correlations between visible brightness, mean cloud temperature and cloud area. An exception was found for clouds >10 000 km2, where area and temperature were uncorrelated, and mean temperature was more effective in discriminating among classes of volume rain than was cloud area. Statistical separation of five- of six-volume rain classes was achieved with mean temperature; however, the probability of occurrence of the classes effectively reduced this to a four-class problem.

Due to the high correlation between visible brightness and infrared temperature, visible data provided largely redundant information. Using a mean cloud brightness threshold of 148 counts, rain/no-rain separation was effected with a POD, FAR, and CSI of 0.98, 0.13, and 0.86, respectively. An infrared threshold (mean temperature of 241 K) produced statistics of 0.88, 0.07 and 0.83, respectively for the POD, FAR and CSI. The standard deviation of visible counts (used as a measure of cloud structure) was poor in explaining the variance of rainrate, yielding no better than rain/no-rain separation.

Time series of the cloud evolution showed that rain volume fluctuations were better “mirrored” by cloud temperature fluctuations than by cloud area. Contrary examples could be found and inconsistency between days was noted. The apportionment of rain volume (assigning rainrates to areas) remained a difficult problem, with significant variability, both within clouds of the same size and between clouds of different size. The coldest 10% cloud area was found to contain 11%–23% of the total rain volume while the coldest 50% area contained 60%–70–. This is in contrast to the rain apportionment used in the Griffith-Woodley Technique.

Abstract

This study examines the relationships between satellite infrared clouds and rainfall, and infrared-threshold visible clouds and rainfall. Clouds are defined by the outline of the 253 K isotherm. Cloud infrared area was highly correlated with rain area (ρ = 0.85) and with volume rainrate (ρ = 0.81). It was poorly correlated with mean cloud rainrate (ρ = −0.28). One-parameter models were as effective in explaining the variance of cloud volume rainrate as multiparameter methods, due to the high correlations between visible brightness, mean cloud temperature and cloud area. An exception was found for clouds >10 000 km2, where area and temperature were uncorrelated, and mean temperature was more effective in discriminating among classes of volume rain than was cloud area. Statistical separation of five- of six-volume rain classes was achieved with mean temperature; however, the probability of occurrence of the classes effectively reduced this to a four-class problem.

Due to the high correlation between visible brightness and infrared temperature, visible data provided largely redundant information. Using a mean cloud brightness threshold of 148 counts, rain/no-rain separation was effected with a POD, FAR, and CSI of 0.98, 0.13, and 0.86, respectively. An infrared threshold (mean temperature of 241 K) produced statistics of 0.88, 0.07 and 0.83, respectively for the POD, FAR and CSI. The standard deviation of visible counts (used as a measure of cloud structure) was poor in explaining the variance of rainrate, yielding no better than rain/no-rain separation.

Time series of the cloud evolution showed that rain volume fluctuations were better “mirrored” by cloud temperature fluctuations than by cloud area. Contrary examples could be found and inconsistency between days was noted. The apportionment of rain volume (assigning rainrates to areas) remained a difficult problem, with significant variability, both within clouds of the same size and between clouds of different size. The coldest 10% cloud area was found to contain 11%–23% of the total rain volume while the coldest 50% area contained 60%–70–. This is in contrast to the rain apportionment used in the Griffith-Woodley Technique.

Save