Using Active Satellite Observations to Characterize Uncertatinty in Long Term Satellite Cloud Liquid Water Path Climatologies

Wednesday, 17 December 2014
Matthew D Lebsock, Jet Propulsion Laboratory, Pasadena, CA, United States
Bias between the Advanced Microwave Scanning Radiometer–EOS (AMSR-E) version 2 and the
Moderate Resolution Imaging Spectroradiometer (MODIS) collection 5.1 cloud liquid water path (Wc)
products are explored with the aid of coincident active observations from the CloudSat radar and the
CALIPSO lidar. In terms of detection, the active observations provide precise separation of cloudy from clear
sky and precipitating from nonprecipitating clouds. In addition, they offer a unique quantification of
precipitation water path (Wp) in warm clouds. They also provide an independent quantification of Wc that is
based on an accurate surface reference technique, which is an independent arbiter between the two passive
approaches. The results herein establish the potential for CloudSat and CALIPSO to provide an independent
assessment of bias between the conventional passive remote sensing methods from reflected solar and
emitted microwave radiation. After applying a common data filter to the observations to account for
sampling biases, AMSR-E is biased high relative to MODIS in the global mean by 26.4gm2. The RMS
difference in the regional patterns is 32.4gm2, which highlights a large geographical dependence in the
bias which is related to the tropical transitions from stratocumulus to cumulus cloud regimes. The
contributions of four potential sources for this bias are investigated by exploiting the active observations: (1)
bias in MODIS related to solar zenith angle dependence accounts for 2.3gm2, (2) bias in MODIS due to
undersampling of cloud edges accounts for 4.2gm2, (3) a wind speed and water vapor-dependent
“clear-sky bias” in the AMSR-E retrieval accounts for 6.3gm2, and (4) evidence suggests that much of the
remaining 18gm2 bias is related to the assumed partitioning of the observed emission signal between
cloud and precipitation water in the AMSR-E retrieval. This is most evident through the correlations between
the regional mean patterns of Wp and the Wc bias within the latitudes of 30°N and 30°S, suggesting that
the assumption of a regionally invariant cloud/precipitation partitioning in the AMSR-E algorithm is the
likely causal factor.