College of Engineering | Apply to CSU | Disclaimer | Equal Opportunity Statement | Privacy | Search CSU

A Comparison of Observed and Modeled Inter-Annual Variability and Trends in Cloud Liquid Water Pah

You must be on the CSU network—either physically or using VPN—to watch this or any of the videos on this site.

December 10, 2015
Andrew Manaster
Hosted by Chris Kummerow (advisor), Chris O’Dell (co-advisor), Dave Randall, Steven Reising (Electrical and Computer Engineering)


Long-term satellite records of changes in cloud properties have only been available for the past several decades and have just recently been used to help diagnose cloud-climate feedbacks. However, due to issues with satellite drift, calibration, and other artifacts, the validity of these changes in cloud properties has been called into question. It is therefore pertinent that we look for other observational datasets that can help to diagnose changes in variables relevant to cloud-radiation feedbacks. One such dataset is the Multi-sensor Advanced Climatology of Liquid Water Path (MAC-LWP), which blends cloud liquid water path (LWP) observations from 12 different passive microwave sensors over the past 27 years. In this study, observed LWP trends from the MAC-LWP dataset are compared to LWP trends from 16 models in the Coupled Model Intercomparison Project 5 (CMIP5) in order to assess how well the models capture these trends and thus related radiative forcing variables.

Mean state values of observed LWP are compared to those of previously observed climatologies and are found to have relatively good quantitative and qualitative agreements. Observed mean state LWP is compared both qualitatively and quantitatively to our suite of CMIP5 models. These models tend to capture mean state LWP features, but the magnitudes exhibit large variations from model to model. Several metrics are used to compare observed mean state LWP and the mean state LWP in models. However, the models' performance in regards to these metrics is found to not be indicative of their abilities to accurately reproduce trends on a regional or global scale.

Global trends in the observations and the model means are compared. It is found that observational trends are roughly 2-3 times larger in magnitude in most regions globally when compared to the model mean, although this is thought to be at least partly caused by cancellation effects due to differing inter-annual variability and physics between models. Several regions have consistent signs in trends between the observations and the model mean while others do not. This is due to spatial inconsistencies in certain trend features in the model mean relative to the observations.

Trends are examined in individual regions. In four of the six regions analyzed, the observational trends are statistically different from zero, while, in most regions, very few models have trends that are statistically significant. In certain regions, the majority of modeled trends are statistically consistent with the observed trends, although this is typically due to large estimated errors in the observations and/or models, most likely caused by large inter-annual variability. The Southern Ocean and globally averaged trends show the strongest similarities to the observed trends. Almost all Southern Ocean trends are robustly positive and statistically significant, with the majority of models being statistically consistent with the observations. Similarly, the observed and global trends are all positive, with the majority being statistically significant and statistically consistent. We discuss why a large positive Southern Ocean trend is unlikely to be due to a trend in cloud phase.

CMIP5 model mean and observational LWP trends are compared regionally to Atmospheric Model Intercomparison Project (AMIP) and European Center for Medium-Range Weather Forecasting (ECMWF) ERA reanalysis trends. It is found that AMIP model mean and ERA LWPs are better than the CMIP5 model mean at capturing the inter-annual variability in the observed time series in most of the regions examined. The AMIP model mean better replicates the observed trends when the inter-annual variability is better captured. The ERA reanalysis tends to better reproduce the observed inter-annual variability when compared to the AMIP model mean in almost every region, but, surprisingly, it is either worse or roughly the same in regards to matching observed trends.

Our results suggest that observed trends are due to a combination of inter-annual and decadal-scale internal variability, in addition to externally forced trends due to anthropogenic influences on the climate system. With a record spanning three decades, many modeled trends are statistically consistent with the observed trends, but a true climatically forced signal is not yet apparent in the models that agrees with the observations. The primary exception to this is in the Southern Ocean, where virtually all models and observations indicate an increasing amount of cloud liquid water path.