The Cloud Retrieval Evaluation Workshops (CREWs) provide a forum for international cloud remote sensing scientists to share their experience with state-of-the-art cloud parameter retrievals from satellite, airborne, and surface observations. The overarching objectives of CREW are to enhance our knowledge on quantitative cloud parameter retrievals from state-of-the-art algorithms and identify shortcomings that need focused attention as a community. Continual improvement in the global description of cloud properties optimizes these algorithms for near-term (nowcasting), short- to medium-term (weather forecasting), and long-term (regional and climatological analyses) applications, as well as for potential improvements in the cloud and convection parameterizations adopted in weather and climate models.
What: A joint Asian–European–U.S. workshop gathered about 70 research scientists to review existing and new approaches to infer cloud parameters from passive and active satellite observations. The priorities of this workshop were to compare and improve level-2 and level-3 cloud products from different teams, increase commonality between these products, and define scientific focal points and collaborations for the next workshop in 2016.
When: 4–7 March 2014
Where: Grainau, Germany
The Fourth CREW (CREW-4) was the held in Germany in March 2014. The workshop was attended by about 70 participants from various universities, research institutes, and satellite agencies in Asia, Europe, and the United States. The workshop was organized along four thematic sessions: “Cloud parameter retrieval methods,” “Cloud parameter retrieval evaluations,” “Nowcasting and severe weather applications,” and “Cloud parameter datasets for weather and climate research.” The cloud parameter retrieval evaluations were facilitated by a common database that comprises, for a number of “golden days,” cloud property retrievals from different algorithms for passive imagers (SEVIRI, MODIS, AVHRR, POLDER, and AIRS; see Table 1 for a complete list of acronyms) and other cloud measurements that serve as a reference (e.g., CALIOP, CPR, and AMSU-E).
A noticeable finding was the increased number of research groups that now implement optimal estimation methods in their operational retrievals. In addition, some research groups have started to combine observations from both passive and active instruments. While the active sensors provide information for only a very small portion of the imager swath, these observations are critical for improving global cloud parameter retrievals.
The preliminary results presented on the assessments of error estimates produced by some of the retrieval schemes were an important step toward quantifying these estimates in a more systematic manner. These assessments reveal that error estimates compare reasonably well in multiple algorithm ensembles or against the true uncertainty between retrieved and observed cloud parameters.
The evaluation of aggregation methods and filtering rules reveal that the manner of aggregating or filtering level-2 data creates systematic differences in level-3 products that tend to vary regionally depending on climate regions and/or surface conditions. Although the differences are smaller than those between level-2 retrievals, they are not negligible. For the community to be able to truly compare their gridded level-3 products, a set of common filtering and aggregation rules need to be identified and integrated by the various teams.
OUTCOME OF CREW-4.
Cloud parameter retrieval methods continue to advance by improving physical approaches, adopting new retrieval methods, or exploiting new types of observations. The session “Cloud parameter retrieval methods” reported on updates made to existing operational cloud parameter retrievals (e.g., MODIS06–006, CM-SAF), as well as on newly developed retrieval methods (e.g., ESA Cloud_cci), such as optimal estimation methods and methods that combine observations from passive and active instruments.
The presentations in the session “Cloud parameter retrieval evaluations” covered the latest results of intercomparison and validation assessments of cloud parameters. Since CREW-3 in 2012, more groups now perform these types of assessments (Hamann et al. 2014; Stengel et al. 2015). Figure 1 shows an example of the CREW type of assessment made for cloud optical depth retrievals from polar-orbiting satellite observations from the VIIRS instrument. Moreover, the validations are performed for different types of cloud conditions, giving better insight to the strengths and weaknesses of the algorithms. Initial results were presented of an assessment of the error estimates produced by some of the retrieval schemes.
The special session “Nowcasting and severe weather applications” highlighted how cloud parameter retrievals can be used to predict severe storms by making use of the local dynamics and the temporal resolution of the geostationary data.
The session “Cloud parameter datasets for weather and climate research” reported on conditions and requirements that need to be satisfied for the generation of well-characterized cloud parameter data records and on the use of these so-called thematic climate data records (TCDRs) in several climate monitoring and climate model evaluation studies.
More focused discussions on cloud retrieval principles and the validation of cloud parameters were held within three clusters of parallel breakout sessions. The topics of these sessions were i) retrieval methods, ii) retrieval and uncertainty evaluations, and iii) nowcasting and climate applications.
CREW endorsed the need for accurate calibration of all passive imager measurements. We hope that GSICS will provide such calibration (visible, near infrared, and infrared) for all passive imagers. The community decided to form subordinate working groups to improve the retrievals, for example, of multilayered clouds, for the treatment of vertical cloud inhomogeneity, and in polar regions. The participants also discussed forming multi-algorithm ensembles to assess uncertainties and sensitivities of, for example, error estimates. There was consensus on the need to streamline intercomparison and validation activities across research groups. Among other recommendations, CREW suggests that the measurement community work toward establishing a network of climate anchor reference sites that operate several reference networks simultaneously (e.g., GRUAN, BSRN, AERONET). The research into using cloud parameter products in severe weather applications generated much discussion and was encouraged through case studies focusing on convection, fog, and severe fires. The breakout session on climate applications discussed ways to accommodate a common approach for generating global decadal gridded (level 3) cloud parameter records and in particular recommended focusing on the definition of a set of essential filtering rules for different cloud parameters and further to investigate the propagation of error estimates in level-3 products. In support of the GEWEX Cloud Assessment, CREW will seek to advance the production of long-term datasets that are well characterized in their strengths and weaknesses. The need was acknowledged for preservation of these data in formats that are widely accessible and to adopt self-describing common data formats.
The participants strongly supported the proposal to establish CREW as an International Clouds Working Group (ICWG) within the Coordination Group for Meteorological Satellites (CGMS). At the 42nd plenary session of CGMS, which was held in China in May 2014, the CGMS-42 plenary endorsed the formation of an ICWG, making CREW a formal entity. The community is now comfortable enough with the CREW forum to collaborate well beyond what has occurred to date and to promote software exchange. It decided to adopt the format of subordinate working groups that could make efficient progress on a variety of topics and report on their findings at the next CREW meeting. In the final discussions, the priority research themes for the coming years were defined. In connection to these themes, a list of potential topics for subordinate working groups and leads was drafted. The next CREW will be held in Lille, France, in 2016.
Improve cloud models used in retrievals to more accurately reflect reality, particularly ice crystal models, vertical inhomogeneity, and multiple layers.
Explore the potential of combining different types of observations in level-2 cloud retrievals methods.
Explore the definition of a set of essential filtering rules in level-3 aggregation methods for different cloud parameters.
Work toward the characterization of uncertainties in level-2 and level-3 products.
Explore production of multi-algorithm ensembles to assess uncertainty/sensitivity.
Explore the production of long-term datasets aimed at stability and accurate assessment of product strengths and weaknesses.
Use common ancillary data and validation procedures for level-2 and level-3 data.
Establish subordinate working groups to make progress on a variety of outstanding issues: for example, multilayered clouds, severe weather applications, and aggregation methods.
More detailed information on the CREW workshops can be found on the CREW website (www.icare.univ-lille1.fr/crew/). The passive imager retrievals and the reference data in the common database are available via our FTP site and can be downloaded (www.icare.univ-lille1.fr/register/register.php). When asked for a “short description of your project,” please state that you want to have an account created for the Cloud Retrieval Evaluation Workshop.
The intercomparison and evaluation of retrievals schemes, done as preparatory work to this workshop, was part of the EUMESTAT Fellowship project cosponsored by EUMETSAT. Financial and organizational contributions for holding this workshop were made by EUMETSAT, ESA, DWD, and CM SAF.