Assessing the uncertainty of expert observations of cloud characteristics based on data from a field campaign in the Arctic ocean in August-September 2021

The sun is the closest natural source of radiation, both shortwave and longwave. The state of the atmosphere, and in particular the Total Cloud Cover (TCC) and the Lower Cloud Cover (LCC), most strongly affects the transfer of incoming solar radiation to the surface. At the moment, the amount and types of clouds are assessed primarily by an expert using visual observation, and such an assessment is considered reliable according to WMO observations guide. However, it is known that the estimates of an observer are subject to errors due to the subjectivity of perception of the visual cloudy scene. Uncertainty in observer estimates may lead to significant inaccuracies in operational weather forecast systems as well as in reanalyses and climatic time series. In addition, the lack of knowledge about the observation error limits one in assessing the corresponding uncertainty of the climatic trends of cloudiness characteristics. In this study, we investigated the uncertainty in the estimates of the TCC, LCC.

To carry out such a study, we conducted an experiment involving the simultaneous observation of the same cloudy situation by several observers. The experiment was carried out on board the Akademik Ioffe research vessel during the AI-58 research cruise from August 18 till September 6 of 2021 in Kara, Baltic and White Seas. The experiment involved 19 volulntary participants. There were 78 observation moments. The number of observers varied from 5 to 19 due to their own duties onboard. On average, the cloud characteristics were assessed by 12 participants.

Thus, in the present study, the uncertainties of cloud characteristics estimated by one forgetful independent observer several times in equivalent conditions were simulated with a large number of experts participating in synchronous observations. We demonstrate that the disparity of opinions is small for simple cloudy situations in which the sky is almost clear or mostly covered by clouds. We also show that the uncertainty in the conditions of moderate cloudiness can reach 1.5 oktas in terms of standard deviation.

This study may help clarifying existing and future models for assessing meteorological characteristics, as well as models used to calculate incoming solar radiation. We plan to assess the uncertainty of cloud types observed by human experts. We will also repeat our experiment in other regions of the World Ocean in order to expand the variety of observed cloud situations, in which a wider range of expert opinions can be expected, as well as to form a dataset balanced w.r.t. synoptic conditions.