Optical thickness of stratiform clouds over the Baltic inferred from on-board irradiance measurements

Abstract

The paper discusses the optical thickness of non-precipitating layer clouds retrieved from irradiance measurements collected during cruises of R/V Oceania to the Baltic Sea from 1994 to 2002. Pyranometer measurements were accompanied by standard meteorological observations. Cloud optical thickness was obtained by a comparison of downward irradiance in the visible part of the spectrum averaged over 1-min intervals with MODTRAN4 (MOD4v2r1) computations. For an individual cloud retrieval, the total statistical error varies from 28% to over 100%. This is mainly attributed to irradiance measurement error, uncertainties in aerosol optical thickness, the lack of information on cloud droplet radius and the assumption of cloud horizontal uniformity. The systematic error (bias) due to the plane-parallel assumption is negative (the cloud optical thickness is overestimated) and is estimated at several percent. Statistical analysis of the optical thickness of layer clouds over the Baltic was performed with two purposes in mind: to look for (1) seasonal variations in cloud optical thickness and (2) for differences in cloud optical thickness for various cloud ‘classes’. The cloud ‘classes’ were distinguished with respect to the following: total cloud cover N and low-level cloud cover N L, and cloud type predominating in the sky (SHIP meteorological reports, WMO cloud classification and coding). Cloud optical thickness distributions for low-level layer clouds can be approximated by a lognormal distribution, the parameters of which depend on the cloud class. The mean base-10 logarithm of τ varies from about 1 ( τ=10) for semi-transparent clouds and cloud classes with N=7 to 1.5 for the class defined by N= N L=8 (overcast sky) and C L=7 (low-level clouds before or after precipitation). The values obtained are consistent with findings by other researchers. For the joint class containing all cases with N=8, the lowest mean logarithm of τ was found for February (about 1, which corresponds to τ=10) and the highest for the spring months, from March to May (1.3–1.4, which correspond to τ=20–25). However, the low number of data available does not allow for an accurate description of the annual cycle.