Quantitative assessment of one year of cloud phase determination from SEVIRI using ground-based cloud radar and lidar data

Accurate determination of cloud phase is of key importance for the retrieval of cloud physical properties from satellite. The past few decades, several methods have been developed. The type of spectral radiances used varies from solely visible, near-infrared and thermal infrared to combinations of these spectral radiances. The validation of the different methods has mostly been restricted to case studies. Little is known on the accuracy of the methods when applied to long term data sets.

The research presented here explores which retrieval method is most appropriate to derive cloud phase for climate monitoring purposes. The research is done in the framework of the Satellite Application Facility on Climate Monitoring (CM-SAF), which aims at providing the climate research community with high quality data sets. The five cloud phase retrieval methods under study are: the KNMI 0.6/1.6 μm reflectance method and thresholds of the brightness temperature 10.8 μm, brightness temperature 12.0 μm, brightness temperature difference 10.8-12.0 μm, and cloud top temperature. A full year of the Spinning Enhanced Visible and Infrared Radiometer Instrument (SEVIRI) data, comprising more than 4300 cloud phase retrievals in total, are compared with collocated cloud phase observations from ground-based cloud radar and lidar at the CloudNET station of Cabauw, The Netherlands. The cloud top temperature is calculated from brightness temperature at 10.8 μm using an emissivity correction obtained from the optical thickness at 0.6 μm. In order to examine the impact of a change in threshold value on the methods' accuracy, the temperature threshold methods (brightness temperature 10.8 and 12.0 μm, and cloud top temperature) are executed at values between 250 and 280 K. The value of 260 K is used in the International Satellite Cloud Climatology Project (ISCCP). Cloud phase retrievals using brightness temperature difference 10.8-12.0 μm are performed using various thresholds. The KNMI 0.6/1.6 μm reflectance method is based on a method by King et al. (1992). Cloud optical thickness, τ, and effective radius, r_eff, are retrieved by comparing measured 0.6 and 1.6 μm reflectances to pre-calculated Lookup Table values for spherical water droplets and imperfect hexagonal ice crystals performed by the DAK (Doubling Adding KNMI) radiative transfer method. Surface reflectance information at 0.6 and 1.6 μm is obtained using the MODIS surface albedo product.

The accuracy of the methods based on SEVIRI data is assessed by evaluation of the seasonally averaged water phase ratio (ratio of water phase cases over all cloudy cases), which are calculated from 2 pixels closest to the Cabauw geolocation. Ground-based observations are taken from a 10-minute time window centered at the SEVIRI time slots. Further, correlation coefficients of daily averaged water phase ratio between SEVIRI and ground-based observations are presented as a second measure of accuracy.

The seasonally averaged water phase ratios obtained from SEVIRI, using the cloud top temperature method with threshold at 265 K and ground-based observations are presented in Figure 1. It can be seen that in the spring and winter months bias (SEVIRI-surface) is fairly low, ~+5%, however it is higher during the summer months, indicating that a higher threshold value is to be used in these months for mid-latitude climate regions. The correlation coefficient at 265 K is high in summer 2004, but drops to less significant values in the winter months. At 260 K, bias has larger positive deviations from the surface reference water phase ratios in summer 2004 than at 265 K. The correlation coefficient for the 0.6/1.6 μm method is significant over the entire period. Values decrease from 0.82 in summer 2004 to 0.67 in winter, followed by an increase to 0.72 in spring 2005. The bias for this method has a constant, but rather high, positive value through the entire investigated period.

Figure 1: Seasonally averaged (90 days running average) water phase ratios for the cloud top temperature method with threshold 265 K (solid line), compared to water phase ratios obtained from ground-based cloud radar and lidar (dashed line).