Assessment of direct normal irradiance assessment from cloud optical depth of geostationary meteorological satellites in Germany

The assessment of Direct Normal Irradiance (DNI) is essential for the solar power industry. Satellite estimations offer many advantages over ground-based measurements with immediate and affordable installation (no hardware installation), inexistant operation and maintenance of sensors, and efficient detection of any malfunction by comparing the production measured on the meter with the satellite estimation. In this work, we compare two estimations models of DNI from geostationary meteorological satellites in Germany.

The first model is provided by Copernicus Atmosphere Monitoring Service (CAMS) solar radiation services. In its latest v4 update, this model uses the APOLLO_NG (APOLLO_NextGeneration) cloud processing scheme that is a probabilistic interpretation of the original APOLLO (AVHRR Processing scheme Over cLouds, Land and Ocean) method to infer the Cloud Optical Depth (COD).

The second model is based on COD of the “Satellite Application Facility on support to NoWCasting and very short range forecasting” (SAFNWC) with the version 2018.1. The algorithm is based on a multispectral threshold technique applied to each pixel of the satellite image and uses Numerical Weather Prediction (NWP) input data for surface temperature and total atmospheric water vapour content.

Both models use a Cloud Mask (CMA) product to delineate all cloud-free pixels in a satellite scene with a high confidence, then DNI is forced to the value obtained by the McClear clear sky model when CMA indicates "cloud free". Otherwise, the COD is combined with the clear sky model in order to compute the effective DNI.

Moreover, we take into account the circumsolar radiation to avoid underestimation of the forward scattered radiation in DNI when compared to ground observations.

The model outputs are compared to 10-minute solar radiation measurements from Deutscher Wetterdienst (DWD) stations located in Germany over the period 2021-04-01 and 2022-03-31. This network measures the Global Horizontal Irradiance (GHI) and Diffuse Horizontal Irradiance (DHI). DNI is computed out of the quality-checked measurements of GHI and DHI. The results are expressed for all available measurement data in terms of relative Root Mean Scare Error (RMSE), RMSE Skill Score, Mean Absolute Error (MAE), MAE Skill Score, and mean bias error.

We conclude that the two models offer similar performances with a MAE between 25% and 35% for Germany.