Towards improved nowcasting using LSA SAF MDSSFTD short wave radiation
Regional applications of satellite derived downward surface shortwave radiation flux in cloudy situations
- July 30, 2024
All operational LSA SAF products have gone through a process of optimizing and testing to ensure that they meet the needs of users. One such product is called MSG Total and Diffuse Downward Surface Shortwave Flux (MDSSFTD, LSA-207), which is very useful for meteorology, energy and agriculture.
Downward surface short wave flux (DSSF), which is a part of global product LSA SAF MDSSFTD, is greatly influenced by the thickness and position of clouds as well as aerosol and water vapour. Estimation of DSSF is more difficult for cloudy situations than for clear sky. More information on the performance in different sky conditions have been reported in the Validation Report.
The DSSF brings a good spatial consistency compared to localized in-situ data. This feature is appealing for improving performance of nowcasting and/or seamless prediction applications, which are primarily based on spatially sparse inputs (e.g., in-situ) and numerical weather prediction (NWP). For such applications, it is of significant importance to first recognize and possibly remove any discrepancies in input fields. The systematic differences of DSSF become significant over a regional domain as will be shown in the rest of this study. An approach will be presented that can be used to adapt DSSF with the goal of providing a representative input to a nowcasting application of the surface shortwave radiation flux over a regional domain. For this study, we are employing Slovenian in-house nowcast applications and our domain of interest fully encompasses Slovenia (see in Fig. 1).
.
Fig. 1. An example of LSA SAF MDSSFTD (downward surface short wave radiation) on 30 May 2023 12 UTC overlaid by in-situ stations (red full circles).
Methodology
The methodology is based on comparison of MDSSFTD with 30 in-situ observations (Fig. 1) over Slovenia in the period 2018-2023. In-situ data are available as both 10 and 30 minutes averages while MDSSFTD measurements are available every 15 minutes. In-situ data are first temporally reorganized to match the satellite measurement times. MDSSFTD’s quality flag is used to classify data based on sky conditions. Additionally, in-situ data are checked for any notable inconsistencies (read more here) compared with the MDSSFTD clear sky conditions.
Our pre-processing of DSSF is based on regional bias correction of DSSF defined in terms of look-up-table (LUT) which is prepared for cloudy situations. The LUT represents median differences between satellite and ground measurements as a function of several parameters. Results are showing that in cloudy situations (see LUT in Fig. 2) systematic differences are more pronounced for lower amplitudes of DSSF, with negative values in the morning and evening and positive values at the solar peak. These results are expected as they are in line with the Validation Report.
Performance
The 2018-2023 period is split into two data sets. The first one contains the first four years of data and is used to create the LUT. The second one covers only the last full year (2022) and is used to test the performance of the method. An example of the performance of the method for the high-altitude location of Rogla, Slovenia is presented in Fig 3.
.
Fig. 2. Median hourly differences between DSSF and in-situ measurements for cloudy conditions over the period 2018-2023. Blue values denote underestimation while orange values indicate overestimation. Differences are shown according to the hour-of-the-day (x-axis) and day of a year in unit months (y-axis). The size of the black circle indicates the sample size. Subplots are generated according to DSSF amplitudes.
.
Fig. 3. Presented are distributions of differences between DSSF and in-situ measurements at the location Rogla, Slovenia for August 2022. Colour of the boxes distinguishes between DSSF (cyan) and adapted DSSF (red).
Distributions of differences between satellite and in-situ measurements for both DSSF and adapted DSSF are presented by the hour-of-the-day for August 2022. The applied technique seems to have biggest effect on reducing differences at the midday and during summer months. Such response is well consistent with the average method performance. The method works well overall, i.e. reduces the largest differences, however, its performance can vary from station to station.
Conclusion
Due to increasing amount of renewable energy sources high quality prediction of solar radiation is necessary, particularly for the needs of the energy sector. Therefore, the first steps have been made towards improving our nowcasting performance using LSA SAF satellite products. The addition of LSA SAF MDSSFTD looks very promising, since it brings a good spatial consistency compared to localized in-situ data. Our analysis over Slovenia has shown that DSSF values are overestimated close to the solar peak while they are underestimated in the morning and evening hours. By doing this analysis, we have found an approach to utilize DSSF data as an input for nowcast applications.