A fine-scale-preserving bias correction approach for global solar irradiance simulated with a numerical weather prediction model over a high-resolution spatial and temporal grid

<p>The feasibility of national and supra-national low carbon power systems (LCPSs) is challenged under high penetration rates of renewables because the inherent variability of renewable sources increases the system’s operation and transmission costs. Optimizing the operation of the systems to make them efficient and profitable requires adapting their design to the regional solar and wind patterns.</p><p>The MET4LOWCAR Spanish project aims at demonstrating the benefits of a synergistic design of the generation and transmission power system that accounts for the regional climatic patterns of both solar and wind renewable resources, using the Spanish territory as a testbed. To that aim, a 30-yr weather integration is being performed with the Weather Research and Forecasting (WRF) model over the Iberian Peninsula and the Balearic Islands throughout a 5-km spatial grid and a 10-min temporal grid. The ultimate goal is to compound a climatic data base of solar- and wind-related variables to simulate the performance of a high-penetration LCPS in multiple scenarios, from long runs to extreme weather events.</p><p>Here, we present an original debiasing approach of the WRF global horizontal solar irradiance (GHI) that guarantees a reliable and more realistic representation of the solar power generation in the LCPS. First, the GHI simulated by WRF is compared against radiometric observations from the Spanish National Radiometric Network to demonstrate that it is indeed affected by a seasonal bias, related to a misrepresentation of convective clouds in the WRF model. Then, the spatial and temporal GHI grid is corrected cell by cell and day by day using the monthly GHI SARAH satellite product as a reference. The method is purposedly designed to interfere only at monthly scale (thus using monthly GHI SARAH as a reference) in order to preserve the original fine-scale spatial and temporal structure of GHI. However, it propagates the debiasing to daily steps using a tailored interpolation approach that prevents any spikes and data artifacts in the bounds between consecutive months.</p><p>The method will be described in detail, and some preliminary results will be shown at least for the period 2010–2020, using ground observations and satellite-based GHI data as a reference to assess the debiasing performance.</p>