Assessment of prediction skill of SEAS5 forecast using ERA5 soil moisture data and relation to crop production

Abstract. Soil moisture is a key component of the climate system and an important parameter of agricultural productivity, but its predictability varies strongly across regions and seasons. In this study, we assess the skill of the SEAS5 seasonal forecasting system in reconstructing soil moisture anomalies relative to ERA5 over 1981–2024. Later, we examine whether this prediction skill can be exploited to estimate seasonal winter wheat and maize yields. SEAS5 shows its strongest performance at short lead times (0–1 months), particularly in the upper soil layer, whereas the forecast skill of near-surface soil moisture decreases rapidly with increasing lead time. In contrast, deeper layers maintain substantially higher skill for several months, especially across central and northern Europe, reflecting the longer hydrological memory of deeper soil moisture, where precipitation and evapotranspiration signals are integrated over time. Projecting SEAS5 anomalies onto the leading ERA5 EOF patterns and reconstructing the first 10 principal components further enhances the agreement between SEAS5 and ERA5, indicating that SEAS5 captures the dominant large-scale, low-frequency modes of soil moisture variability. It was found that the first principal component of the deepest soil layer is contaminated by non-physical discontinuities associated the parallel production streams in ERA5 and the transition for hindcast to forecast in SEAS5. Reconstructing components 2–10 in both ERA5 and SEAS5 soil-moisture anomalies to remove this non-physical errors further improves the correlations. The SEAS5 prediction skill was found to be potentially relevant for agriculture. Winter wheat shows moderate correlation to soil moisture conditions during autumn establishment and spring regrowth, with pronounced relationships in the Balkans, Hungary, Romania, and central Europe. Maize exhibits an even stronger dependence on soil moisture throughout its growing season, especially in rain-fed regions where yield variability is primarily controlled by water availability. In the Balkan region, maize yields closely track soil moisture anomalies, demonstrating the potential for using SEAS5 as an early season predictor of crop outcomes. Overall, the principal component reconstruction of SEAS5 and ERA5 improves the correlation between the two datasets, demonstrating that SEAS5 prediction skill benefits from filtering out high-frequency noise. The refined signal provides meaningful soil-moisture predictability, which is particularly valuable for planning crops up to six–seven months ahead in rain-fed regions where yields are tightly linked to soil moisture variability. Integrating soil moisture forecasts with extended seasonal climate information can therefore strengthen drought preparedness and support climate-informed agricultural decision-making across Europe.