Seasonal climate forecast can inform the European agricultural sector well in advance of harvesting

Agricultural production faces increasing challenges from unfavorable climate events and extremes like heat stress, drought, and excessive rainfall, impacting crop yields and global markets. Climate change is expected to worsen this trend. Seasonal climate predictions (lead time up to 1 year) offer valuable information for farmers, aiding in decisions regarding sowing, variety selection, fertilization, irrigation, and disease treatment. These forecasts also benefit policymakers in market, trade, and food security matters. While seasonal forecasts hold potential, challenges include lower prediction skill in regions like Europe and the dependence of skillful forecasts on spatial scale and timing. This study assesses the skill and reliability of the ECMWF's SEAS5 seasonal forecasting system in predicting agro-climate indicators for European winter wheat, using the co-designed Clisagri agro-climate service, which directly incorporates farmer and agronomist input. The spatial assessment evaluates forecasts at different winter wheat growth stages, exploring predictability of various climate events and identifying opportunities linked to increased lead times of skillful predictions.

Existing research demonstrates the added value of seasonal climate forecasts for agricultural decision-making in various regions. However, challenges remain in meeting end-user expectations due to inconsistent prediction skill across key areas, including Europe. The dependency between forecast skill and the spatial scale of climate events presents another hurdle. Interpreting probabilistic forecast uncertainty for effective integration into decision-making processes is also a significant challenge. Previous studies have often focused on assessing skill for essential climate variables, while this study adopts an integrated perspective, considering indicators relevant to climate risks throughout the growing season.

This study utilizes the European Centre for Medium-Range Weather Forecasts (ECMWF) seasonal forecasting system SEAS5 to predict agro-climate indicators relevant to European winter wheat production. The analysis uses retrospective forecasts (re-forecasts) from 1993-2019, consisting of 25 ensemble members. Three key climate variables—minimum and maximum daily temperatures, and daily total precipitation—are used. Forecasts were retrieved for six initialization times (November, February, March, April, May, and June), each with a 7-month lead time. The data is bias-adjusted using quantile mapping with MarsMet as the reference dataset. The Clisagri agro-climate service provides a set of dynamic agro-climate indicators tailored to winter wheat, characterizing climate events such as drought (using SPEI), excessive wetness, heat stress, and cold stress. The timing of these indicators is dynamically linked to crop phenological stages, considering sowing dates, varieties, and climate conditions. The Fair Ranked Probability Skill Score (FRPSS) assesses prediction skill against climatology, while reliability diagrams evaluate the correspondence between predicted probabilities and observed frequencies. Five reliability categories are defined: perfect, very useful, marginally useful, not useful, and dangerously useless.

The study reveals skill in predicting wheat flowering in central and eastern Europe as early as November forecasts. This skill increases with later initialization times. The reliability of flowering predictions is useful for decision-making for late and early flowering categories after February, although predicting the 'normal' category remains difficult. The predictability of agro-climate indicators is strongly linked to forecast initialization time and the indicator type. Hydrological balance indicators (based on SPEI) show improved predictability with later initialization, particularly for drought prediction during the entire growing season. However, excessive wetness indicators show limited skill throughout the season. Temperature stress indicators exhibit better skill, particularly in southern Europe. Predicting drought between heading and maturity is crucial, and while skill is limited early in the year, it improves with later forecast runs, especially in southern and eastern Europe. The 2018 drought event in central and northern Europe highlights the limited early-season prediction skill, with the June forecast showing better accuracy. The study also finds a positive correlation between soil moisture conditions in late spring and summer drought events in eastern Europe.

The findings confirm the usefulness of seasonal climate forecasts for decision-making in the European wheat sector. The ability to predict flowering time early in the season supports critical agricultural choices. The dependence of forecast skill on initialization time and indicator type highlights the need for further research to enhance the predictability of events like excessive wetness. Alternative approaches, such as using large-scale atmospheric patterns, should be explored to improve forecasts. The limited predictability in certain regions (e.g., western Europe during spring) suggests the influence of factors such as extra-tropical jets and the difficulty in predicting atmospheric regime transitions. While this study provides a European-level assessment, the impact of climate events can vary based on soil properties. The study's indicators prioritize climate risks relevant to winter wheat production, but regional prioritization should consider specific environmental conditions.

This study demonstrates the valuable potential of seasonal climate forecasts for informing decisions within the European wheat sector, offering benefits to farmers and policymakers alike. The findings highlight the need for further research to improve the skill of extreme weather event predictions, particularly excessive wetness, by exploring alternative forecasting methods and higher-resolution data. Co-design approaches with end-users are essential for effectively translating climate predictability into actionable information.

The study's findings are based on a specific seasonal forecasting system (SEAS5) and a particular set of agro-climate indicators. The results may not be directly generalizable to other regions, crops, or forecasting systems. The accuracy of the predictions is affected by the inherent limitations of seasonal climate forecasts and their ability to capture the complexity of regional climate variability. The study focused on winter wheat, and the findings may not fully apply to other crops with differing phenological stages or sensitivity to climate extremes.