Land-surface initialisation improves seasonal climate prediction skill for maize yield forecast

Abstract

Seasonal crop yield forecasting represents an important source of information to maintain market stability, minimise socio-economic impacts of crop losses and guarantee humanitarian food assistance, while it fosters the use of climate information favouring adaptation strategies. As climate variability and extremes have significant influence on agricultural production, the early prediction of severe weather events and unfavourable conditions can contribute to the mitigation of adverse effects. Seasonal climate forecasts provide additional value for agricultural applications in several regions of the world. However, they currently play a very limited role in supporting agricultural decisions in Europe, mainly due to the poor skill of relevant surface variables. Here we show how a combined stress index (CSI), considering both drought and heat stress in summer, can predict maize yield in Europe and how land-surface initialised seasonal climate forecasts can be used to predict it. The CSI explains on average nearly 53% of the inter-annual maize yield variability under observed climate conditions and shows how concurrent heat stress and drought events have influenced recent yield anomalies. Seasonal climate forecast initialised with realistic land-surface achieves better (and marginally useful) skill in predicting the CSI than with climatological land-surface initialisation in south-eastern Europe, part of central Europe, France and Italy.

When Will Current Climate Extremes Affecting Maize Production Become the Norm?

“…Globally, the sum of the country‐level production anomalies estimated by the CSI captures 49% of the variance, in agreement with previous studies (Ceglar et al, ; Ray et al, ).…”

“…In a statistical framework, the main effects of climate anomalies can be diagnosed through heat and water stress indicators, here the Heat Magnitude Day (HMD; Zampieri et al, , ; Zampieri, Ceglar, Dentener & Toreti, ) and the Standardized Precipitation Evapotranspiration Index (SPEI; Vicente‐Serrano et al, ). These two indicators are computed for a specific period before harvesting, encompassing the flowering stage and the grain filling period (Ceglar et al, ; Zampieri, Cegla, Dentener & Toreti, ). The HMD integrates the amplitude and duration of maximum daily temperature anomalies exceeding the 90th percentile, which is seasonally and spatially varying.…”

“…Like previous statistical studies for maize (Ceglar et al, ), we use a two‐month time window for computing the HMD and a four‐month window for the SPEI, both counting backward from the second month before harvesting. Thus, the period after ripening, when the crop is left in the field to dry up and is not any more sensitive to climate variations, is excluded.…”

“…When focusing on the interannual yield variation, assuming constant harvested area and removing the baseline trend of yields and climate indexes, the CSI explains 50% of maize yield interannual variability in Europe (Ceglar et al, ).…”

“…In equation , β and γ are regression coefficients. The calibrated linear combination of HMD and SPEI is the CSI (Ceglar et al, ; Zampieri, Ceglar, Dentener & Toreti, ), which is proportional to the production loss P CSI through the α coefficient.…”