Temperature and precipitation are the most important factors responsible for agricultural productivity variations. In 2018 spring/summer growing season, Europe experienced concurrent anomalies of both. Drought conditions in central and northern Europe caused yield reductions up to 50% for the main crops, yet wet conditions in southern Europe saw yield gains up to 34%, both with respect to the previous 5‐year mean. Based on the analysis of documentary and natural proxy‐based seasonal paleoclimate reconstructions for the past half millennium, we show that the 2018 combination of climatic anomalies in Europe was unique. The water seesaw, a marked dipole of negative water anomalies in central Europe and positive ones in southern Europe, distinguished 2018 from the five previous similar droughts since 1976. Model simulations reproduce the 2018 European water seesaw in only 4 years out of 875 years in historical runs and projections. Future projections under the RCP8.5 scenario show that 2018‐like temperature and rainfall conditions, favorable to crop growth, will occur less frequent in southern Europe. In contrast, in central Europe high‐end emission scenario climate projections show that droughts as intense as 2018 could become a common occurrence as early as 2043. While integrated European and global agricultural markets limited agro‐economic shocks caused by 2018's extremes, there is an urgent need for adaptation strategies for European agriculture to consider futures without the benefits of any water seesaw.

In the present study an exploratory analysis of the relationship between regional fAPAR time-series and yield inter-annual variability in the period 1999-2012 has been conducted, evaluating its reliability for operational crop yield forecasting activities. Overall, the results indicated strong correlation between fAPAR and official yields in water-limited yield agro-climatic conditions for all the three crops studied. Where crops are subjected to frequent water stress, e.g. the Mediterranean basin, the Black Sea area, most of the yield inter-annual variability is explained by leaf area formation along the growing season, and can be well captured by regional fAPAR time-series. By contrast, in the regions of high yields, where generally fAPAR inter-annual variability is low as consequence of rainfall regimes close to potential evapotranspiration (or intensive irrigation in the specific case of grain maize) the correspondence between fAPAR and yield anomalies is weak. Yield variations are explained by other factors than leaf area formation (e.g. excess of rain during harvest), and therefore extreme yields in those regions are not necessarily corresponding to large fAPAR anomalies. The use of remote sensing in operational crop yield forecasting at regional scale therefore requires the consideration of additional, independent indicators: weather parameters, crop models, etc. EO biophysical products provide, indeed, a synoptic view of the vegetation status and leaf area formation along the growing season. Nevertheless, the results of this study indicates that their interpretation may be sometimes misleading when used alone to explain extreme yields.JRC.H.4-Monitoring Agricultural Resource

Early-stage plant density is an essential trait that determines the fate of a genotype under given environmental conditions and management practices. The use of RGB images taken from UAVs may replace the traditional visual counting in fields with improved throughput, accuracy, and access to plant localization. However, high-resolution images are required to detect the small plants present at the early stages. This study explores the impact of image ground sampling distance (GSD) on the performances of maize plant detection at three-to-five leaves stage using Faster-RCNN object detection algorithm. Data collected at high resolution () over six contrasted sites were used for model training. Two additional sites with images acquired both at high and low () resolutions were used to evaluate the model performances. Results show that Faster-RCNN achieved very good plant detection and counting () performances when native high-resolution images are used both for training and validation. Similarly, good performances were observed () when the model is trained over synthetic low-resolution images obtained by downsampling the native training high-resolution images and applied to the synthetic low-resolution validation images. Conversely, poor performances are obtained when the model is trained on a given spatial resolution and applied to another spatial resolution. Training on a mix of high- and low-resolution images allows to get very good performances on the native high-resolution () and synthetic low-resolution (