Cereal production in Europe

Cereals are grown on half of the European Union's (EU) farms, occupying a third of EUs agricultural area and accounting for a quarter of its crop production value. On a global scale, Europe accounts for 20% of the total cereal production, of which about 63% is produced in the countries of the EU28. Cereals in the EU28 are mainly used for animal feed (61%) and human consumption (24%), while smaller other purposes include alcoholic beverages (5%), bio-energy (4%) and seeds (3%) (EU, 2016). Wheat and rye are used almost equally for animal feed and human consumption, while barley, maize, sorghum, oats and triticale are predominantly used for animal feed. In recent years, Europe has been a net exporter of around 15% of its cereal production (FAO Food Outlook, 2016). The exported cereals are mainly wheat and barley, while the imports consist mainly of maize.

Approach

The global yield gap atlas presents actual and potential yields for wheat, barley and maize in Europe. Together these three crops represent 90% of the European cereal production. We use a country-by-country bottom-up approach to establish statistical estimates of actual grain yield, and compare these to modelled estimates of potential yields. In brief, the approach distinguishes the following main steps: (1) selection of representative climate zones (CZ) based on dominant crop areas, (2) selection of reference weather stations (RWS) that represent the selected CZs, (3) selection of dominant soil types and cropping systems in a 100 km radius around the RWS, (4) crop model simulations to establish rainfed or irrigated yield potential, and (5) estimation of actual yields from statistical surveys. Detailed information and justification is available in separate publications on climate zones, upscaling from RWS to CZ and national scale with area-weighted averages, and data selection methods. For 22 out of the 39 countries, country agronomists were involved, representing 94% of the wheat and barley area, and 82% of the maize area. They had a pivotal role in data access, expert estimates and evaluation of results. The next sections present a point-by-point account of the underlying data and methods used for Europe.

Selected climate zones

On average, there are 15 CZs per country, but it varied from 2 (the Netherlands) to 48 (Spain). Harvested areas for wheat, barley and grain maize were taken from the Spatial Production Allocation Model (SPAM) of Harvestchoice, presenting harvested areas around the year 2005 at a 5' grid, which is the most recent version to date. Both layers were combined to calculate the harvested area per climate zone per country. The harvested areas of wheat (34 Mha), barley (20 Mha) and maize (13 Mha) were distributed over 80 CZs for wheat and barley, and 75 CZs for maize (Table).

Selection of the CZs was carried out in two steps, with the aim to cover at least 50% of the harvested area. First, all CZs with at least 5% of the national harvested area were selected. However, for some countries with many relatively small climate zones this approach resulted in a low crop area covered by the CZs. Therefore, also CZs with less than 5% of the national harvested area were selected if a suitable weather station was present for that particular CZ. The final number of selected CZs was on average 5 per country, with a range of 1 (Luxemburg) to 16 (Spain). On average, 43% of the number of CZs in a country were selected representing on average 88% of the national crop area, but it ranged from 26% (Bosnia Herzegovina – maize) to 100% (the Netherlands) of the national crop area.

Reference weather stations (RWS)

Weather data were collected for stations in the selected CZs with at least ten years of consecutive daily data and less than 20% missing data for each variable. Precipitation, minimum and maximum temperature, vapour pressure and wind speed were used from stations in the NOAA Global Surface Summary of the Day (https://data.nodc.noaa.gov/cgi-bin/iso?id=gov.noaa.ncdc:C00516). If suitable weather stations with adequate data were unavailable, gridded weather data from Agri4Cast (http://agri4cast.jrc.ec.europa.eu/) were used. Daily radiation was derived from NASA (http://power.larc.nasa.gov/). For precipitation, zero was assigned to missing values, while for other variables linear interpolation was used to fill data gaps.

The total number of selected RWSs for wheat, barley and maize was 287, 271 and 202, respectively. This equates to an average of 1.3 weather stations per climate zone. In 22 cases (Albania, Bosnia Herzegovina, Kosovo, Macedonia, Moldova and Montenegro) a suitable RWS with actual weather data was unavailable, and gridded data were used that represented the location of the specific weather station.

Soils and crops

Data on water content at field capacity (pF=2.5) minus water content at wilting point (pF=4.2), root penetrable soil depth and hydrogeological class of soils were taken from the 1 km x 1 km grid of the European Soil Database. Within the 100 km zone around each RWS, the three dominant Soil Map Units (SMUs) were selected, based on the harvested area. Each SMU comprises a varying number of Soil Type Units (STUs) of which the soil parameters were used as input to the model.

The yield gap assessment in this study covers rainfed wheat, barley and maize, as well as irrigated maize. For wheat and barley, both spring sown and autumn sown cropping systems were taken into account if they covered at least 15% of the total wheat or barley area (Table). Common wheat is the default wheat type, but for Mediterranean regions durum wheat was also considered. In this study the term "maize" stands for grain maize; forage maize was not included.

Crop modelling

We implemented the WOFOST crop model version 7.1.7, which was calibrated for many regions throughout Europe. WOFOST computes daily biomass accumulation and its distribution over crop organs during the growth period using a photosynthesis minus respiration approach. Crop yield was simulated for the potential (Yp) and the water-limited (Yw) production situation. Yp is determined by temperature, day length, solar radiation and genetic characteristics assuming absence of any water or other stress factors. Yw is limited by water supply, and hence influenced by rainfall, soil type and depth. Soil water dynamics in the root zone were simulated with a daily time step. The model does not account for capillary rise of water from below the root zone. To prevent overestimation of water limitation for soils with shallow groundwater (e.g. large parts of the Netherlands and the Po valley in Italy), Yw was assumed to equal Yp in these regions. For both Yp and Yw non-limiting nutrient supply was assumed. Yield losses caused by pests, diseases, weed and/or extreme weather events were considered for neither Yp nor Yw. Vernalization was not implemented in WOFOST and crop growth and phenological development for autumn sown crops were therefore calculated from January 1^st onwards, assuming a fixed initial dry biomass of 210 kg ha^-1.

The original crop parameters for European regions were updated in two steps to represent conditions covering approximately the last 15 to 20 years. First, additional phenology data were collected from a variety of sources, comprising either geo-referenced networks of observations, experimental sites or expert estimates by country agronomists. For spring sown crops, the observations consisted of day of sowing, emergence, anthesis and maturity. For autumn sown crops, emergence was fixed on January 1^st. All phenology data were grouped per country per climate zone, and subsequently temperature sums between sowing and emergence, between emergence and anthesis, and between anthesis and maturity were calculated for each RWS. Initial model runs were carried out with these data and the simulated harvest index (HI) and maximum leaf area index (LAIM) were checked against a plausible range.. If more than 10% of the number of simulations were outside the plausible range for HI (minimum: 0.35 to 0.40; maximum 0.55 to 0.60, depending on crop) or LAIM (minimum 3 to 4; maximum 6 to 7, depending on crop), crop parameters were adjusted within biologically plausible values as per the guidelines for regional calibration of WOFOST. The relevant parameters for calibration were, in order of importance, SLATB (specific leaf area [ha kg^-1]), the partitioning of assimilates to the various organs (FSTB for stems, FLTB for leaves and FOTB for storage), SPAN (life span of leaves growing at 35^oC [d]) and AMAX (maximum leaf CO₂ assimilation [kg ha^-1 h^-1]). The final evaluation step consisted of an expert assessment by the country agronomist, and if available, supported by annual maximum yields observed in variety trials or agronomic experiments. If the average simulated yields were more than 15% lower or higher than the observed maximum yields, or if the experts qualified the simulations not plausible, further fine-tuning of the crop parameters, as described above, was carried out.

The simulations were carried out for each STU within a RWS zone. The simulation results were up-scaled successively to SMU, RWS, CZ and country, using harvested area per STU as weighing factor. All simulated crop yields are presented at standard moisture content, i.e. 13.5% for wheat and barley, and 15.5% for maize.

Actual yields

The actual annual grain yields (Ya) were obtained from national statistical offices at the best available spatial resolution (NUTS levels 1, 2 or 3 representing increasing regional resolution; https://ec.europa.eu/eurostat/web/nuts/overview) and at the most disaggregated crop and cropping system level (spring sown or autumn sown common or durum wheat, spring sown or autumn sown barley, rainfed maize, and irrigated maize). Yield data were collected for a recent period of 10 years for rainfed crops and 5 years for irrigated crops. A linear trend analysis was carried out to identify countries with a significant (P<0.05) increasing trend. For those countries with increasing trends, only the recent 5 year period was used to avoid effects of technological developments. The actual yields were re-scaled from the NUTS region to the RWS zone, and subsequently to CZ and country, using the harvested area as weighing factor. All actual crop yields are presented at standard moisture content, i.e. 13.5% for wheat and barley, and 15.5% for maize.

More details and references are presented in Schils et al. (2018) Cereal yield gaps across Europe https://doi.org/10.1016/j.eja.2018.09.003

Legume production in Europe

Approach

Selection of countries

For faba bean we selected the following countries with their water regime: Denmark (irrigated, rainfed), Italy (rainfed), Finland (rainfed), France (rainfed), Germany (rainfed), Sweden (rainfed), United Kingdom (rainfed). And for pea the following countries with their water regime were selected: Denmark (irrigated, rainfed), Finland (rainfed), France (rainfed), Germany (rainfed), Sweden (rainfed), Romania (rainfed), Spain (rainfed, irrigated), Ukraine (rainfed, irrigated). For soybean the following countries were selected with their water regime: Italy (irrigated), republic of Moldova (rainfed), Romania (irrigated, rainfed), France (irrigated), Ukraine (irrigated, rainfed), Croatia (irrigated), Hungary (rainfed), Republic of Serbia (irrigated).

Reference weather stations (RWS)

For soybean RWS were selected based on SPAM2010 crop area mask. As for faba bean and peas no SPAM2010 crop area map is available, therefore a crop area mask was generated based on national statistics.

Weather data were collected for stations in the selected CZs with at least ten years of consecutive daily data and less than 20% missing data for each variable. Precipitation, minimum and maximum temperature, vapour pressure and wind speed were used from stations in the NOAA Global Surface Summary of the Day (https://data.nodc.noaa.gov/cgi-bin/iso?id=gov.noaa.ncdc:C00516). If suitable weather stations with adequate data were unavailable, gridded weather data from Agri4Cast (http://agri4cast.jrc.ec.europa.eu/) were used. Daily radiation was derived from NASA (http://power.larc.nasa.gov/). For precipitation, zero was assigned to missing values, while for other variables linear interpolation was used to fill data gaps.

Soils

Data on soil saturation limit, soil drained upper limit, soil extractable moisture, and root penetrable soil depth were taken from the 1 km x 1 km grid of the European Soil Database. Within the 100 km zone around each RWS, the three dominant Soil Map Units (SMUs) were selected, based on the harvested area. Each SMU comprises a varying number of Soil Type Units (STUs) of which the soil parameters were used as input to the model.

Crop modelling

For all three legumes we used the generic crop growth model Simple Simulation Model, SSM-iCrop2, a simpler version of SSM crop model (Soltani, 2012; Soltani et al., 2020b); https://sites.google.com/view/ssm-crop-models). The crop related parameters including the cardinal temperatures, thermal time requirement for the phenological stages, radiation use efficiency, light extinction coefficient, leaf area expansion parameters, dry matter partitioning coefficients were extracted from literature (Martin and Jamieson, 1996; Moot and McNeil, 1995; Sadras et al., 2019; Soltani et al., 2020a). 

For faba bean the model was validated using three European datasets, two winter sowings from Auzeville, France (Bedoussac, 2009),  a spring sowing for two years in Wageningen, the Netherlands (Wang et al., 2021), and six spring sowings for Viikii, Finland (Lizarazo et al., 2015; Skovbjerg et al., 2020).

For pea we the model was validated using two European datasets, two winter and spring sowings for five years in Auzeville, France (Bedoussac, 2009), and a spring sowing for two years in Wageningen, Netherlands (Wang et al., 2021). For Denmark, Finland and Sweden validation was done using experimental data of Antichi et al. (2023).

For soybean the model was first calibrated to fit European cultivars using three datasets. The first dataset from Geel and Merelbeke in Belgium (Pannecoucque et al., 2018) contains data from 2014 and 2015 and compares the suitability of 28 different soybean varieties under rainfed conditions. The second and third dataset is from Muncheberg in Germany (Reckling et al., 2020; Reckling and Rosner, 2020). The first dataset from this location contains data from 2014 – 2017 of two different soybean varieties, and was performed under both irrigated and rainfed conditions. A second experimental dataset from the same location contains data from 2017 – 2020, testing several soybean varieties.

SSM validation for soybean was done using two datasets. The first is from Fundulea in Romania (Cociu et al., 2013) and contains data from 2008-2012 of soybean grown under rainfed conditions. For the second dataset is data obtained from the LegValue project (http://www.legvalue.eu/) from experiments in France on 15 different locations under both irrigated and rainfed conditions, all experiments take 1 season (years 2014 – 2018). On each location on average 13 cultivars were tested.

For peas, faba bean, and soybean simulations were carried out for each STU within a RWS zone. The simulation results were up-scaled successively to SMU, RWS, CZ and country, using harvested area per STU as weighing factor. All simulated crop yields are presented at standard moisture content, i.e. 14% for peas, 16% for faba beans, and 13% for soybean.

Actual yields

The actual annual grain yields (Ya) were obtained from national statistical offices at the best available spatial resolution (NUTS levels 1, 2 or 3 representing increasing regional resolution) and at the most disaggregated crop and cropping system level.

References

Antichi, D., Pampana, S., Tramacere, L. G., Biarnes, V., Stute, I., Kadžiulienė, Ž., Howard, B., Duarte, I., Balodis, O., and Bertin, I. (2023). An experimental dataset on yields of pulses across Europe. Scientific Data 10, 708.

Bedoussac, L. (2009). Analyse du fonctionnement des performances des associations blé dur-pois d'hiver et blé dur-féverole d'hiver pour la conception d'itinéraires techniques adaptés à différents objectifs de production en systèmes bas-intrants.

Lizarazo, C. I., Lampi, A. M., Liu, J., Sontag‐Strohm, T., Piironen, V., and Stoddard, F. L. (2015). Nutritive quality and protein production from grain legumes in a boreal climate. Journal of the Science of Food and Agriculture 95, 2053-2064.

Martin, R., and Jamieson, P. (1996). Effect of timing and intensity of drought on the growth and yield of field peas (Pisum sativum L.). New Zealand Journal of Crop and Horticultural Science 24, 167-174.

Moot, D. J., and McNeil, D. L. (1995). Yield components, harvest index and plant type in relation to yield differences in field pea genotypes. Euphytica 86, 31-40.

Pannecoucque, J., Goormachtigh, S., Heungens, K., Vleugels, T., Ceusters, J., Van Waes, C., and Van Waes, J. (2018). Screening for soybean varieties suited to Belgian growing conditions based on maturity, yield components and resistance to Sclerotinia sclerotiorum and Rhizoctonia solani anastomosis group 2-2IIIB. The Journal of Agricultural Science 156, 342-349.

Reckling, M., Bergkvist, G., Watson, C. A., Stoddard, F. L., and Bachinger, J. (2020). Re-designing organic grain legume cropping systems using systems agronomy. European Journal of Agronomy 112, 125951.

Reckling, M., and Rosner, G. (2020). Data on the effect of cultivars and irrigation on soybean yield and rotational effects. Leibniz Centre for Agricultural Landscape Research (ZALF)

Sadras, V. O., Lake, L., Kaur, S., and Rosewarne, G. (2019). Phenotypic and genetic analysis of pod wall ratio, phenology and yield components in field pea. Field Crops Research 241, 107551.

Skovbjerg, C. K., Knudsen, J. N., Füchtbauer, W., Stougaard, J., Stoddard, F. L., Janss, L., and Andersen, S. U. (2020). Evaluation of yield, yield stability, and yield–protein relationship in 17 commercial faba bean cultivars. Legume Science 2, e39.

Soltani, A. (2012). "Modeling physiology of crop development, growth and yield," CABi.

Soltani, A., Alimagham, M., and Nehbandani, A. (2020a). Modeling plant production at country level as affected by availability and productivity of land and water. Agric. Syst.(Minor Revision Is Needed).

Soltani, A., Alimagham, S., Nehbandani, A., Torabi, B., Zeinali, E., Dadrasi, A., Zand, E., Ghassemi, S., Pourshirazi, S., and Alasti, O. (2020b). SSM-iCrop2: A simple model for diverse crop species over large areas. Agricultural Systems 182, 102855.

Wang, Z., Dong, B., Evers, J. B., Stomph, T., van der Putten, P., and van der Werf, W. (2021). Strip intercropping of maize, wheat, pea and faba bean in the Netherlands. In "Intercropping for sustainability", Vol. 146, pp. 221-224.

Potato production in Europe

Approach

Selection of countries

The Global Yield Gap Atlas presents actual and potential yields for both rainfed and irrigated potato cropping in the major production areas of Europe. We selected the following countries with their water regime: Belarus (rainfed), Belgium (rainfed), France (irrigated, rainfed), Germany (irrigated, rainfed), Italy (irrigated), the Netherlands (irrigated), Poland (irrigated, rainfed) and Ukraine (rainfed). Together these major production areas represent 56% of the European potato cultivation area.

 

Selected climate zones

Climate zones were selected based on the SPAM2010 crop area mask (IFPRI, 2019), and then verified with the country agronomists. The number of climate zones varied from one (the Netherlands) to nine (Poland) and in total there were 39 CZs among the eight selected countries.

 

Reference weather stations (RWS)

Within the selected climate zones, reference weather stations were selected also based on the SPAM2010 crop area mask and in consultation with the country agronomists. This resulted in a total selection of 96 different weather stations.

Daily weather data were collected for each RWS for the years 2006 – 2020. Data on precipitation, vapour pressure, wind speed and minimum and maximum temperature were obtained from the MARS database (https://agri4cast.jrc.ec.europa.eu/DataPortal/Index.Aspx) and radiation data was derived from NASA (https://power.larc.nasa.gov/).

 

Soils

Data on water content at field capacity (pF=2.5) minus water content at wilting point (pF=4.2), root penetrable soil depth and hydrogeological class of soils were taken from the 1 km x 1 km grid map of the European Soil Database in a vector format. Within the 100 km zone around each RWS, the three dominant Soil Map Units (SMUs) were selected, based on the harvested area. Each SMU comprises a varying number of Soil Type Units (STUs) of which the soil parameters were used as input to the model.

 

Crop modelling

We implemented the WOFOST crop model version 7.2, which was recently calibrated for modern potato cultivars by ten Den et al (2022). In this yield gap study, the cultivars Fontane, Innovator, Festien, Premiere and Markies were used for the simulations. Cultivar information together with planting and harvest window was obtained from the country agronomists per RWS, and depending on this information we used one of the listed cultivars or used a slightly adjusted one. Cultivar adjustment was done through correction phenology (TSUM1 and TSUM2) and was based on the provided growth duration. All simulated crop yields were set at 21.5% dry matter content.

 

Actual yields

Actual yield data was gathered from either national statistical agencies or from Eurostat at the finest available spatial resolution (NUTS levels 1, 2 or 3 representing increasing regional resolution) and collected for a period of 10 years (2012-2021). There was no distinction between irrigated and rainfed potato in the actual yield data; we corrected this based on the fraction irrigated and rainfed area and the ratio between water-limited potential yield and the potential yield.

All crop yields were standardized at 21.5% dry matter content.

 

References

IFPRI (2019). Global Spatially-Disaggregated Crop Production Statistics Data for 2010 Version 2.0.  (H. Dataverse, ed.).

ten Den, T., van de Wiel, I., de Wit, A., van Evert, F. K., van Ittersum, M. K., & Reidsma, P. (2022). Modelling potential potato yields: Accounting for experimental differences in modern cultivars. European Journal of Agronomy, 137, 126510. https://doi.org/10.1016/j.eja.2022.126510