Vol.

44: 55–68, 2010 CLIMATE RESEARCH

Published online October 28
doi: 10.3354/cr00925 Clim Res

Contribution to CR Special 22 ‘Agriculture in a changing climate’

OPEN
ACCESS

Modelling climate impacts on crop yields in Belgium

A. Gobin*
Environmental Modelling Unit, Flemish Institute for Technological Research, Boeretang 200, 2400 Mol, Belgium

ABSTRACT: In the last 2 decades, Belgium has experienced more monthly extremes than in any other
decade since observations began in 1833. During the past 60 yr, yields have increased, on average,
by 0.1 t ha–1 yr–1 for winter cereals, 0.4 t ha–1 yr–1 for potato and 0.6 t ha–1 yr–1 for sugar beet. A total of
60 to 74% of the variability in yields between 1960 and 2008 was explained by meteorological vari-
ables. Multivariate analysis showed significant statistical relationships between yield and vapour-
pressure deficit, temperature and growing season length, and water logging and drought. These vari-
ables were therefore included in a new regional dynamic crop model (REGCROP) developed to
assess climate impacts on regional arable crop production. Nash-Sutcliffe model efficiencies were
between 0.68 and 0.84 between simulated and observed national yields for the period 1960–2008.
REGCROP was subsequently run for 3 climate change scenarios and 3 typical Belgian soils (clay, loam
and loamy sand) and climate impacts were compared with historical weather impacts (1960–1989).
Higher temperatures increase crop development and shorten the growing season. Strong projected
changes in seasonality affect cumulative drainage for both winter and summer crops and result in a
drier water balance regime with climate change. Average yield losses of 12 to 27% were simulated
for sugar beet and 23 to 44% for potatoes owing to drought and heat stress. Projected losses for win-
ter cereals are 5 to 12%, mainly owing to waterlogging, whereas yield increases up to 6 or 7% are
projected as a result of temperature increases and favourable vapour-pressure deficits.

KEY WORDS: Arable crop yield · Climate change impact · Regional dynamic crop model ·
Time series · Multivariate analysis
Resale or republication not permitted without written consent of the publisher

1. INTRODUCTION track the speed of climate change (Menzel et al.

2006a).
During the past decades, climate change has had a Experimental observation of physiological plant re-
marked influence on European agriculture (Orlandini sponses to meteorological variables has resulted in the
et al. 2008, Reidsma et al. 2009). The impacts of the development of a wide range of models applied at the
2003 heat wave in Europe, with temperatures up to plant, plot or field scale, such as CropSyst (Stockle et
6°C above long-term means and precipitation deficits al. 2003), Daisy (Abrahamsen and Hansen, 2000),
up to 300 mm, resulted in an estimated loss of €13 bil- wheat simulation models (Porter et al. 1993), DSSAT
lion for the European agricultural sector (Ciais et al. (J. W. Jones et al. 2003) and STICS (Brisson et al. 2003).
2005). Global coupled climate model results show that Crop models in conjunction with regional climate mod-
future heat waves in Europe will become more intense, els enable the assessment of climate impacts on re-
more frequent and longer lasting in the 21st century gional agriculture (Jacob et al. 2007). When employed
(Meehl & Tebaldi 2004). With climate defined as the in climate impact studies, inter-model comparisons
average weather type during a period of a minimum of demonstrate that uncertainties and sources of variation
30 yr, effects are on a temporal scale that are not ac- in projected climate impacts on agriculture depend not
counted for in agricultural decision making. Conse- only on the emission scenarios and climate models, but
quently, the annual activities of farmers might not also on the crop models used (Olesen et al. 2007).

*Email: anne.gobin@vito.be © Inter-Research 2010 · www.int-res.com

56 Clim Res 44: 55–68, 2010

Crops often respond nonlinearly to changes in their climate change on 4 arable crops in Belgium were as-
growing conditions and have threshold responses; this sessed using time series and multivariate analysis of
greatly increases the importance of climatic variability meteorological and yield observations, a newly devel-
and frequency of extreme events for yield, yield stabil- oped regional dynamic crop model (REGCROP), re-
ity and quality (Porter & Semenov 2005). Many of the gional IPCC climate scenarios and economic evalua-
crop models were originally developed to predict crop tion indicators. Model results are presented for
responses to local weather conditions and might not historical observations and 3 climate-change scenarios
capture some of the current findings on feedback on 3 typical Belgian soils. Together, the 4 crops se-
mechanisms and threshold effects under new climatic lected represent €843 million or 80% of the 2008 total
conditions or extreme weather events, nor might they arable crop production in Belgium. According to 2008
be suited to applications at the regional scale. Belgian agricultural census data, winter wheat Triti-
With climate change, increased air temperature and cum aestivum L. (2086 km2) and winter barley Horde-
CO2 enrichment are projected to significantly affect um vulgare L. (488 km2) represent 71% of the total
crop phenology, reduce stomatal conductance and cereal production area. Potato Solanum tuberosum L.
transpiration, improve water-use efficiency and stimu- (639 km2) and sugarbeet Beta vulgaris L. (643 km2)
late higher rates of photosynthesis (Drake et al. 1997, each represent 37% of the industrial crop production
Chmielewski et al. 2004, Menzel et al. 2006b, Kattge & area.
Knorr 2007). Results from the Free Air Carbon Enrich-
ment (FACE) experiments cast doubts on whether CO2
fertilisation will fully offset yield losses resulting from 2. METHODS
decreased soil moisture or increased heat stress under
climate change (Ewert et al. 2002, Kimball et al. 2002, 2.1. Weather and yield data analysis
Long et al. 2006). The discrepancy between modelling
efforts and experiments is possibly related to the fact Historical weather records for the period 1947–2008
that the majority of crop models predict yield with a from the Belgian Royal Meteorological Institute (RMI)
non-limiting supply of water and near-optimum tem- were used to calculate seasonal, monthly and 10-daily
perature for crop growth. Current impacts of climate rainfall, minimum and maximum air temperature, solar
change or extreme weather events on agricultural pro- radiation, wind speed, dewpoint temperature and re-
duction enable further exploration of upper or lower lative humidity. Vapour pressure deficit (VPD) and
model boundary conditions and result in improved potential evapotranspiration (PET) were calculated
models suited to simulate climate impacts. using the FAO Penman-Monteith equation (Allen et
Seasonal patterns in agro-meteorological variables al. 1998):
have a major impact on regional crop production
e 0 (T min) + e 0 (T max)
(Wheeler et al. 2000, Challinor et al. 2003). All of the VPD = − e 0 (Tdew)
2
processes involved in agricultural production take
with (1)
( )
place at different spatial, temporal and organisational
17.27 × T
scales. Reconciliation across this range of scales re- e (T ) = 0.6108 exp
0
T + 273.3
quires explicit integration over temporal distributions
of meteorological, soil, crop and agro-economic data. where e0(T ) is saturation vapour pressure (in kPa) at
The physical basis for regional crop modelling relies the air temperature T (in °C), T min is minimum temper-
on a quantified relationship between weather and ature, T max is maximum temperature, T dew is tem-
yield data at particular temporal scales. Regional crop perature at dewpoint. Extreme values were defined as
models have been developed to a limited extent, e.g. to the days with non-zero-value 10th and 90th percentiles
investigate the effects of climate change on grain for evapotranspiration and rainfall, and days with mean
yields in China (Mo et al. 2009) and groundnut yields temperatures above 30°C. Days with VPD between 0.45
in India (Challinor et al. 2004). The focus of the present and 1.25 were used to identify healthy air-moisture
study is the development of a regional crop model that conditions for plant production; VPD values outside
can be used to assess climate impacts on regional these boundaries were considered extreme.
arable crop production. A generic soil-moisture balance was developed to
Based on a long-term temperature analysis for generate agri-meteorological variables that reflect
Ukkel, the Belgian Royal Meteorological Institute (RMI soil-moisture conditions. The soil available water
2009) distinguished a first period (1910–1987) with an capacity was set to 200 mm; the fraction of plant avail-
average mean temperature of 9.7°C followed by an ab- able water to 0.5 and the rooting depth to 1.5 m. Daily
rupt rise in 1988 resulting in a second period with an actual evapotranspiration (AET) of the soil–plant –
average mean temperature of 10.9°C. The impacts of atmosphere continuum was described as a dynamic
 Gobin: Modelling climate impacts on arable crop yields 57

function of rainfall (P) and PET, taking into account a the beginning of the growing season to the maximum
daily 75% loss of storage capacity (ST). For rainfall rooting depth at the middle of the growing season.
events < 0.1 mm during the growing season, the ratio Daily soil moisture (θt, Eq. 4) is confined between field
between actual and potential evapotranspiration was capacity (θFC) and permanent wilting point (θWP)
set at the average between 0.5 and the ratio: (Table 1); losses such as runoff (L) are assumed to be
10% of the total rainfall. Soil-moisture balance perfor-
STt
kt = with STt = 0.9 × STt −1 + Pt (2) mance indicators include deficit with respect to plant
PETt available water, surplus with respect to field capacity,
where t is the time step (i.e. 1 d). excess with respect to saturation and extreme values
The growing season was defined in accumulated defined as the days exceeding non-zero-value 10th
thermal sums above the base temperature 0°C. Vege- and 90th percentiles.
tation growth was assumed between the days of the
(kt + kt2 )
year when the accumulated thermal sum equals 100 AETt = PETt for P > 0.1 mm (3)
and when the accumulated thermal sum for a period of
(1 + kt + kt2 )
10 d is below 5, i.e. roughly from March until October. θt = θt −1 + (Pt − Lt − AETt ) × Δt (4)
The soil-moisture balance operates as a single-layer
bucket model that varies with the rooted soil layer. The All meteorological variables were integrated be-
rooting depth increases at constant rate, from 10 cm at tween average planting and harvesting dates for the 4
crops (Table 2) and Δt equals 1 d.
Table 1. Soil variables after Saxton & Rawls (2006). Textures are based on USDA Agricultural census statistics on crop
classifications yields (Belgian Federal Government
2008) were detrended using linear re-
Texture Sand Clay Moisture content (vol %) at Ksat Bulk gression. Detrending helped to define
(%) (%) Wilting Field Saturation (mm h–1) density extreme yields, in terms of yield re-
point capacity (g cm– 3)
siduals below the 10th percentile and
above the 90th percentile, which were
Loamy sand 82 6 5.7 12.1 45.7 91.3 1.44
used in further analysis of REGCROP
Silty loam 8 13 10.0 33.0 48.9 13.9 1.35
Loam 30 25 16.4 31.9 47.2 9.7 1.40 model results. A multivariate correla-
tion analysis was carried out between

Table 2. Crop parameters for 4 Belgian crops. Data based on Blaes & Defourny (2003), Allen et al. (1998) and Vanclooster et al. (1995)

Crop parameter Definition Units Winter wheat Winter barley Potato Sugar beet

Tb Base temperature °C 4 4 7 6
To Optimal temperature °C 25 25 30 30
Tmax Maximum temperature °C 35 35 35 35
Plant date – – 15 Oct 15 Oct 15 Apr 20 Mar
Harvest datea – – 01 Aug 15 Jul 30 Sep 21 Oct
PTUb Potential thermal units °C 1700 1450 1350 1800
a – 3.055 3.055 3.055 2.08
b – 13.38 13.38 13.38 15.02
LAImax Maximum leaf area index – 5.5 5.5 5.5 6
HI Harvest index – 0.60 0.50 0.95 0.60
LUE Light-use efficiency kg ha–1 MJ–1 m2 30 28 30 33
Rooting depth – m 1.60 1.60 0.50 0.95
Height – m 0.85 0.80 0.70 0.4
P factor Depletion factor – 0.55 0.55 0.35 0.55
Kcb initialc Initial crop base factor – 0.17 0.17 0.15 0.15
Kcb mid Mid-season crop base factor – 1.1 1.1 1.1 1.15
Kcb late Late-season crop base factor – 0.3 0.3 0.35 0.5
kst initial/vegetatived Sensitivity factors start; end 0.2; 0.2 0.2; 0.2 0.45; 0.45 0.5; 0.8
kst reproductived start; end 0.2; 0.6 0.2; 0.6 0.8; 0.8 0.8; 1.1
kst ripening/harvestd start; end 0.5; 0.2 0.5; 0.2 0.7; 0.2 1.0; 0.7

a
Used in all meteorological variables. bUsed in REGCROP. cFor winter cereals, 0.5 is used during dormancy if the crop
cover is >10%. dUsed in Eq. (18)
58 Clim Res 44: 55–68, 2010

annual crop yields and different meteorological vari- variety trials in central Belgium (Veredelingsstation
ables aggregated to the crop growth season, defined as Heverlee [now out of use], data available from author).
the time between average planting and harvesting Crop characteristics and parameters were linked to the
dates (Table 2). All meteorological variables and yields different growth stages and expressed in thermal time
were compared between the RMI-defined climate peri- as a fraction of the required PTU (Table 2).
ods (RMI 2009) using ANOVA. Linear regression equa- The water balance is a single-layer varying size
tions were developed between crop yield and single bucket model, adapted from Allen et al. (1998) to incor-
meteorological variables with significant correlations. porate effects of reduced growth conditions due to
Non-linear multiple regression equations were devel- water stress (drought, water logging) or heat stress
oped using the 2 most significant and non-collinear (temperature, VPD) on crop growth. Crop characteris-
meteorological variables. Multi-collinearity between tics include rooting depth, crop factors linking AET to
the meteorological variables was revealed by means of reference evapotranspiration, crop cover and the frac-
principal components derived from the correlation ma- tion of plant available water (depletion factor). The
trix. In addition, a variance minimizing non-hierarchical critical moisture content (θCR), or the point below
(k-means) cluster analysis involving the selected mete- which restricted plant transpiration takes place,
orological variables was used to investigate the effect of depends on the plant-specific depletion factor and
climate period on the regression between the selected varies with rooting depth. Water stress is defined as the
meteorological variables and yield. number of days and amount of deficit with respect to
the critical moisture level or as surplus with respect to
field capacity (θFC) and saturation (θSAT). Crop stress is
2.2. REGCROP model calculated as the degree of waterlogging (WLt) or
drought (DRt) for a period of 10 d:
A regional dynamic crop model (REGCROP) was de-
1 9 ⎛ θt − i − θFC ⎞
veloped to assess climate impacts on regional crop pro- WLt = ∑
10 i =0 ⎝ θSAT − θFC ⎠
for θt − i > θFC (5)
duction through the dynamic coupling of a water bal-
ance and a biomass production model run at a daily
1 9 ⎛ θt − i − θ WP ⎞
time step. The physical basis relies on quantified rela- DRt = ∑
10 i =0 ⎝ θCR − θ WP ⎠
for θt − i < θCR (6)
tionships between weather and yield at the regional
scale. In thermodynamic terms, the radiation use effi- The actual evapotranspiration (AETt) is a function of
ciency of crop production is defined as the ratio be- soil evaporation, plant transpiration and soil moisture.
tween energy output (carbohydrates) to energy input The soil evaporation factor (Ket) is calculated from the
(solar radiation) (Monteith & Moss 1977). The amount of rainfall interval (RIt) and the reference evapotranspira-
solar radiation intercepted depends on the seasonal dis- tion, multiplied by the fraction of bare soil, taken as the
tribution of leaf area, the development of which relies reverse of the cropped soil (Cc). Plant transpiration
on ambient temperature and moisture supply in both depends on the plant variety, growth stage and cli-
soil and atmosphere. These biophysical concepts are matic zone. Crop-specific Kcb factors (Eq. 7; Table 1)
related to the rate and duration of different crop devel- were adjusted for wind speed, relative humidity and
opment phases and are implemented in REGCROP. plant height, resulting in higher transpiration rates for
Phenological crop development in REGCROP is the majority of cases (Allen et al. 1998).
based on thermal time using planting dates and crop-
AETt = ( K cbt + K et ) PETt (7)
specific upper and lower threshold temperatures
(Table 2), which form the boundaries of phenological
K et = [1.286 − 0.274 × ln(RIt ) − 0.162 ln(PETt )] × (1 − Cc ) (8)
activity. The threshold temperatures take into account
the lag between air and soil temperatures, which is Ratios of AET and PET rates differ from Kcb factors
important for the early crop development stages. (Piccinni et al. 2009, Kjaergsaard et al. 2008), as they
Crops are grown from planting dates until the accumu- are mainly designed for irrigated crops and, hence,
lated thermal units (TU) equal the required potential may not take into account plant physiological behav-
thermal units (PTU). Depending on the crop and crop iour under drought conditions. To allow for reduced
development stage, the thermal time is adjusted for transpiration under drought stress, Eq. (2) was run in
vernalisation, dormancy and day length. For winter parallel and used when the soil moisture (θt) was below
cereals, the roots become dormant after frost and re- the critical moisture level (θCR).
sume their growth when the average air temperature The regional dynamic biomass model was adapted
exceeds the minimum required air temperature. The from the Monteith & Moss (1977) approach to incorpo-
PTU were derived on the basis of unpublished plant- rate extremes in soil moisture, temperature and vapour
ing, harvesting and general phenological data from pressure. The photosynthetic active radiation (Monsi &
 Gobin: Modelling climate impacts on arable crop yields 59

Saeki 1953) as a function of radiation (RAD, MJ m–2) ∑ TUt

( )
0.2
t =0 1 − TUIt
and leaf area index (LAI) was converted into potential for TUIt = > 0.8, LAIt = LAIat 0.8 (15)
PTU 1 − 0.8
daily increase in biomass using a light-use efficiency
factor (LUE) (Table 2) and accounting for the effects of The biomass generation part of the crop growth
day length (DL) (after Monteith & Moss 1977): model provides a feedback mechanism (Eq. 16) to
adjust leaf area development and hence actual crop
ΔBt = LUE × 0.5 × RADt × (1 − e−0.65LAI ) × DLt × Rt (9)
cover (Cc), refining the partitioning between covered
Temperature, water supply and vapour pressure are and uncovered soil:
the main climatic constraints on crop growth efficiency.
Cc = 1 − e−0.65LAIt (16)
The reduction factor (Rt) is computed as the minimum
of 4 climatic constraints for crop growth—waterlogging The water balance returns information to the crop
(WL, Eq. 5), drought (DR, Eq. 6), temperature stress growth module concerning water stress in terms of
(TR) or stress due to vapour pressure deficit (VPDR): waterlogging and soil-moisture deficit and accounting
1 2 π (T − T ) for the water-use efficiency.
TRt = ∑ sin ⎛⎝ 2 ⎡⎢⎣1 − (T t −i − To ) ⎤⎥⎦⎞⎠
3 i =0
(10) Both the standing and root biomass are derived from
max o
the daily biomass growth (ΔBt).
1 2 π (VPD − VPD )
VPDRt = ∑ sin ⎛⎝ 2 ⎡⎣⎢1 − (VPD t −i − VPDo ) ⎤⎦⎥⎞⎠ (11)
3 i =0
BAGt = BAGt −1 + (0.6 + 0.2TUI)ΔBt (17)
max o
The yield is subsequently derived from the standing
Rt = min ( WLt , DRt , TRt , VPDRt ) (12) biomass (BAG, Eq. 17) before senescence occurs, using
a harvest index (HI; Table 2). Calculations were made
This is incorporated in the VPD reduction term by set- with nested iterations to close the energy balances and
ting VPDo to 0.9 kPa and VPDmax to 2 kPa. conserve the mass balance. Reported national yields
Negative values result in zero additional growth in provide for an assessment of the model’s performance
biomass; the maximum value for Rt equals 1. Tempera- using model efficiency criteria such as the Nash-
ture stress reflects the physiological behaviour with Sutcliffe efficiency (Nash & Sutcliffe 1970) and related
respect to crop-specific optimal (To), maximum (Tmax) indicators (Willmott 1981).
and base temperatures (Tb) (Table 2). The reduction
term for VPD (Eq. 9) was developed to reflect the
response of the crop to humidity. Values below 2.3. Yield projections under climate scenarios
0.45 kPa reflect a cool and humid environment, and
values above 1.25 kPa a hot and dry environment; opti- According to global climate model predictions, Bel-
mal growth is considered to be around 0.9 kPa. VPD gium is situated on the boundary between wetter and
values above 2 kPa have been reported as the break- drier climatic regimes. Three climate-change scenarios
point for transpiration rates (Fletcher et al. 2007, were selected on the basis of a multi-criteria analysis
Rodriguez & Sadras 2007). (Ntegeka et al. 2008) of the PRUDENCE RCM runs
The leaf area index (LAI) is an important structural (Christensen & Christensen 2007), capturing the future
property of the crop canopy that influences water climate variability in Belgium (Fig. 1). Goodness-of-
interception, photosynthesis, respiration and senes- fit between observed meteorological data and control
cence. The leaf area is reduced by stresses related to runs for rainfall, temperature and evaporation were used
temperature, moisture and vapour pressure. The LAI to retain low (RCM: Arpège with GCM: HadCM3-B2),
curve (Williams et al. 1989) is adapted to incorporate medium (RCM: RCAO with GCM: HadAM3H-B2)
growth reduction factors. Based on 2 scaling factors — and high (RCM: RCAO with GCM: ECHAM4/OPYC-B2)
the temperature unit index (TUI) and the temperature scenarios. For the 3 selected PRUDENCE RCM runs, all
unit factor (TUF) — LAI follows a sigmoid curve until meteorological variables were obtained using Interac-
the crop reaches its maximum at 80% of PTU: tive Data Language scripts for the grid nearest to Ukkel
t
(central Belgium), which has the longest meteorological
∑ TUt time series, beginning in 1833, and is located in the cen-
t =0 TUIt
for TUIt = ≤ 0.8 : TUFt = (13) tre of the most productive agricultural area in Belgium.
PTU TUIt + exp (a − b × TUIt )
Yield changes under climate-change scenarios were
compared to simulations with observed meteorological
ΔLAIt = ΔTUFt × LAImax × Rt (14)
data using simulated net primary production (NPP)
After reaching its maximum, a power function de- divided by potential NPP in the case of summer crops
scribes the decline at the onset of senescence or at the and using simulated BAG divided by potential BAG in
grain filling stage in cereals: the case of winter cereals. Potential NPP and potential
 60 Clim Res 44: 55–68, 2010

Evapotranspiration (mm)
180.0 180.0
Obs Obs
160.0 160.0
Low Low
140.0 140.0
Rainfall (mm)

Medium Medium
120.0 120.0
High High
100.0 100.0
80.0 80.0
60.0 60.0
40.0 40.0
20.0 20.0
0.0 0.0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Month Month
Fig. 1. Rainfall (left) and Penman-Monteith evapotranspiration (right) for 3 climate-change scenarios (low, medium and high)
compared with 1960–1989 observations (Obs)

BAG were simulated without limiting reduction factors. Belgium, the past 2 decades have experienced more
Average yield reductions were calculated as the frac- monthly extremes than any other decade since the
tion of actual over potential biomass production (BAG, observations started in 1833 (Fig. 2).
NPP) per major soil type for each climate change Results of detrending show a high goodness-of-fit
scenario and divided by the historical yield fractions. between Belgian yields and time; R2 varied from 0.79
The yield reductions (or increases) were subsequently for sugar beet to 0.94 for winter barley (Fig. 3). During
weighted according to the proportional division of the past 60 yr, yields increased at a mean ± SE rate
agricultural parcels on each of the major soil types, as of 0.102 ± 0.004 t ha–1 yr–1 for winter wheat, 0.095 ±
derived on the basis of the 2008 land parcel informa- 0.003 t ha–1 yr–1 for winter barley, 0.440 ± 0.025 t ha–1
tion system (LPIS) database. According to LPIS, the yr–1 for potatoes and 0.580 ± 0.038 t ha–1 yr–1 for sugar
relative proportion of loamy sand:loam:clay is 12:62:26 beet. Yields differed significantly (p < 0.001) between
for winter cereals, 10:74:16 for sugar beet and for the 2 climate periods (0: pre 1987, 1: 1988–2008). Yield
33:57:10 potatoes. residuals did not significantly differ between the 2 cli-
Yield reduction indicators were developed to evalu- mate periods (p = 0.44 to 0.82). High correlations be-
ate yield changes on the basis of yield reduction factors tween arable crop yields and meteorological variables
(Eq. 12). The yield fraction of actual yield (Yact) over integrated over the cropping season (Table 3) show
potential yield (Ypot) was also expressed as the product significant correlations between yield and the temper-
of a sequence of terms based on the reduction factor ature-related variables VPD, PET, THU and Tmin, low
(Rt ) and a sensitivity factor (kst ; Table 2) in order to rainfall and soil moisture conditions. Moreover, these
weight for sensitive crop stages during the growth sea- meteorological variables are significantly different
son (S, d; Eq. 18): between the 2 climate periods (Table 4), and appear to
S reflect the long-term meteorological variability or cli-
Yact
= ∏ [1 − kst (1 − Rt )]
1/S
(18) mate condition. Correlations between yield residuals
Ypot t =1 and meteorological variables (Table 3) demonstrate
The effects of temperature and water stress were ac- the importance of the water balance (Eq. 2), the maxi-
counted for separately by replacing Rt in Eq. (18) with mum of Tmax, soil-moisture conditions and average
WLt (Eq. 5), DRt (Eq. 6), TRt (Eq. 10) or VPDRt (Eq. 12), rainfall. The majority of these meteorological vari-
thus resulting in different yield reduction indicators ables, as well as the yield residuals, are not signifi-
that enable quantification of the distance between cantly different between the 2 climate periods and ap-
actual and potential yields due to different stress types. pear to capture short-term meteorological conditions
Changing atmospheric CO2 concentrations, which can or interannual weather variability. These results sug-
be related to an increase in light-use efficiencies, have gest that detrending yields prior to investigating cli-
not been taken into account. mate impacts might have accentuated interannual
weather variability, but might have removed all or part
of the long-term climate variability alongside the ef-
3. RESULTS fects of technological developments.
The hypothesis that weather conditions have a sub-
3.1. Weather and yield observations stantial impact on crop yield in Belgium was tested by
analysing statistical relationships between yield and
The frequency and magnitude of extreme weather meteorological variables. Principal component analy-
events are likely to increase with climate change. In sis confirmed multi-collinearity among the meteoro-
 Gobin: Modelling climate impacts on arable crop yields 61

2006
8 24
1996 1997
7.6 22
7.2 Average daily rainfall per month 2009
6.8 (no. of extremes after 1990: 11) 2001 20 1949
6.4 2002 18 2008
6 1980 1991
1999 16 2007 2001
5.6
Precipitation (mm)

1974

Temperature (°C)
1963 14
5.2 2004
1965 1841 1833
4.8 1988 2001 12 1994
1957
4.4 10 1923
1990 1912
4 2007
8 1988
3.6 1902
3.2 6
2.8 4
1881
2.4 1837
2 2
1.6 1845
0 1858
1.2 –2
0.8 Monthly mean temperature
0.4 1953
–4 (no. of extremes after 1990: 9)
1955 1963
0 1959 1976 1983 –6
1997 1959 1993 2007 2006
1975 1879
–0.4 1838 1956
–8
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Month Month
Fig. 2. Monthly precipitation and temperature, median (circles) and extremes at Ukkel since 1833. Labels indicate years with
monthly extreme(s) (RMI 2009)

10 10
y = –195.25 + 0.102x; R 2 = 0.91 y = –182.13 + 0.095x; R2 = 0.94
Yield winter wheat Yield winter barley
8 Residuals 8 Residuals

6 6

4 4

2 2

0 0
Yield (t ha –1)

–2 –2
1940 1950 1960 1970 1980 1990 2000 2010 1940 1950 1960 1970 1980 1990 2000 2010

60 90
y = –1099.77 + 0.58x; R 2 = 0.79
y = –838.13 + 0.44x; R 2 = 0.84 80
Yield sugar beet
50 Yield potatoes
70 Residuals
Residuals
40 60
50
30
40
20
30
10 20
10
0
0
–10
–10
–20 –20
1940 1950 1960 1970 1980 1990 2000 2010 1940 1950 1960 1970 1980 1990 2000 2010

Fig. 3. Time series (1947–2008) of observed and detrended yield of 4 Belgian arable crops
 62 Clim Res 44: 55–68, 2010

For feature engineering

Table 3. Correlation coefficients of yield and yield residuals with meteorological variables, aggregated during the growing season
of 4 arable crops (top 5 rank). VPD: vapour pressure deficit; PET: potential evapotranspiration; TU: thermal units; Tmin and Tmax :
minimum and maximum temperature, respectively; P: rainfall. Significant at p < 0.05, unless otherwise stated (ns: not significant)

Meteorological variable Winter wheat Winter barley Potato Sugar beet

Yield
Days with VPD > 0.5 kPa 0.83 (1) 0.78 (2) 0.71 (1) 0.66 (1)
ΣPET 0.67 (2) 0.66 (3) 0.32 0.34
Average wind speed –0.67 (3) –0.66 (4) –0.69 (2) –0.66 (2)
ΣTU 0.58 (5) 0.35 0.37 (4) 0.38 (4)
Average Tmin 0.59 (4) 0.59 (5) 0.52 (3) 0.59 (3)
Σ(Days with P < 0.2 mm) 0.43 0.83 (1) 0.08 ns 0.17 ns
Σ(Days with P > 15 mm) 0.10 ns –0.03 ns 0.36 (5) 0.37 (5)
Σ(Waterlogged conditions days) –0.37 –0.37 –0.18 ns –0.19 ns
Σ(Wet moisture conditions days) 0.33 0.32 0.31 0.28
Σ(Dry moisture conditions days) 0.23 ns 0.15 ns 0.00 ns 0.00 ns
Σ(Radiation) –0.23 –0.31 –0.36 –0.36

Yield residuals
Σ(Radiation) 0.44 (1) 0.32 (1) 0.03 ns 0.02 ns
Σ(Days with dry water balance) 0.40 (2) 0.18 ns –0.18 ns –0.28 (5)
Σ(Days with P < 0.2 mm) 0.37 (3) 0.20 ns –0.32 (2) –0.37 (2)
Σ(Rainfall) –0.36 (4) –0.13 ns 0.28 (4) 0.37 (3)
Σ(Days with wet water balance) –0.31 (5) –0.16 ns 0.18 ns 0.27
Σ(Days with Tmax > 30°C) –0.05 ns –0.02 ns –0.46 (1) –0.46 (1)
Σ(Days with P > 15 mm) –0.27 –0.18 ns 0.21 ns 0.21 ns
Σ(Dry moisture condition days) 0.27 0.04 ns –0.32 (3) –0.36 (4)

logical variables. The first principal component axis Non-linear regressions explained 74% of the variabil-
(PC1) explained 42.9% of the variation in meteorolog- ity in winter wheat yield, 73% in winter barley, 70% in
ical variables and revealed high loadings from temper- potatoes and 60% in sugar beet, which demonstrates the
ature and VPD, THU and PET. PC2 explained 17.3% of effect of the selected meteorological variables on yield.
the variation and had high loadings from rainfall, A variance-minimizing non-hierarchical (k-means) clus-
water balance and soil-moisture-related variables. PC1 ter analysis involving the selected meteorological vari-
and PC2 significantly differed between the 2 climate ables demonstrates that there was a significant effect
groups (p < 0.001). For each crop, 2 non-collinear vari- of days with VPD > 0.5 (p < 0.001) on the separation of
ables with high correlations with yield (Table 3) and clusters for each of the crops. Days with rainfall < 0.2 mm
significant differences between the 2 climate periods significantly (p < 0.001) influenced the clustering of
were selected (Table 4): the sum of days with VPD winter cereal yields. For summer crops, the effect of days
> 0.5 kPa; the sum of days with rainfall < 0.2 mm, in the with rainfall >15 mm was significant at p < 0.025.
case of winter cereals; and the sum of days with rainfall Changes in single meteorological variables had a
>15 mm, in the case of summer root crops. significant impact on yields. Linear regressions be-

Table 4. Comparison of yield and meteorological variables, aggregated during the growing season of 4 arable crops, for 2 climate periods (M0:
1947–1987; M1: 1988–2008). VPD: vapour pressure deficit; P: rainfall; TU: thermal units; PET: potential evapotranspiration; Tmin: minimum
temperature. p-values: *< 0.05; **< 0.01; ***< 0.001; ns: not significant

Yield Days with Days with Days with ΣTU (°C) ΣPET (mm) Mean
(t ha–1) VPD > 0.5 kPa P < 0.2 mm P > 15 mm Tmin (°C)

Winter wheat M0 4.35 ± 1.12 49.9 ± 22.8 132.6 ± 18.0 6.0 ± 3.3 1651.4 ± 116.6 507.5 ± 56.0 4.7 ± 0.6
M1 7.78 ± 0.84*** 91.8 ± 11.9*** 151.9 ± 20.7*** 7.0 ± 2.8 (ns) 1910.0 ± 138.8*** 597.8 ± 40.8*** 6.0 ± 0.7***
Winter barley M0 4.22 ± 1.13 41.3 ±19.1 122.0 ± 17.0 5.3 ± 3.1 1385.1 ± 112.0 445.8 ± 50.1 4.2 ± 0.6
M1 7.25 ± 0.79*** 76.5 ± 10.6*** 138.5 ± 19.5*** 6.4 ± 2.9 (ns) 1610.7 ± 139.8*** 521.3 ± 33.6*** 5.5 ± 0.7***
Potatoes M0 28.75 ± 5.95 64.3 ± 32.0 79.5 ± 13.2 3.7 ± 2.3 1327.6 ± 125.2 490.6 ± 64.5 10.0 ± 0.6
M1 42.86 ± 4.58*** 110.1 ± 10.4*** 89.9 ± 13.0** 5.2 ± 2.0* 1483.3 ± 89.9*** 556.3 ± 38.0*** 11.1 ± 0.7***
Sugar beet M0 45.34 ± 8.79 68.5 ± 35.3 94.9 ± 14.8 4. 6 ± 2.5 1709.5 ± 144.2 538.2 ± 68.5 9.8 ± 0.6
M1 63.72 ± 5.73*** 119.8 ± 12.3*** 106.6 ± 14.9** 6.2 ± 2.4* 1872.3 ± 108.5*** 610.3 ± 42.9*** 10.9 ± 0.7***
 Gobin: Modelling climate impacts on arable crop yields 63

10 9
Winter wheat Winter barley
8
8
7

6 6

5
4
4
CL1 (1947–1983, 1987)
y = 2.93 + 0.026x, R² = 0.38 CL1 (1947–1981, 1983, 1987)
3 y = 2.7 + 0.034x; R² = 0.47
2 CL2 (1974, 1976, 1982,
1984–2008) CL2 (1973, 1982–2008)
y = 5.31 + 0.026x; R² = 0.27 2 y = 5.93 + 0.016x; R² = 0.25
y = 1.95 + 0.057x; R² = 0.71 y = 1.94 + 0.062x; R² = 0.69
Yield (t ha –1)

0 1
0 20 40 60 80 100 120 0 20 40 60 80 100 120
80
Potato Sugar beet
50 70

60
40

50
30
40

CL1 (1947–1985, 1987)

20 CL1 (1947–1976, 1980, 1983) 30 y = 41.13 + 0.053x; R² = 0.20
y = 24.09 + 0.056x; R² = 0.25 CL2 (1971, 1973, 1976, 1979,
CL2 (1974, 1977–2008) 1981, 1982,1986–2008)
y = 23.21 + 0.17x; R² = 0.19 20 y = 28.36 + 0.28x; R² = 0.24
10 y = 19.27 + 0.18x; R² = 0.63 y = 34.27 + 0.20x; R² = 0.50
10
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 160
Sum of VPD > 0.5 Sum of VPD > 0.5

Fig. 4. Linear regression between arable crop yield (t ha–1) and days with VPD > 0.5. Clustering is based on days with VPD > 0.5
and rainfall days above 15 mm for summer root crops or rainfall days below 0.2 mm for winter cereals

tween yields and growing days with VPD > 0.5 (Fig. 4) Sutcliffe efficiency was highest for sugar beet (0.84)
revealed that the form of the relationship did not sig- and lowest for winter wheat (0.68). The simulated win-
nificantly change between the 2 different climate peri- ter cereal yields accounted for 80% of the variation
ods. Rates of yield increase per growing day with VPD in the observed yields (Fig. 5). Overall, winter cereal
> 0.5 kPa were not significantly different between the yields were overestimated, with an overprediction of
clusters; the intercept of the regression, however, was the lower yields. Simulated sugar beet yields captured
significantly different between the clusters. The results 84% of the variation in observed yields, with an over-
confirm that the majority of climate impacts took place estimation of lower yields. The simulated potato yields
between the 1970s and 1980s, with increasingly more accounted for 83% of the variation in observed yields;
years belonging to a warmer climatic regime. lower potato yields were overestimated and higher
yields were underestimated.
On the basis of detrending analysis (Fig. 3), years
3.2. REGCROP model output with yield residuals below the 10th percentile or
above the 90th percentile were defined as extreme.
Not only the frequency and magnitude of mete- The quantitative measures of model performance
orological events, but also their timing in relation to showed that the differences between simulated and
crop development will determine their effects. The observed yields are not the same for years with ex-
REGCROP model output shows that yields are simu- treme yields as compared with normal years (Table 5).
lated well, with a tendency to slightly overestimate For all arable crops, the normal year simulations (N;
with a relative difference of 0.5 to 4% between ob- Table 5) did not differ from the overall simulated
served and simulated yields (Table 5). The Nash- years (All; Table 5), but they differed from the years
64 Clim Res 44: 55–68, 2010

Table 5. Quantitative measures of REGCROP model performance according to (p = 0.05) and WL (p = 0.05). Model
high, low and normal years. H: 90th percentile; L: 10th percentile; N: other performances for extreme summer
yields; All: all years; Obs: mean observed yield (t ha–1); Sim: mean simulated
root crops were different from normal
yield (t ha–1); MBE: mean bias error; MAE: mean absolute error; RMSE: root
mean square error; E: model efficiency; d: index of agreement years (Table 5); differences in mean
and absolute error between normal
and low-yielding years were signifi-
Class Obs Sim MBE MAE RMSE E R2 d
cant (potato, p = 0.02; sugar beet, p =
Winter wheat 0.04), but not for the high-yielding
All 10.395 10.834 0.439 0.910 1.056 0.68 0.80 0.928 years (p > 0.05). High and low potato
H 12.808 13.072 0.264 0.550 0.668 0.93 0.72 0.985 yields were related to significant dif-
L 8.443 9.374 0.931 0.968 1.126 0.75 0.66 0.901
N 10.501 10.868 0.367 0.937 1.076 0.61 0.79 0.920 ferences in the yield reduction indica-
Winter barley
tors for temperature (p = 0.05) and
All 9.564 9.845 0.281 0.651 0.859 0.71 0.80 0.933 VPD (p = 0.004). Extreme sugar beet
H 11.510 11.768 0.258 0.958 1.155 0.66 0.46 0.932 yields differed significantly in yield-
L 8.218 9.015 0.797 0.964 1.228 0.67 0.68 0.879 reduction indicators for temperature
N 9.467 9.668 0.201 0.551 0.719 0.74 0.85 0.945
(p = 0.01) and VPD (p = 0.05). Differ-
Potato ences between simulated and ob-
All 36.249 37.276 1.027 2.551 3.504 0.78 0.83 0.927
H 44.814 43.702 –1.112 1.605 2.066 0.95 0.86 0.987 served yields were the same in years
L 28.245 32.778 4.533 5.850 6.744 0.58 0.60 0.786 with extreme and normal yields, with
N 35.918 36.776 0.858 2.185 2.881 0.79 0.82 0.936 the exception of low tuber crop yields.
Sugar beet
All 55.574 55.876 0.302 2.764 3.558 0.84 0.84 0.957
H 57.886 57.680 –0.206 3.373 4.144 0.656 0.68 0.913 3.3. Effects of climate change
L 49.654 52.998 3.344 3.449 4.846 0.762 0.81 0.916
N 56.037 56.014 –0.023 2.598 3.270 0.868 0.87 0.965
Although yearly balances fail to re-
flect seasonal dry spells or soil water
with extreme yields. Model performances for extreme storage, a comparison between scenarios or between
winter cereal yields were different from normal years crops under a certain climate scenario provides an
(Table 5), but differences in mean and absolute error indication of water shortage and surplus (Fig. 6). Sum-
were not significant at the 0.05 level. In the case of mer root crops experience increased drought stress,
winter wheat, differences between high- and low- particularly when the sensitive crop stages coincide
yielding years were linked to significant differences with dry spells. Yields may be further reduced when
in different yield reduction indicators for temperature wet springs cause water logging in the field and delay
(p = 0.05), VPD (p = 0.044) and WL (p = 0.04). Differ- planting dates. Rising temperatures enable earlier
ences between high and low winter barley yields maturity of winter cereals, which, in turn, will reduce
could be linked to significant differences in the yield yield reductions due to drought stress or increased
reduction indicators for temperature (p = 0.04), VPD heat stress during the flowering period. Despite lower

16 80
Simulated yield (grain + straw) (t ha–1)

Simulated yield (t ha–1)

14

60
12

10
40

8 Sugar beet
Winter barley
Winter wheat Potato
6 20
6 8 10 12 14 20 40 60 80
Observed yield (grain + straw) (t ha–1) Observed yield (t ha–1)
Fig. 5. Simulated versus observed yields for winter cereals (left) and summer root crops (right)
 Gobin: Modelling climate impacts on arable crop yields 65

1200 1200
Winter wheat Year Rain Winter barley
Season ET0
1000 Season Rain 1000
ETa-BL
ETa-BLZ
800 ETa-BU 800

600 600

400 400
Water balance (mm)

200 200

0 0
90

90
80

95

80

95
85

85
70

70
75

75
20

20
20

20

20

20
20

20
20

20
20

20
1200 1200
Potato Sugar beet
1000 1000

800 800

600 600

400 400

200 200

0 0
90

90
80

95

80

95
85

85
70

70
75

75
20

20
20

20

20

20
20

20
20

20
20

20

Fig. 6. Water balance components for arable crops on 3 typical Belgian soils (BL: Belgian loam; BLZ: Belgian loamy sand; BU: Bel-
gian clay) for the high climate-change scenario. ET0: potential evapotranspiration (PET); ETa: actual evapotranspiration (AET)

2500
Winter wheat BL-LOW Potato
6000 BL-LOW
BLZ-LOW BLZ-LOW
Cumulative drainage (mm)

BU-LOW BU-LOW
BL-MID 2000 BL-MID
5000
BLZ-MID BLZ-MID
BU-MID
BU-MID
4000 BL-HIGH
BLZ-HIGH 1500 BL-HIGH
BU-HIGH BLZ-HIGH
3000 BL BU-HIGH
BLZ BL
BU 1000
BLZ
2000 BU

500
1000

0 0
2070 2075 2080 2085 2090 2095 2100 2070 2075 2080 2085 2090 2095 2100

Fig. 7. Cumulative drainage for winter wheat (left) and potato (right) under 3 typical Belgian soils (loamy sand, loam and clay)
for 3 climate-change scenarios (2070–2099; low, medium and high) and observed weather conditions (1960–1989), projected
on the same axis

summer precipitation predictions for future climate in tion, drainage under winter cereals is expected to de-
Belgium, winter cereals may suffer from water logging crease in heavy textured soils (clay and loam soils in
in the field, particularly during early spring, which is Fig. 7). Rising temperatures cause higher evapotran-
the case for the low climate-change scenario. spiration rates and result in a drier water balance re-
Strong projected changes in seasonality (Fig. 1) af- gime with climate change. Only for sandy soils was a
fect cumulative drainage for both winter and summer higher drainage under winter cereals simulated for the
crops. Despite projections of higher winter precipita- low and high climate-change scenarios as compared
 66 Clim Res 44: 55–68, 2010

25 25
Winter wheat OBS Winter barley OBS
20 LOW 20 LOW
MID MID
15 15
HIGH HIGH
10
Yield reduction indicators (%)

10

5 5

0 0
VPD TR DR WL R VPD TR DR WL R

80 80
Potato OBS Sugar beet OBS
LOW LOW
60 60
MID MID
HIGH HIGH
40 40

20 20

0 0
VPD TR DR WL R VPD TR DR WL R
Fig. 8. Yield reduction indicators for arable crops under 3 climate-change scenarios (low, medium and high). Error bars indi-
cate ranges as simulated for 3 typical Belgian soils (loamy sand, loam and clay). VPD: vapour pressure deficit; TR: temperature
reduction; DR: drought reduction; WL: water logging; R: overall reduction

with the historical time series. This can be explained 4. DISCUSSION

by high spring rainfall in the low climate-change sce-
nario and high autumn rainfall in the high climate- Examples from other regions in the world have
change scenario (Fig. 1) in combination with high infil- demonstrated the usefulness of detrending yields in
tration rates on sandy soils. For summer crops, the dual relation to seasonal weather (e.g. agricultural insur-
effect of rising temperatures and lower rainfall results ances: Bielza et al. 2008) or climate variability (Nicholls
in lower cumulative drainage under all scenarios of cli- 1997, Trnka et al. 2007, Baigorria et al. 2008) to amplify
mate change, as compared with historical data (Fig. 7). the interannual yield variations owing to weather fac-
Yield reduction indicators (Fig. 8) show a favourable tors and to control for technological advances such as
effect of VPD and higher temperatures on winter improved varieties and management. The results of
cereal yields; increased waterlogging in winter and the present study demonstrate that climate and tech-
early spring has negative effects on winter cereal nological advances cannot easily be separated using
yields. Poor summer root crop yields are attributed to detrending analysis. Yields differed significantly be-
drought in late spring and summer, and to higher sum- tween the 2 climatic periods (pre- and post-1987) and
mer temperatures, which cause heat stress. For the 3 were linked to meteorological variables; yield residu-
climate change scenarios (low, medium, high), average als, however, did not differ significantly. Yield trends
yield losses of 12% (low), 17% (medium) and 27% contain information on the effects of climate variability.
(high) were simulated for sugar beet and 23% (low), Technological advances might also indirectly result
37% (medium) and 44% (high) for potatoes as com- from climate change, such as improved varieties that
pared with historical yields. Projected losses were 5% allow for earlier sowing dates. Effects of climate and
for winter wheat and 12% for winter barley under a technological advances on annual yield rates are
low climate-change scenario. Yield increases of 6% for closely related but can be separated using data from
winter wheat and 3% for winter barley were simulated variety trials, as demonstrated by Jaggard et al. (2007).
for the medium scenario. For the high scenario, pro- Understanding climate–yield relationships and the
jected yield increases were 2% for winter wheat and effects of recent climate trends on regional crop pro-
7% for winter barley. ductivity is an important step in predicting agricultural
 Gobin: Modelling climate impacts on arable crop yields 67

production at the national scale (Lobell & Asner 2003, effects of cropping calendars, weather and soil on crop
Tao et al. 2008). Results from Europe confirm the effect yields at the regional scale.
of climate change on national agricultural production A drier water balance regime is projected with all
(Moonen et al. 2002, Chmielewski et al. 2004, Trnka et the selected climate-change scenarios. Pronounced
al. 2009). A total of 60 to 74% of the variability in Bel- yield losses, mainly owing to drought and heat stress,
gian arable crop yields is explained by meteorological occur for all scenarios, to a lesser extent in the case of
variables during the growing season. At the Belgian winter cereals.
national scale, higher winter cereal yields are attrib-
uted to higher temperatures, drier springs and more LITERATURE CITED
days with favourable VPD values. Low tuber crop
yields, however, are caused by drought and heat ➤ Abrahamsen P, Hansen S (2000) DAISY: an open soil–crop–
atmosphere system model. Environ Model Softw 15:
stress. These findings are consistent with other re-
313–330
search. By the time heat waves strike, winter sown Allen RG, Pereira LS, Raes D, Smith M (1998) Crop evapo-
cereals have terminated growth and are less affected transpiration: guidelines for computing crop require-
by the summer drought, as confirmed by an analysis of ments. Irrigation and Drainage Paper No. 56, FAO, Rome
the 2003 heat wave impact across Europe (Ciais et al. ➤ Baigorria GA, Hansen JW, Ward N, Jones JW, O’Brien JJ
(2008) Assessing predictability of cotton yields in the
2005). Tuber crop yields in extreme years tend to be southeastern United States based on regional atmospheric
highly variable: drought stress was the major factor circulation and surface temperatures. J Appl Meteorol Cli-
causing yield loss of sugar beet (Pidgeon et al. 2001) matol 47:76–91
and potato (2t’astná et al. 2010). Belgian Federal Government (2008) Landbouw: statistieken
en cijfers. Belgian Federal Government, Brussels, available
Yields were projected with the REGCROP model and
at http://statbel.fgov.be/nl/statistieken/cijfers/economie/
effects were calculated for 4 arable crops in Belgium landbouw/index.jsp (in Dutch)
under 3 scenarios of climate change. Increased fre- Bielza M, Costanza G, Catenaro R, Gallego FJ (2008) Regional
quency of heat stress, drought and floods negatively yield insurance for arable crops in EU-27. Paper presented
affected yields. Analysis from several studies suggests at the 12th Congress of the European Association of Agri-
cultural Economists, Ghent, Belgium, 26–28 Aug 2008.
that crops with the highest relative increases in yield University Press, Ghent
are the most adaptable to interannual weather vari- Blaes X, Defourny P (2003) Retrieving crop parameters based
➤
ability and climate change, with the best adapted crop on tandem ERS 1/2 interferometric coherence images.
being winter wheat and the least adapted potato. Remote Sens Environ 88:374–385
➤ Brisson N, Gary C, Justes E, Roche R and others (2003) An
Under future climate scenarios, wheat may be resilient
overview of the crop model STICS. Eur J Agron 18:309–332
to drought but not heat stress (Semenov 2009). Potato Challinor AJ, Slingo JM, Wheeler TR, Craufurd PQ, Grimes
➤
(Holden et al. 2003) and sugar beet (P. D. Jones et al. DIF (2003) Toward a combined seasonal weather and crop
2003) yields will increasingly suffer from the effects of productivity forecasting system: determination of the
drought. working spatial scale. J Appl Meteorol 42:175–192
➤ Challinor AJ, Wheeler TR, Craufurd PQ, Slingo JM, Grimes DIF
(2004) Design and optimisation of a large-area process-based
model for annual crops. Agric For Meteorol 124:99–120
5. CONCLUSIONS ➤ Chmielewski FM, Muller A, Bruns E (2004) Climate changes
and trends in phenology of fruit trees and field crops in
Germany, 1961–2000. Agric For Meteorol 121:69–78
The sensitivity of agricultural yield variability to
➤ Christensen JH, Christensen OB (2007) A summary of the
climate change depends on future technological PRUDENCE model projections of changes in European
progress, crop adaptation and meteorological effects, climate by the end of this century. Clim Change 81:7–30
among many other factors. The results show that ➤ Ciais Ph, Reichstein M, Viovy N, Granier A and others (2005)
detrending yields prior to studying climate impacts Europe-wide reduction in primary productivity caused by
the heat and drought in 2003. Nature 437:529–533
may not capture all climatic trends. Multivariate analy- Drake BG, Gonzalez-Meler MA, Long SP (1997) More effi-
➤
sis showed that there were significant statistical rela- cient plants: a consequence of rising atmospheric CO2?
tionships between yield and VPD, temperature, grow- Annu Rev Plant Physiol Plant Mol Biol 48:609–639
ing season length, water logging and drought, and that ➤ Ewert F, Rodriguez D, Jamieson P, Semenov MA and others
(2002) Effects of elevated CO2 and drought on wheat: test-
these meteorological variables significantly differed
ing crop simulation models for different experimental and
between 2 climate periods. The meteorological vari- climatic conditions. Agric Ecosyst Environ 93:249–266
ables were therefore included in a new regional ➤ Fletcher AL, Sinclair TR, Allen LH Jr (2007) Transpiration
dynamic crop model (REGCROP) developed to assess responses to vapor pressure deficit in well watered ‘slow-
climate impacts on regional arable crop production. wilting’ and commercial soybean. Environ Exp Bot 61:
145–151
The model predicts historical yields well and is able to Holden NM, Brereton AJ, Fealy R, Sweeney J (2003) Possible
➤
simulate extreme yields, which may approximate cli- change in Irish climate and its impact on barley and potato
mate change conditions. The model can integrate the yields. Agric For Meteorol 116:181–196
 68 Clim Res 44: 55–68, 2010

➤ Jacob D, Barring L, Christensen OB, Christensen JH and oth- ➤ Olesen JE, Carter TR, Diaz-Ambrona CH, Fronzek S and others
ers (2007) An inter-comparison of regional climate models (2007) Uncertainties in projected impacts of climate change
for Europe: model performance in present-day climate. on European agriculture and ecosystems based on scenar-
Clim Change 81:31–52 ios from regional climate models. Clim Change 81:123–143
➤ Jaggard KW, Qi A, Semenov MA (2007) The impact of climate Orlandini S, Nejedlik P, Eitzinger J, Alexandrov V and others
change on sugarbeet yield in the UK: 1976–2004. J Agric (2008) Impacts of climate change and variability on euro-
Sci 145:367–375 pean agriculture results of inventory analysis in COST 734
Jones JW, Hoogenboom G, Porter CH, Boote KJ and others countries. Trends Dir Clim Res 1146:338–353
(2003) The DSSAT cropping system model. Eur J Agron ➤ Piccinni G, Ko JH, Marek T, Howell T (2009) Determination of
18:235–265 growth-stage-specific crop coefficients (K-C) of maize and
Jones PD, Lister DH, Jaggard KW, Pidgeon JD (2003) Future sorghum. Agric Water Manage 96:1698–1704
climate impact on the productivity of sugar beet (Beta vul- ➤ Pidgeon JD, Werker AR, Jaggard KW, Richter GM, Lister DH,
garis L.) in Europe. Clim Change 58:93–108 Jones PD (2001) Climatic impact on the productivity of
➤ Kattge J, Knorr W (2007) Temperature acclimation in a bio- sugar beet in Europe, 1961–1995. Agric For Meteorol 109:
chemical model of photosynthesis: a reanalysis of data 27–37
from 36 species. Plant Cell Environ 30:1176–1190 ➤ Porter JR, Semenov MA (2005) Crop responses to climatic
➤ Kimball BA, Kobayashi K, Bindi M (2002) Responses of agri- variation. Philos Trans R Soc B 360:2021–2035
cultural crops to free-air CO2 enrichment. Adv Agron 77: ➤ Porter JR, Jamieson PD, Wilson DR (1993) Comparison of the
293–368 wheat simulation models, AFCWHEAT2, CERES-Wheat
➤ Kjaergsaard JH, Plauborg F, Mollerup M, Petersen CT, Hansen and SWHEAT for non-limiting conditions of crop growth.
S (2008) Crop coefficients for winter wheat in a sub-humid Field Crops Res 33:131–157
climate regime. Agric Water Manage 95:918–924 ➤ Reidsma P, Ewert F, Lansink AO, Leemans R (2009) Vulnera-
➤ Lobell DB, Asner GP (2003) Climate and management contri- bility and adaptation of European farmers: a multi-level
butions to recent trends in US agricultural yields. Science analysis of yield and income responses to climate variabi-
299:1032 lity. Reg Environ Change 9:25–40
➤ Long SP, Ainsworth EA, Leakey ADB, Nosberger J, Ort DR RMI (Royal Meteorological Institute) (2009) Klimatologisch
(2006) Food for thought: lower-than-expected crop yield overzicht van het jaar 2008. Royal Meteorological Institute
stimulation with rising CO2 concentrations. Science 312: of Belgium, Brussels, available at www.meteo.be/meteo/
1918–1921 view/nl/2827848-2008.html (in Dutch)
➤ Meehl GA, Tebaldi C (2004) More intense, more frequent, ➤ Rodriguez D, Sadras VO (2007) The limit to wheat water use
and longer lasting heat waves in the 21st century. Science efficiency in eastern Australia. I. Gradients in the radiation
305:994–997 environment and atmospheric demand. Aust J Agric Res
➤ Menzel A, von Vopelius J, Estrella N, Schleip C, Dose V 58:287–302
(2006a) Farmers’ annual activities are not tracking the ➤ Saxton KE, Rawls WJ (2006) Soil water characteristic esti-
speed of climate change. Clim Res 32:201–207 mates by texture and organic matter for hydrologic solu-
➤ Menzel A, Sparks TH, Estrella N, Koch E and others (2006b) tions. Soil Sci Soc Am J 70:1569–1578
European phenological response to climate change matches ➤ Semenov MA (2009) Impacts of climate change on wheat in
the warming pattern. Glob Change Biol 12:1969–1976 England and Wales. J R Soc Interface 6:343–350
➤ Mo XG, Liu SX, Lin ZH, Lin ZH, Guo RP (2009) Regional crop 2t’astná M, Toman F, Dufkova J (2010) Usage of SUBSTOR
yield, water consumption and water use efficiency and model in potato yield prediction. Agric Water Manag 97:
their responses to climate change in the North China 286–290
Plain. Agric Ecosyst Environ 134:67–78 ➤ Stockle CO, Donatelli M, Nelson RL (2003) CropSyst, a crop-
Monsi M, Saeki T (1953) Über den Lichfaktor in den ping systems simulation model. Eur J Agron 18:289–307
Pflanzengesellschaften und seine Bedeutung für die Stoff- Trnka M, Hlavinka P, Semeradova D, Dubrovsky M, Zalud Z,
produktion. Jpn J Biotechnol 14:22–52 Mozny M (2007) Agricultural drought and spring barley
➤ Monteith JL, Moss CJ (1977) Climate and the efficiency of crop yields in the Czech Republic. Plant Soil Environ 53:306–316
production in Britain. Philos Trans R Soc B 281:277–294 Tao F, Hayashi Y, Zhang Z, Sakamoto T, Yokozawa M (2008)
➤ Moonen AC, Ercoli L, Mariotti M, Masoni A (2002) Climate Global warming, rice production, and water use in China:
change in Italy indicated by agrometeorological indices developing a probabilistic assessment. Agric For Meteorol
over 122 years. Agric For Meteorol 111:13–27 148:94–110
➤ Nash JE, Sutcliffe JV (1970) River flow forecasting through Trnka M, Eitzinger J, Hlavinka P, Dubrovsky M and others
conceptual models. I. A discussion of principles. J Hydrol (2009) Climate-driven changes of production regions in
10:282–290 central Europe. Plant Soil Environ 55:257–266
➤ Nicholls N (1997) Increased Australian wheat yield due to ➤ Vanclooster M, Viaene P, Diels J, Feyen J (1995) A determin-
recent climate trends. Nature 387:484–485 istic validation procedure applied to the integrated soil
Ntegeka V, Boukhris O, Willems P, Baguis P, Roulin E (2008) crop model WAVE. Ecol Modell 81:183–195
Climate change impact on hydrological extremes along ➤ Wheeler TR, Craufurd PQ, Ellis RH, Porter JR, Prasad PVV
rivers and urban drainage systems in Belgium. II. Study of (2000) Temperature variability and the yield of annual
rainfall and ETo climate change scenarios. CCI-HYDR crops. Agric Ecosyst Environ 82:159–167
project. Report to the Belgian Federal Ministry of Science Williams JR, Jones CA, Kiniry JR, Spanel DA (1989) The EPIC
Policy. Catholic University, Leuven, available at www. crop growth model. Trans ASAE 32:497–511
kuleuven.be/hydr/cci/reports/CCI-HYDR_II-Climate Willmott CJ (1981) On the validation of models. Phys Geogr 2:
ChangeScenarios.pdf 184–194

Submitted: March 1, 2010; Accepted: August 17, 2010 Proofs received from author(s): September 27, 2010