180 P.A. Harrison et al. Agricultural and Forest Meteorology 101 2000 167–186
model to AFRCWHEAT2 and NWHEAT, but a greater value was found for the SIRIUS model. The range
of predicted dates from the spatial model overlaps to a considerable degree with all the site-based models.
Differences in the predictions between the broad-scale model and the site models are not significantly greater
than differences between the site models themselves. The timing of other development stages at Seville also
fell within the range of the site model predictions.
4. Model sensitivity
The broad-scale development model performs sat- isfactorily under the range of climatic variation for
which it has been developed and validated. However, a change in mean climate, which may arise from the
enhanced greenhouse effect, would lead to a shift in the range of climatic conditions under which winter
wheat is grown in the future at a given location. Cou- pled with the likelihood of a change in mean climatic
conditions, there is also the possibility of a change in the variability of climate. Studies have shown that
crops show non-linear responses to changes in envi- ronmental variables Semenov and Porter, 1995; Moot
et al., 1996. Hence, it is important to test both the robustness and sensitivity of the broad-scale model to
changes in the mean and variability of individual input variables.
4.1. Changes in mean temperature The sensitivity of the spatial development model
to incremental increases in minimum and maximum temperature was evaluated. Temperatures were in-
creased by 1–4
◦
C in 1
◦
C intervals in a uniform manner throughout the year. As temperatures increase
there is a progressive northward and eastward ex- pansion in the area of suitability due to a longer and
warmer growing season. A 4
◦
C increase in temper- ature causes the limit of suitability to extend into
mid Fenno-Scandinavia and most of Russia within the study region. The area of winter wheat suitability
expands by, on average, 400,000 km
2
or 5.4 of the area of suitability under current conditions per 1
◦
C warming. This equates to a northward expansion of
the limit of production in Finland of approximately 90 km per 1
◦
C warming, on average. The rate of development increases with higher mean
temperatures causing a reduction in the length of the growing period. This is because thermal time accu-
mulates more rapidly with warmer temperatures and, hence, the thresholds of thermal time for each devel-
opment stage are reached faster. An increase of 1
◦
C causes the duration from sowing to maturity to de-
crease by 1–2 weeks across most of Europe. For each further 1
◦
C increase in temperature the length of the growing period reduces by, on average, approximately
a week. With a 4
◦
C increase the duration from sow- ing to maturity decreases by 5–6 weeks in northwest
Europe and by 3–4 weeks in southern and eastern Eu- rope.
Increases in temperature affect the rate of develop- ment during different development phases to different
degrees. The effect of a 2
◦
C rise in temperature on three development phases is shown in Fig. 8 for cul-
tivar Avalon. There is a strong reduction in the length of the phase between sowing and double ridges of
4–8 weeks in northwest Europe and 1 and 3 weeks in southern and eastern Europe. The pattern of response
is reversed for the phase from double ridges to anthe- sis. Increases in phase length of 1–3 weeks are found
in northwest Europe, whilst either small increases or reductions of 1–2 weeks are observed in southern and
eastern Europe. The large reductions in the duration of the previous phases up to double ridges cause this
phase to occur significantly earlier in the season. Tem- peratures experienced during this phase are actually
cooler than at present, even with an increase in tem- perature, because of its earlier timing. The length of
the grain filling period is reduced by less than 1 week throughout most of Europe. Changes to this phase are
relatively small because it has also moved to slightly earlier in the season due to the combined effects of
changes in the previous phases.
The spatial pattern of developmental rate response to higher temperatures is similar for all six cultivars.
The greatest increases occur in northwest Europe and the smallest in southern Europe. There are several
reasons for this phenomenon. Firstly, temperatures are closer to the base temperature over winter in northern
Europe and, hence, a small increase in temperature has a much more substantial effect on development than
when the same increase occurs in southern Europe. Secondly, the effectiveness of winter temperatures for
vernalization are generally below the optimum range
P.A. Harrison et al. Agricultural and Forest Meteorology 101 2000 167–186 181
Fig. 8. Change in duration of three winter wheat cv. Avalon development phases following a uniform increase in mean minimum and maximum temperature of 2
◦
C: a sowing to double ridges; b double ridges to anthesis; and c beginning to end of the grain filling period.
in northern Europe, but within this range in southern Europe. Hence, the vernalization response is more sen-
sitive to increases in temperature in northern Europe. Thirdly, photoperiod differences between northern
and southern Europe affect the accumulation of tem- peratures which determines development over winter
and spring through photo-vernal-thermal time and photo-thermal time, respectively. In winter, photope-
riod is longer in southern Europe, but the influence of photoperiod is low due to the inhibiting effects of the
182 P.A. Harrison et al. Agricultural and Forest Meteorology 101 2000 167–186
vernalization factor. However, photoperiod is the only factor influencing the accumulation of temperatures
in spring from double ridges to anthesis. Days are shorter in southern Europe in spring and this reduces
the effectiveness of the same temperature in contribut- ing to thermal time, thus, resulting in a lower sensitiv-
ity of development to increases in temperature. Butter- field and Morison 1992 reached similar conclusions
in a comparison of the sensitivity of winter wheat development at northern and southern sites in the UK.
There are small differences in the magnitude of re- sponse between the cultivars. The cultivars that simu-
late development after anthesis using a base tempera- ture of 9
◦
C e.g., Hustler, Caribo and Alcala are more sensitive to increases in temperature than those culti-
vars that assume a base of 0
◦
C e.g., Avalon, Riband and Slepner. This may not accurately reflect physical
differences between the varieties, but rather be a con- sequence of model formulation and calibration. Fur-
ther, Caribo, which has a narrower range of vernaliz- ing temperatures than all the other cultivars, exhibits
the most non-linear response to increases in tempera- ture. This results in smaller reductions in the length of
the over winter development stages and a larger short- ening of the grain filling period. Cultivar Alcala ex-
periences a similar reduction in the over winter devel- opment stages to Avalon. However, decreases in the
length of the grain filling period are smaller and in some parts of Europe small increases in the length of
this phase occur. This is because the timing of the grain filling period is fairly early under current climatic con-
ditions, ranging from May to June. Hence, a forward shift in the timing of this period due to higher tem-
peratures causes it to move from summer into spring when temperatures are significantly cooler.
4.2. Changes in temperature variability The sensitivity of the broad-scale model to both in-
creases and decreases in temperature variability was evaluated. Doubling the inter-annual variability of
temperatures causes the mean date of double ridges to occur from 1 to 20 days earlier throughout central,
northern and eastern Europe. The greatest shift in the date of this stage is centred over southern Sweden,
Denmark, northern Germany, Poland and the Baltic states. A slightly later date of double ridges, by up
to 5 days, is predicted in Ireland, western France and the Mediterranean region. The mean date of anthesis
and maturity change to a lesser extent, but in a sim- ilar pattern to that described for double ridges. Both
stages occur slightly later, by 1–5 days, in Ireland, western France and the Mediterranean region and
slightly earlier, by 1–8 days, throughout the rest of Europe. Alternatively, halving the annual temperature
variability causes either no change or a very slight delay of 1–2 days in the mean date of occurrence of
all phenological stages.
The standard deviation of all phenological stages increases with a doubling of temperature variability
and decreases with a halving of temperature variabil- ity. The largest increases or decreases in standard de-
viation for a doubling or halving of variability, respec- tively, occur in eastern Europe. Increases range from
3 to 13 days for a doubling and decreases range from 0 to 6 days for a halving of temperature variability.
Changes in the standard deviation of anthesis are sim- ilar to those described for maturity. The standard de-
viation of double ridges changes to a much greater extent for both perturbations. A doubling of variabil-
ity causes the largest increases in standard deviation, ranging from 12 to 20 days, to occur in southern
Sweden, Denmark, northern Germany, Poland and the Baltic states. This corresponds with the greatest shift
earlier in the mean date of double ridges. A halving of temperature variability causes the largest decreases in
standard deviation, ranging from 6 to 8 days, to occur in northwest Europe.
5. Discussion and conclusions