The vulnerability of winter wheat in Germany to air temperature, precipitation or compound extremes is shaped by soil-climate zones

Abstract

Whether hydroclimatic extremes cause yield losses or failures not only depends on their intensity but also on local environmental conditions. These conditions shape the capacity to buffer climatic shocks and thus necessitate a regionally specific impact assessment and adaptation planning. However, the degree to which different environmental conditions affect climate impacts on yields and its spatiotemporal variability across Germany is relatively unknown. In this study, we use a regression-based crop-climate modelling approach for 71 regions, classified according to soil and climate characteristics and investigate region-specific vulnerabilities of winter wheat yields to hydroclimatic extremes for the period 1991-2019. We account for the co-occurrence of temperature and moisture impacts (i.e. compound effects) as well as for local soil-climate conditions. On average, our models can explain approx. 67 % of past winter wheat yield variations. Despite the rather homogeneous climate in Germany, the results reveal clear geographic differences across different soil-climate regions. While the northeastern regions show a clear dominance of drought impacts, southern regions show stress due to moisture excess. Heat impacts can clearly be linked to the warm regions along the western part of the country. Overall, compound dry-hot extremes pose the strongest and most widespread risk for winter wheat yields in Germany, being responsible for approx. 38 % and in some regions for up to 50 % of past yield variations. Based on the identified regional differences in hydroclimate susceptibility, we can define four geographic risk clusters, which exhibit vulnerability to climatic extremes such as summer droughts, winter droughts, summer heat waves, and winter moisture excess. The identified risk clusters of heat and moisture stresses could inform regional-specific adaptation planning.

Full text

The vulnerability of winter wheat in Germany to air temperature,
precipitation or compound extremes is shaped by soil-climate zones
Rike Becker
a,b,c,*
, Bernhard Schauberger
b,d
, Ralf Merz
e
, Stephan Schulz
f
, Christoph Gornott
a,b
a
University of Kassel, Nordbahnhofstr. 1a, 37213 Witzenhausen, Germany
b
Potsdam Institute for Climate Impact Research, Telegraphenberg 31, 14473 Potsdam, Germany
c
Imperial College London, South Kensington Campus, London SW7 2AZ, UK
d
University of Applied Sciences Weihenstephan-Triesdorf, Am Staudengarten 1, 85354 Freising, Germany
e
UFZ Helmholz Centre for Environmental Research, Theodor-Lieser-Straße 4, 06120 Halle, Germany
f
TU Darmstadt, Schnittspahnstraße 9, 64287 Darmstadt, Germany
ARTICLE INFO
Keywords:
Winter wheat
Climate impact
Heat impact
Drought impact
Compound effects
ABSTRACT
Whether hydroclimatic extremes cause yield losses or failures not only depends on their intensity but also on
local environmental conditions. These conditions shape the capacity to buffer climatic shocks and thus neces-
sitate a regionally specic impact assessment and adaptation planning. However, the degree to which different
environmental conditions affect climate impacts on yields and its spatiotemporal variability across Germany is
relatively unknown. In this study, we use a regression-based crop-climate modelling approach for 71 regions,
classied according to soil and climate characteristics and investigate region-specic vulnerabilities of winter
wheat yields to hydroclimatic extremes for the period 1991–2019. We account for the co-occurrence of tem-
perature and moisture impacts (i.e. compound effects) as well as for local soil-climate conditions. On average,
our models can explain approx. 67 % of past winter wheat yield variations. Despite the rather homogeneous
climate in Germany, the results reveal clear geographic differences across different soil-climate regions. While
the north-eastern regions show a clear dominance of drought impacts, southern regions show stress due to
moisture excess. Heat impacts can clearly be linked to the warm regions along the western part of the country.
Overall, compound dry-hot extremes pose the strongest and most widespread risk for winter wheat yields in
Germany, being responsible for approx. 38 % and in some regions for up to 50 % of past yield variations. Based
on the identied regional differences in hydroclimate susceptibility, we can dene four geographic risk clusters,
which exhibit vulnerability to climatic extremes such as summer droughts, winter droughts, summer heat waves,
and winter moisture excess. The identied risk clusters of heat and moisture stresses could inform regional-
specic adaptation planning.
1. Introduction
The summers of 2003 and 2018 are egregious examples during which
very high temperatures and simultaneously low precipitation rates led to
signicant yield losses in Germany and Europe (Beillouin et al., 2020;
Ciais et al., 2005;Webber et al., 2020). With increasing climate vari-
ability and an intensifying hydrological cycle, temperature and precip-
itation extremes already occur more frequently than in the past and are
predicted to become even more frequent and intense in the future.
Climate extremes may also occur more frequently as spatially and
temporally compound events (Lesk et al., 2022). This rise in extremes
will challenge future wheat production, necessitating the need for
appropriate adaptation strategies to enhance the stability of winter
wheat yields, one of Germany’s major staple crops. Yet, adaptation
measures have to be tailored not only to risks but also to regions, as local
environmental characteristics like soil conditions, topography and local
climate variability might inuence the capacity of a region to buffer and
react to climate extremes. While in years with pronounced temperature
peaks, like 2003, during which most regions experienced substantial
yield losses, some regions were still able to produce above-average
wheat yields (Fig. 1, grey bars). Hence, regions are affected differently
by weather events owing to their varying response to climatic stresses,
due to locally specic environmental conditions or particular manage-
ment strategies. As B¨
onecke et al. (2020) and Riedesel et al. (2024) point
* Corresponding author at: Imperial College London, South Kensington Campus, London SW7 2AZ, UK.
E-mail address: r.becker@imperial.ac.uk (R. Becker).
Contents lists available at ScienceDirect
Agricultural and Forest Meteorology
journal homepage: www.elsevier.com/locate/agrformet
https://doi.org/10.1016/j.agrformet.2024.110322
Received 11 March 2024; Received in revised form 18 November 2024; Accepted 20 November 2024
Agricultural and Forest Meteorology 361 (2025) 110322
0168-1923/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ).

out, a particular focus should be laid on regional soil characteristics, as
they inuence the capacity to buffer adverse climatic effects.
While risks for winter wheat cultivation in Germany and beyond,
along with adaptation options to climate change, have been discussed
before (Schmitt et al., 2022;Webber et al., 2020), the inuence of soil,
regional climate conditions and crop phenology have rarely been dis-
entangled. We therefore explicitly separate the impacts of individual
from compound extremes and consider regional-specic environmental
conditions in our study. We chose a regression-based statistical model-
ling approach and implicitly account for temperature and precipitation
extremes, as well as compounding effects. In addition, we account for
spatial variation of impacts. We base our analysis on soil-climate regions
(hereafter referred to as SCR) of Germany as dened by Roßberg et al.
(2007), accounting for heterogeneities in both climate and soil attri-
butes. This is an important difference to most other statistical
crop-climate impact assessments, which often base their analysis on
administrative regions, such as districts or counties, which are dened
largely independent of physical characteristics (Conradt et al., 2016;
Gornott and Wechsung, 2016;Lüttger and Feike, 2018;Mirschel et al.,
2014).
Therefore, the central objectives of this study are to (i) identify the
main climatic drivers of winter wheat yield variations in Germany,
dissected into soil-climate zones and seasonal components, (ii) study
differences in their regional impacts, and (iii) quantify the relative
contributions of single and compound effects to yield losses. Lastly, we
discuss the methodology presented in this paper and evaluate its
strengths and weaknesses.
2. Data and methods
2.1. Input data
Our regression-based crop-climate models are constructed based on
district-level yield data (NUTS3 level), averaged over each soil-climate
region (dependent variable) and agroclimatic indices, which present
variables representing heat, drought and compound stresses for winter
wheat cultivars (independent variables; Table 1). Yield data was ob-
tained from German federal and regional ofces for statistics. Further
information and links to the data sets can be found in SI, Table S.1. The
indices are derived from temperature, precipitation and solar radiation
data sets (EOBS; Copernicus Climate Change Service, 2020) as well as
from data on soil moisture decit, plant transpiration (GLEAM-Hybrid;
Koppa et al., 2022) and soil moisture contents (SALTO; Merz et al.,
2020). We dene our climatic indices based on soil-climate regions
(SCR; Roßberg et al., 2007), extracting the mean daily climate,
soil-water and plant-water data within each SCR. We then use the
aggregated data to calculate regional agroclimatic indices. Yield data as
well as climate and hydrological data are remapped to these regions.
Remapping is done by calculating the mean of all districts/pixels, which
fall within one SCR polygon. A district/pixel falls into a polygon, when
it’s centre lays within a specic soil-climate region boundary. We do not
assign weights based on the proportion of the pixel, which falls into a
polygon. All data sets cover the time period 1991–2019.
Fig. 1. Winter wheat yield anomalies in Germany between 1991 and 2019 (A) and regional differences in average yields (B). Grey bars in A encompass the 5th and
95th percentiles of winter wheat yield across all soil-climate regions. Relative anomalies are calculated based on detrended region-specic yield time series, sub-
tracting a quadratic trend, and dividing the detrended time series by the trend-dened expected value for each year and each soil-climate region.
Table 1
Data sets used in this study.
Variable Data set Source Description, temporal and
spatial resolution
Yield District-
level
yield
Federal and
regional ofces
for statistics
Annual district-level winter
wheat yield data (NUTS3
level)
Temperature,
precipitation,
solar radiation
EOBS Copernicus
Climate Change
Service 2020
The data provides
observation-based,
gridded, historical climate
data over Europe. In this
study we use EOBS data at a
spatial resolution of 0.25◦x
0.25◦, at daily temporal
resolution.
Soil moisture
decit (SMD)
and ET
GLEAM-
Hybrid
Koppa et al.
2022:Martens
et al., 2017
The data set provides ET
and SMD data at a spatial
resolution of 0.25◦x 0.25◦
and at daily temporal
resolution. GLEAM is a
process-based land-
atmosphere model, which
dynamically simulates
environmental state
variables such as SMD and
ET for historical periods.
Soil moisture SALTO Merz et al. 2020 The data set provides soil
moisture data at a spatial
resolution of 8 km x 8 km
and at daily temporal
resolution. Soil moisture is
given as a relative unit [%],
indicating the current soil
moisture status compared
to a maximum possible soil
moisture content at eld
capacity. Soil moisture is
dynamically simulated
using the process-based
hydrological SALTO model.
Soil-climate
regions
SCR Roßberg et al.
2007
Soil-climate regions are
delineated based on soil
quality characteristics (dt.
‘Bodengüte”) and
dominating temperature
and precipitation patterns.
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
2

2.2. Denition of agroclimatic indices for heat, drought and compound
stresses
To identify the main climatic drivers of winter wheat yield uctua-
tions in Germany, we dene a set of agroclimatic indices which repre-
sent heat, moisture and compound heat-moisture stresses as well as
growth factors for winter wheat. The selection is made based on
commonly chosen indices in statistical crop-climate modelling studies
(B¨
onecke et al., 2020;Laudien et al., 2020;Lobell et al., 2011;Roma-
novska et al., 2023) and on a commentary on the importance of com-
pound effects by (Lesk et al., 2022). The indices chosen for this study are
listed in Table 2 and their respective relevance for winter wheat growth
are outlined in the SI, Section 4. Here we also outline, why we chose to
use soil moisture decit (SMD) as well as soil moisture, even though they
are linked and might exhibit a similar stress for crop growth.
All indices are calculated based on daily data using the data sets
listed in Table 1, which we remap to match the spatial extend of the soil-
climate regions (SCR). Subsequently, we calculate each indicator at the
spatial resolution of SCRs. Each of the indicators is calculated twice:
once for the vegetative season (with a sufx of “_v”to the name) and
once for the reproductive season (sufx “_r”). Details of single indices
and their units can be found in the SI, Section 4.
2.3. Data pre-processing and variable selection procedure
All agroclimatic indices are calculated separately for the vegetative
and the reproductive phases of the growing cycle and separately for each
soil-climate region. For winter wheat in Germany, we dene the vege-
tative phase as the period from sowing (October) to heading (March)
and the reproductive phase from end of heading (April) to maturity/
harvest (July), which aligns with the mean observed phenological pe-
riods in Germany (DWD Climate Data Center, 2021). The vegetative
phase covers the sub-phases of germination (BBCH 0–9), emergence and
leaf development (BBCH 10–19), tillering (BBCH 20–29), stem elonga-
tion (BBCH 30–39) and booting (BBCH 40–49). The reproductive phase
covers the sub-phases of heading (BBCH 50–59), owering (BBCH
60–69), milk ripening (BBCH 71–89), and senescence (BBCH 90–99)
until harvest (see also B¨
onecke et al., 2020).
All indices are standardized subtracting their mean and dividing by
their standard deviation, to allow for comparing and jointly interpreting
individual agroclimatic impacts on winter wheat yields using variable
coefcients. Soil-climate region specic yield time series are detrended
to remove management-dependent long-term trends (SI Fig. S.2 and
section 5). In addition, they are log-transformed to ensure a Gaussian
normal distribution of our dependent variable. Subsequently, we apply a
two-step variable selection, starting with pre-selecting the most inu-
ential compound variables to reach an equal number of compound and
single variables for the second step, as otherwise the much larger
number of compounds would skew the selection procedure. We limit the
selection to the group size of heat and moisture variable groups,
comprising 18 variables. Next, the pre-selected multiplicative interac-
tion terms are merged with the explicitly calculated compound variables
VPD and ETact, as well as with the single heat and moisture variables to
total of maximum 58 variables and the selection is re-run. For both se-
lection steps we use LASSO regression, separately for each soil-climate
region, dening the most inuential variables as those with the high-
est R
2
contribution. Variables are pruned before the regression by (i)
removing variables with zero variances and (ii) removing collinear
variables (correlation coefcient r>0.7). We allow at most seven var-
iables per model, showing the highest correlation with yield variations,
to avoid model-overtting. Our approach thus closely follows Laudien
et al. (2020), but adds the separate lter step for compounds.
2.4. Model setup
To quantify the impacts of heat and drought stresses on winter wheat
yields in Germany in a spatially discrete manner, we use a separately
estimated time series model for each region (STSM: Gornott and
Wechsung, 2016), which can account for site-specic climatic impacts.
Compound stress variables (c) are included as explicitly calculated
aggregate variables (ET
act
and VPD) or as multiplicative interaction
terms between moisture and temperature variables (Table 1).
The multiple regression model is given by the function:
log(yieldit) =logβi0∑
J
j=1
βij logwijt +∑
K
k=1
βik loghikt +∑
L
l=1
βil logcilt +loguit (1)
2.5. Model validation
Based on the approach of Laudien et al. (2020) and Romanovska
et al. (2023), we conduct a commonly used leave-one-out cross valida-
tion (LOOCV –level 1) and a more stringent LOOCV–level 2 validation
(SI, section 6). The evaluation of the model performances is based on the
coefcient of determination (R
2
) between simulated and observed ab-
solute yields, for single SCRs as well as for entire Germany. To assess the
explanatory power of each predictor variable, we calculate separate
explained variances for each selected predictor, dividing the individual
sum of squared residuals of a single independent variable by the total
sum of squared residuals of all independent variables.
To obtain the annual share of explained yield variation by temper-
ature, moisture and compounding stresses, we weigh the normalized
indicators by their respective share of explained variation and get an
annual index of temperature, moisture and compound impacts as fol-
lows:
yield =yield time series as trend anomaly;
i=spatial subunit (here:soil −climate regions);
t=time step in years;
β0=intercept;
β=coefficient of independent variables w,h,and c;
j=number of all selected moisture stress related climate variables for model i;
k=number of all selected heat stress related climate variables for model i;
l=number of all selected combined moisture and heat stress climate variables for model i;
w=water stress variable (climatic indicator);
h=heat stress variable (climatic indicator);
c=compound stress variable (climatic indicator);
u=error term
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
3

Indexxt=∑
I
i=1
∑
X
x=1
V normxti ∗Explained variancexi
Total explained variancei
(2)
with :
x=moisture,temperature or compound variable;
t=time (year);
i=soil −climate region;
Vnorm =normalized variable
2.6. Denition of climate risk clusters
A climate risk cluster is dened as a region which reacts particularly
sensitive to a specic category of climatic stresses (i.e. temperature,
moisture or compound category). For example, in the case of particu-
larly dry conditions in summer, indicated by high importance of
moisture related agroclimatic indices during the reproductive phase, a
region would be classied as “too dry in summer”. To identify a risk
cluster, we select the most important variables for each SCR (explained
variance multiplied by the selection frequency of the respective variable;
SI section 7) and examine whether they show a clear dominance of a
particular variable group (i.e. moisture, heat, or compound group),
hinting towards the prevalence of a specic risk factor for yields (e.g.
risk of summer droughts). A region is dened as part of a risk cluster if
the total explained variance of all variables which contribute to the risk
factor is greater than 10 % (for 5 %- and 20 %-thresholds see SI, Fig. S.4).
3. Results and discussions
3.1. Model performances
Average total simulated wheat yields for Germany of the full (in-
sample) model show a good t with observed yield uctuations (R
2
=
0.81; Fig. S.3). Similarly, the two cross-validation levels (LOOCV-1 and
LOOCV-2) present clear correlations with R
2
-values of 0.70 and 0.55
respectively (Table 3). The fact that model performances between all
three simulations do not strongly vary hints towards a robust variable
selection, which is independent of the different selection methods
applied for LOOCV-1 and LOOCV-2. Model overtting can therefore be
ruled out, underlining that substantial parts of yield variance can be
explained by climate uctuations.
3.2. Impact of heat, moisture and compound impacts on winter wheat
yields
Our results show that temperature, moisture and compound effects
signicantly impact winter wheat yields in Germany. On average,
approx. 67 % of yield variability can be explained by climatic impacts in
each SCR (Table 3). At the same time, our results show that region-
specic soil-climate conditions matter strongly in dening the strength
and the cause of yield losses. Disentangling the impacts of moisture,
heat, and compound heat-moisture impacts reveals that the individual
regions are differently responsive (Fig. 2).
3.2.1. Yield losses explained by moisture impacts
Moisture impacts were found to affect yield variations particularly in
the warm and dry north-east and central-north of the country, but also in
the wetter southern regions, such as in the Black Forest and in the re-
gions of the Danube and Inn river valleys (Fig. 2A; a map with the
geographic regions highlighted in this publication can be found in
Fig. S1, in the supplementary material). On average, approx. 16 % of
yield uctuations can be ascribed to moisture impacts alone, which
highlights the importance of optimal moisture condition for stable
winter wheat yields. Similarly, the interannual variations of moisture
anomalies (Fig. 2D) reveal the importance of moisture impacts on yields
in Germany and show that years with high yield losses (1992, 2003, and
2018) show a particularly high contribution of moisture stress. Year
2007, which likewise shows high yield losses, experienced on average
low moisture stress (very dry April but wet remaining reproductive
season), indicating that yield losses in this year have to be attributed to
climatic stresses which cannot be captured by the selected indicators or
to non-climatic impacts, e.g. plant viruses such as the yellow dwarf virus
which severely affected the 2007-yields (Bayrische Landesanstalt für
Landwirtschaft, n.d.). The most important moisture stress variables are
soil moisture decit during the spring and summer seasons (SmdI_sum_r),
long dry spells during winter (PrecipD_b01_v) and long dry spells during
spring and summer (PrecipD_b01_r; Fig. 3).
Our regional assessment shows, that soil moisture decits during the
reproductive season (SmdI_sum_r), was found to be the strongest cause
for yield failures in central-north and north-east Germany, explaining on
average 16 % of past yield variations (Fig. S.4A). In single soil-climate
regions, such as the inland areas of the north-eastern lowlands, this
Table 3
Average model performances of all 71 soil-climate region models.
Performance criteria Training (in-
sample)
Validation LOOCV-1
(out-of-sample)
Validation
LOOCV-2
(out-of-
sample)
Mean R
2
(single soil-
climate regions)
0.67 0.44 0.29
Mean R
2
(entire
Germany)
0.81 0.70 0.55
Table 2
Agroclimatic indices.
Agroclimatic Index Long name and denition
Temperature 
Tmax Median of maximum daily temperatures
Tmin Median of minimum daily temperatures
GDD Growing Degree Days –sum of days with optimal growing
conditions
EDD_I27 Intensity of hot days –sum of Extreme Degree Days with
Tmax >27 ◦C
EDD_F27 Frequency of hot days –number of Extreme Degree Days
with Tmax >27 ◦C
EDD_D27 Duration of hot days –maximum consecutive Extreme
Degree Days with Tmax >27 ◦C
EDD_I31 Intensity of very hot days –sum of Extreme Degree Days
with Tmax >31 ◦C
EDD_F31 Frequency of very hot days –number of days with Tmax >
31 ◦C
EDD_D31 Duration of very hot days –maximum consecutive Extreme
Degree with Tmax >31 ◦C
Moisture 
PrecipI_sum Intensity of precipitation –sum of precipitation
PrecipF_b01 Frequency of dry periods –number of days with
precipitation <0.1 mm
PrecipD_b01 Duration of dry periods –maximum consecutive days with
precipitation <0.1 mm
SmdI_sum Intensity of soil moisture decit (SMD) –sum of daily SMD
[min. stress =1, max. stress =0]
SmdF_b1 Frequency of soil moisture decit –number of days with
SMD <1
SmdD_b1 Duration of soil moisture decit –maximum consecutive
days with SMD <1
SmI_sum Intensity of soil moisture availability –sum of soil moisture
SmF_b06 Frequency of low soil moisture contents –number of days
with soil moisture <60 % at eld capacity
SmD_b06 Duration of low soil moisture contents –maximum
consecutive days with soil moisture <60 %
Compound 
ETact Actual plant evapotranspiration –sum of ETact (crop-
physiological interaction)
VPD Vapor pressure decit –mean of VPD (heat-moisture
interaction)
[Moisture]*
[Temperature]
Moisture variable*temperature variable (crop-atmosphere
interaction); this comprises 324 variables in total (18
temperature * 18 moisture variables)
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
4

variable explains up to 47 % of yield variations. The north-eastern re-
gions present drought-prone areas of sandy and diluvial soils with low
water holding capacities (Kahlenborn et al., 2021;Roßberg et al., 2007)
and cover areas of lowest winter wheat yield potential (Fig. 1). Our
results underline their high vulnerability towards climatic impacts and
their low potential to buffer adverse effects of dry periods during spring
and summer months.
The north-west of the country is comparably less affected by soil
moisture decits, but also shows a clear negative impact of summer
droughts on yield. Here, prolonged durations of dry spring and summer
periods (PrecipD_b01_r) can explain on average 7 % of yield variations
(Fig. S.4F). Similar to the regional distribution of plant water decit
impacts, the highest impacts can be found in regions with sandy soils (e.
g. sandy north-western lowlands).
Our results align with ndings of Schmitt et al. (2022), who found
that summer droughts are the main driver of wheat yield failures in
Germany. This is due to winter wheat production in Germany being
nearly entirely rain-fed and hence highly dependent on precipitation as
its only moisture source (Webber et al., 2018). The importance of
moisture availability during the reproductive (spring/summer) season
can be explained by the need for optimal moisture conditions particu-
larly during the months around anthesis. A lack of moisture availability
during this critical phenological phase can drastically reduce the plant’s
photosynthetic activity and negatively inuence yields (Reynolds et al.,
2022).
In the central part of the country, yields are found to react more
sensitive to long durations of dry periods during the winter season
(PrecipD_b01_v; Fig. S.4C). In some regions up to 29 % of yield variation
can be explained by this predictor alone. These are primarily regions,
characterized by fertile loess soils with high water holding capacities
and favourable conditions for farming (Roßberg et al., 2007). Suf-
ciently wet winter periods seem crucial for lling soil moisture storages
in these regions, from which the winter crop can benet during dry
spring and summer months.
A low frequency of dry spells between October to March (Pre-
cipF_b01_v) is found to be the main cause for wheat yield failures in the
south of Germany (Fig. 4C), explaining on average 10 % and in some
regions up to 29 % of yield variability (Fig. S.5I). Recurring wet spells
during the vegetative period in these regions, particularly during the
sowing phase, have delayed sowing in the past (Webber et al., 2020),
which increases the risk of frost damages during the early growth stage
and therefore the risk for yield failures. Based on these results, we can
distinguish three moisture risk clusters, which show the need to adapt
winter wheat cultivation to i. spring and summer droughts (Fig. 4A); ii.
winter drought (Fig. 4B), and iii. periods of moisture excess during
winter (Fig. 4C).
Fig. 2. Regional specic impacts of moisture (A), heat (B) and compound effects (C) and the average annual share of each impact group (i.e. moisture, heat or
compound effects; left y axis) on winter wheat yields (right y axis) between 1991 and 2019 (D). Positive impact anomalies describe below average stresses, and
negative anomalies describe above average stress. Grey regions in A, B and C represent regions where no moisture (A), heat (B), or compound impact (C) are found.
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
5

3.2.2. Yield losses explained by heat impacts
Heat related effects show stronger explanatory power in models in
the west of Germany, particularly along the heat-prone Rhine River
valley, in the low mountain ranges in the central-west and in the lowland
areas in the north-west of Germany (Fig. 2B). On average, heat impacts
explain 9 % of yield variations. In the north-east of the country, they
show limited impacts on yield level uctuations. Notably, the main
temperature-related impact affecting wheat yields independently of co-
occurring moisture decits is the minimum temperature during the
reproductive season (Tmin_r); this is the second most selected predictor
variable (Fig. 3). High minimum temperatures can explain on average 8
%, and in single regions up to 25 % of yield variations. The negative
impact of high minimum temperatures (i.e. night-time temperatures)
has been identied by a series of studies as an important factor for
causing yield failures (García et al., 2015,2016;Hein et al., 2019;Sadok
and Jagadish, 2020). High night-time temperatures, particularly during
the seed-lling period, as captured by our Tmin_r variable, could clearly
be linked to declines in winter wheat crops (Hein et al., 2019). The main
physiological cause behind these yield losses is an insufcient carbon
supply for plant and seed growth due to enhanced night-time respiration
and possibly a faster leave senescence (Sadok and Jagadish, 2020).
Considering that global night-time temperatures are rising faster than
day-time temperatures (Sillmann et al., 2013) this phenomenon, which
already shows a measurable impact, might pose an increasing risk for
winter wheat yields in Germany. However, heat impacts vary signi-
cantly from region to region, as well as from year to year (Fig. 2D),
which underlines the importance of region-specic adaptation strate-
gies. Based on our results we dene one heat risk cluster, which covers
primarily the regions in the centre and west of the country, encom-
passing favourable arable lands of the central agricultural regions and
central low mountain ranges (Fig. 4D).
3.2.3. Yield losses explained by compound impacts
Our results reveal that compounding effects pose the strongest risk
for winter wheat yields, explaining on average approx. 38 % and in some
regions up to 50 % of past yield variations. The results are consistent
with recent ndings of compound effects posing a particular risk for crop
yields (Heino et al., 2023;Lesk et al., 2022). Even though overall large
knowledge gaps still exist regarding the interplay of heat and water
stress and its effect on plant growth (Siebert, Webber, and Rezaei, 2017,
2017;Zampieri et al., 2018), recent studies suggest that co-occurring
heat and drought events results in synergistic effects and have a stron-
ger negative impact than single stressors alone (Rezaei et al., 2023). For
Germany, this is particularly evident with regard to heat stress. As
outlined above, high day-time temperatures alone rarely cause signi-
cant yield losses. But the co-occurrence of high day-time temperatures
and low precipitation rates during the reproductive phase (Precip_-
sum_r*Tmax_r and Precip_sum_r*EDD_I27_r) are amongst the most deci-
sive factors controlling wheat yields (Fig. 3). The impacts of compound
effects on the plant physiology are complex. A common effect is stomata
closure as a response to a lack in moisture availability, entailing less
transpiration and thus a reduced cooling of plant tissue. This eventually
increases leaf temperatures, particularly when ambient heat prevails,
which can cause signicant damage to the crop (Abdelhakim et al.,
2021).
The interannual variation of compound effects between 1991 and
2019 (Fig. 2D) shows that co-occurring heat and moisture stresses vary
signicantly between years. Our results furthermore reveal that in those
regions which show pronounced vulnerability towards compound ef-
fects, single heat and moisture stress effects are negligible or non-
existent (Figs. 2A–C and S.3). Unlike the separate moisture and tem-
perature impacts, single compound impacts cannot clearly be attributed
to specic soil-climate regions. This might be due to their complex
combinations, which lead to the selection of highly specic predictor
variables for each region. Areas which are strongly affected by co-
occurring heat and moisture stresses are the Alpine foothills and adja-
cent northern areas, as well as productive agricultural lands in loess
regions in the south-west and centre-east of Germany (Fig 2C). Based on
our results, we cannot detect a clear risk cluster for compounding im-
pacts, but Fig. 4E highlights that their impacts have to be considered in
vast parts of the country.
3.3. Strengths and weaknesses of the proposed methodology to assess
climate impacts on winter wheat yields
Our approach of using a spatially distributed linear regression model
shows promising results in identifying main climatic drivers of winter
wheat yield variations and in ascribing these drivers to soil-climate
Fig. 3. Selection frequency, importance (R
2
) and impact on yield (negative/positive) of the eight most important predictor variables. The weighted importance (=
selection frequency*R
2
) is ranked from top to bottom. There are three levels of information: rst, the direction indicates if the respective variable shows a negative
(dark blue) or positive (light blue) correlation with yield. Second, the extent on the X-axis represents the frequency by which the variables were selected (possible
maximum =71; total number of models). Third, the R
2
values in each box indicate the average explained variance of each variable. See Fig. S.5 for full list.
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
6

regions as well as to seasonal components. We explicitly account for
compounding effects, which are often missing in statistical crop-climate
models and only recently have been integrated (B¨
onecke et al., 2020).
Our results show that compound impacts are the dominating stress
factors which control yield levels and should consequently be integrated
in climate impact studies. We present a rst step towards an interpret-
able modelling framework which allows for the inclusion of spatially
and seasonally differentiated compounding effects.
The approach furthermore allows for a separation of heat, moisture
and compound stresses on winter wheat at SCR level. This could be of
interest for the selection of stress- and region-dependent adaptation
strategies. Grouping multiple impacts of one stress-category into larger
risk clusters (Fig. 4) can give further information on region- and season-
specic adaptation requirements. Each risk cluster demands a different
set of adaptation measures to optimally cope with the region-specic
adverse climatic impacts on yields. In addition, the approach enables
us to quantify the relative contributions of single and compound effects
to yield losses and assess their interannual differences (Fig. 2). This can
be of importance when planning and prioritizing adaptation strategies
for the most frequent and most important drivers of yield losses.
However, despite these strengths, our simple mathematical formu-
lation (e.g. multiplicative compounds and linear unidirectional re-
lationships between seasonal yield and seasonal climatic impacts) has
limitations. Due to its mathematical form as well as the coarse split of
the growing season into only two phenological phases, our approach
may fall short of the physiological realities of weather impacts, which
may be non-linear, bidirectional, seasonally divergent and temporally
staggered. An alternative approach to address these limitations could be,
for example, the application of machine learning based approaches with
a temporal resolution based on most important phenological phases.
Another point to consider is that the nature of our regression-based
approach may gloss over minor climatic impacts when there is a
clearly dominating impact on yield dynamics. For example, the sandy
regions in the north-east of the country stand out as regions where
summer dryness affects wheat yield the strongest (Fig. 4A). This does not
mean that other stresses (e.g. winter dryness) are absent in this region.
Compared to the dominating effect they are however statistically less
important and are therefore ‘neglected’. Comparing the results with
those from process-based crop models (e.g. Nendel et al., 2023;Webber
et al., 2020) with a ner temporal stress-disaggregation, could further
assist in increasing the robustness of the results. A ner separation of
phenological phases, as done by B¨
onecke et al. (2020) and Riedesel et al.
(2024), would help to better identify critical phenological phases, which
are particularly sensitive to adverse climate impacts. Considering only
the vegetative and the reproductive season, as done in this study, does
not resolve impacts at important phenological steps, e.g. temperature
Fig. 4. Risk cluster of a particular climatic stress showing regions of drought risks in summer (A), regions of drought risks in winter (B), regions of water excess in
winter (C), regions of heat risks in summer (D), and regions where compound effects impact wheat yields (E).
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
7

stress around anthesis. However, spatially distributed data availability
of temporal and spatial highly variable starting and ending dates of
single phases is limited. Model stability is another trade-off factor, as
more seasonal indices along with their interactions increase the number
of explanatory variables.
4. Conclusion
In this study, we were able to detect the main climatic drivers of
winter wheat yield anomalies in Germany based on a spatially distrib-
uted multiple regression model approach which accounts for regionally
distinct soil-climate characteristics. The integration of compound im-
pacts, as multiplicative interaction terms, allowed us to consider co-
occurring effects of heat and moisture extremes. Our results highlight
that these compounding effects are the dominating factor for causing
yield losses and are hence crucial to consider in modelling frameworks
which analyse climate impacts on yields. We could show that our
approach offers a consistent framework to assess multiple stressors and
their numerous interactions, which does not rely on extensive eld and
laboratory data collection. The assessment presented here can be done
using publicly available data sets only. This could be of particular in-
terest in regions where farm-level data is unavailable and a nationalwide
analysis is desired which accounts for difference in regional soil-climate
conditions.
Finally, using our modelling framework, we were able to identify
stress-specic risk clusters of adverse climate impacts for winter wheat
in Germany. Bearing in mind that adaptation measures depend primarily
on a farm-specic cost-benet analyses, skills, technology and local
environmental and economic conditions, such region-specic results can
be an inception for basing adaptation incentives on region- and season
specic risks.
Funding
This research did not receive any specic grant from funding
agencies in the public, commercial, or not-for-prot sectors.
CRediT authorship contribution statement
Rike Becker: Writing –original draft, Visualization, Validation,
Methodology, Investigation, Formal analysis, Data curation, Conceptu-
alization. Bernhard Schauberger: Writing –review &editing, Super-
vision, Methodology. Ralf Merz: Writing –review &editing,
Methodology. Stephan Schulz: Writing –review &editing, Methodol-
ogy. Christoph Gornott: Writing –review &editing, Supervision,
Methodology, Conceptualization.
Declaration of competing interest
The authors declare that they have no conict of interest.
Acknowledgements
We thank Dietmar Roßberg, Volker Michel, Rudolf Graf und Ralf
Neukampf (Julius-Kühn Institute) for the data set of Germany’s soil-
climate regions. We furthermore thank Paula Romanowska and Rahel
Laudien (Potsdam Institute for Climate Impact Research) for their kind
and helpful advice regarding the model set-up and model validation. We
also thank Tini Plitt (University of Kassel) for sharing her agricultural
expertise. Finally, we thank two anonymous reviewers for their time to
review our manuscript and for giving us valuable recommendations and
feedback on how to further improve it.
Supplementary materials
Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.agrformet.2024.110322.
Data availability
I have shared the link to my data and further data can be made
available on request.
References
Abdelhakim, L.O.A., Palma, C.F.F., Zhou, R., Wollenweber, B., Ottosen, C.-O.,
Rosenqvist, E., 2021. The effect of individual and combined drought and heat stress
under elevated CO2 on physiological responses in spring wheat genotypes. Plant
Physiol. Biochem. 162, 301–314. https://doi.org/10.1016/j.plaphy.2021.02.015.
Bayrische Landesanstalt für Landwirtschaft. (n.d.). Weizenverzwergungsvirus (Wheat dwarf
monogeminivirus, WDV). Retrieved 25 September 2024, from https://www.l.
bayern.de/ips/getreide/018686/index.php.
Beillouin, D., Schauberger, B., Bastos, A., Ciais, P., Makowski, D., 2020. Impact of
extreme weather conditions on European crop production in 2018. Philosoph.
Transact. Roy. Soc. B: Biolog. Sci. 375 (1810), 20190510. https://doi.org/10.1098/
rstb.2019.0510.
B¨
onecke, E., Breitsameter, L., Brüggemann, N., Chen, T., Feike, T., Kage, H.,
Kersebaum, K., Piepho, H., Stützel, H., 2020. Decoupling of impact factors reveals
the response of German winter wheat yields to climatic changes. Glob. Chang. Biol.
26 (6), 3601–3626. https://doi.org/10.1111/gcb.15073.
Ciais, Ph., Reichstein, M., Viovy, N., Granier, A., Og´
ee, J., Allard, V., Aubinet, M.,
Buchmann, N., Bernhofer, Chr., Carrara, A., Chevallier, F., De Noblet, N., Friend, A.
D., Friedlingstein, P., Grünwald, T., Heinesch, B., Keronen, P., Knohl, A., Krinner, G.,
Valentini, R., 2005. Europe-wide reduction in primary productivity caused by the
heat and drought in 2003. Nature 437 (7058), 529–533. https://doi.org/10.1038/
nature03972.
Conradt, T., Gornott, C., Wechsung, F., 2016. Extending and improving regionalized
winter wheat and silage maize yield regression models for Germany: enhancing the
predictive skill by panel denition through cluster analysis. Agric. For. Meteorol.
216, 68–81. https://doi.org/10.1016/j.agrformet.2015.10.003.
Copernicus Climate Change Service. (2020). E-OBS daily gridded meteorological data for
Europe from 1950 to present derived from in-situ observations. https://cds.climate.co
pernicus.eu/cdsapp#!/dataset/insitu-gridded-observations-europe?tab=o
verview</Dataset>.
DWD Climate Data Center. (2021). Eintrittsdaten verschiedener Entwicklungsstadien
landwirtschaftlicher Kulturpanzen von der Bestellung bis zur Ernte (Jahresmelder,
historisch), Version v007 (DWD (urn:x-wmo:md:de.dwd.cdc::obsgermany-phenology-
annual_reporters-crops-historical)) . </Dataset>.
García, G.A., Dreccer, M.F., Miralles, D.J., Serrago, R.A., 2015. High night temperatures
during grain number determination reduce wheat and barley grain yield: a eld
study. Glob. Chang. Biol. 21 (11), 4153–4164. https://doi.org/10.1111/gcb.13009.
García, G.A., Serrago, R.A., Dreccer, M.F., Miralles, D.J., 2016. Post-anthesis warm nights
reduce grain weight in eld-grown wheat and barley. Field Crops Res. 195, 50–59.
https://doi.org/10.1016/j.fcr.2016.06.002.
Gornott, C., Wechsung, F., 2016. Statistical regression models for assessing climate
impacts on crop yields: a validation study for winter wheat and silage maize in
Germany. Agric. For. Meteorol. 217, 89–100. https://doi.org/10.1016/j.
agrformet.2015.10.005.
Hein, N.T., Wagner, D., Bheemanahalli, R., ˇ
Sebela, D., Bustamante, C., Chiluwal, A.,
Neilsen, M.L., Jagadish, S.V.K, 2019. Integrating eld-based heat tents and cyber-
physical system technology to phenotype high night-time temperature impact on
winter wheat. Plant Method. 15 (1), 41. https://doi.org/10.1186/s13007-019-0424-
x.
Heino, M., Kinnunen, P., Anderson, W., Ray, D.K., Puma, M.J., Varis, O., Siebert, S.,
Kummu, M., 2023. Increased probability of hot and dry weather extremes during the
growing season threatens global crop yields. Sci. Rep. 13 (1), 3583. https://doi.org/
10.1038/s41598-023-29378-2.
Kahlenborn, W., Porst, L., Voß, M., Fritsch, U., Renner, K., Zebisch, M., Wolf, M.,
Sch¨
onthaler, K., &Schauser, I. (2021). Klimawirkungs- und Risikoanalyse 2021 für
Deutschland—Kurzfassung. Umweltbundesamt - Climate Change 26/2021.
Koppa, A., Rains, D., Hulsman, P., Poyatos, R., Miralles, D.G., 2022. A deep learning-
based hybrid model of global terrestrial evaporation. Nat. Commun. 13 (1), 1912.
https://doi.org/10.1038/s41467-022-29543-7.
Laudien, R., Schauberger, B., Makowski, D., Gornott, C., 2020. Robustly forecasting
maize yields in Tanzania based on climatic predictors. Sci. Rep. 10 (1), 19650.
https://doi.org/10.1038/s41598-020-76315-8.
Lesk, C., Anderson, W., Rigden, A., Coast, O., J¨
agermeyr, J., McDermid, S., Davis, K.F.,
Konar, M., 2022. Compound heat and moisture extreme impacts on global crop
yields under climate change. Nat. Rev. Earth Environ. 3 (12), 872–889. https://doi.
org/10.1038/s43017-022-00368-8.
Lobell, D.B., B¨
anziger, M., Magorokosho, C., Vivek, B., 2011. Nonlinear heat effects on
African maize as evidenced by historical yield trials. Nat. Clim. Chang. 1 (1), 42–45.
https://doi.org/10.1038/nclimate1043.
Lüttger, A.B., Feike, T., 2018. Development of heat and drought related extreme weather
events and their effect on winter wheat yields in Germany. Theor. Appl. Climatol.
132 (1), 15–29. https://doi.org/10.1007/s00704-017-2076-y.
Martens, B., Miralles, D.G., Lievens, H., van der Schalie, R., de Jeu, R.A.M., Fern´
andez-
Prieto, D., Beck, H.E., Dorigo, W.A., Verhoest, N.E.C., 2017. GLEAM v3: satellite-
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
8

based land evaporation and root-zone soil moisture. Geoscientif. Model Develop. 10
(5), 1903–1925. https://doi.org/10.5194/gmd-10-1903-2017.
Merz, R., Tarasova, L., Basso, S., 2020. Parameter’s controls of distributed catchment
models—how much information is in conventional catchment descriptors? Water
Resour Res 56 (2). https://doi.org/10.1029/2019WR026008 e2019WR026008.
Mirschel, W., Wieland, R., Wenkel, K.-O., Nendel, C., Guddat, C., 2014. YIELDSTAT –a
spatial yield model for agricultural crops. Eur. J. Agron. 52, 33–46. https://doi.org/
10.1016/j.eja.2013.09.015.
Nendel, C., Reckling, M., Debaeke, P., Schulz, S., Berg-Mohnicke, M., Constantin, J.,
Fronzek, S., Hoffmann, M., Jakˇ
si´
c, S., Kersebaum, K.-C., Klimek-Kopyra, A.,
Raynal, H., Schoving, C., Stella, T., Battisti, R., 2023. Future area expansion
outweighs increasing drought risk for soybean in Europe. Glob. Chang. Biol. 29 (5),
1340–1358. https://doi.org/10.1111/gcb.16562.
Reynolds, M.P., Slafer, G.A., Foulkes, J.M., Grifths, S., Murchie, E.H., Carmo-Silva, E.,
Asseng, S., Chapman, S.C., Sawkins, M., Gwyn, J., Flavell, R.B., 2022. A wiring
diagram to integrate physiological traits of wheat yield potential. Nat. Food 3 (5),
318–324. https://doi.org/10.1038/s43016-022-00512-z.
Rezaei, E.E., Webber, H., Asseng, S., Boote, K., Durand, J.L., Ewert, F., Martre, P.,
MacCarthy, D.S., 2023. Climate change impacts on crop yields. Nat. Rev. Earth
Environ. 4 (12), 831–846. https://doi.org/10.1038/s43017-023-00491-0.
Riedesel, L., M¨
oller, M., Piepho, H.-P., Rentel, D., Lichthardt, C., Golla, B., Kautz, T.,
Feike, T., 2024. Site conditions determine heat and drought induced yield losses in
wheat and rye in Germany. Environ. Res. Lett. 19 (3), 034024. https://doi.org/
10.1088/1748-9326/ad24d0.
Romanovska, P., Schauberger, B., Gornott, C., 2023. Wheat yields in Kazakhstan can
successfully be forecasted using a statistical crop model. Eur. J. Agron. 147, 126843.
https://doi.org/10.1016/j.eja.2023.126843.
Roßberg, D., Michel, V., Graf, R., Neukampf, R., 2007. Denition von Boden-Klima-
R¨
aumen für die Bundesrepublik Deutschland. Heft 7 59 (7), 155–161.
Sadok, W., Jagadish, S.V.K., 2020. The hidden costs of nighttime warming on yields.
Trend. Plant Sci. 25 (7), 644–651. https://doi.org/10.1016/j.tplants.2020.02.003.
Schmitt, J., Offermann, F., S¨
oder, M., Frühauf, C., Finger, R., 2022. Extreme weather
events cause signicant crop yield losses at the farm level in German agriculture.
Food Policy 112, 102359. https://doi.org/10.1016/j.foodpol.2022.102359.
Siebert, S., Webber, H., Rezaei, E.E., 2017a. Weather impacts on crop yields—Searching
for simple answers to a complex problem. Environ. Res. Lett. 12 (8), 081001.
https://doi.org/10.1088/1748-9326/aa7f15.
Siebert, S., Webber, H., Zhao, G., Ewert, F., 2017b. Heat stress is overestimated in
climate impact studies for irrigated agriculture. Environ. Res. Lett. 12 (5), 054023.
https://doi.org/10.1088/1748-9326/aa702f.
Sillmann, J., Kharin, V.V., Zhang, X., Zwiers, F.W., Bronaugh, D., 2013. Climate extremes
indices in the CMIP5 multimodel ensemble: part 1. Model evaluation in the present
climate. J. Geophys. Res.: Atmosph. 118 (4), 1716–1733. https://doi.org/10.1002/
jgrd.50203.
Webber, H., Ewert, F., Olesen, J.E., Müller, C., Fronzek, S., Ruane, A.C., Bourgault, M.,
Martre, P., Ababaei, B., Bindi, M., Ferrise, R., Finger, R., Fodor, N., Gabald´
on-
Leal, C., Gaiser, T., Jabloun, M., Kersebaum, K.-C., Lizaso, J.I., Lorite, I.J.,
Wallach, D., 2018. Diverging importance of drought stress for maize and winter
wheat in Europe. Nat. Commun. 9 (1), 4249. https://doi.org/10.1038/s41467-018-
06525-2.
Webber, H., Lischeid, G., Sommer, M., Finger, R., Nendel, C., Gaiser, T., Ewert, F., 2020.
No perfect storm for crop yield failure in Germany. Environ. Res. Lett. 15 (10),
104012. https://doi.org/10.1088/1748-9326/aba2a4.
Zampieri, M., Ceglar, A., Dentener, F., Toreti, A., 2018. Understanding and reproducing
regional diversity of climate impacts on wheat yields: current approaches, challenges
and data driven limitations. Environ. Res. Lett. 13 (2), 021001. https://doi.org/
10.1088/1748-9326/aaa00d.
R. Becker et al. Agricultural and Forest Meteorology 361 (2025) 110322
9