Projecting climate change impacts on grain maize based on three different crop model approachesby A. Holzkämper, P. Calanca, M. Honti, J. Fuhrer

Agricultural and Forest Meteorology


Agronomy and Crop Science / Forestry / Atmospheric Science / Global and Planetary Change


Agricultural and Forest Meteorology 214–215 (2015) 219–230

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology j our na l ho me page: www.elsev ier .com/ locate /agr formet

Project aiz three d

A. Holzkä a Agroscope, In CH-80 b MTA-BME Wa pest, H a r t i c l

Article history:

Received 7 De

Received in re

Accepted 21 A


Climate chang


Grain maize

Adaptation pla


Crop modellin aptat clim d up e – th exp , dow ated hains bigu t clim , mor crop-specific adaptation planning. © 2015 Elsevier B.V. All rights reserved. 1. Introdu

The asse studies has mates of im productivit tation plan estimates b ties as poss et al., 2013) tainty helps and to deriv

A comm mates is to u with differe

Using mult mon proced (e.g. Masut 2011; Islam

Höglind et addition, im ∗ Correspon

E-mail add http://dx.doi.o 0168-1923/© ction ssment of uncertainties in agricultural climate impact gained increasing attention in recent years. Since estipacts of projected climate changes on agricultural y are intended to support decision-making in adapning, it is desirable to increase confidence in these y providing as much information on known uncertainible (Challinor, 2009; Challinor et al., 2013; Vermeulen . Quantifying and separating different sources of uncerto improve understanding of uncertainties in impacts e decision-relevant information (Challinor et al., 2013). on approach to quantify uncertainties in impact estise ensembles, where impacts are estimated repeatedly nt models or different realizations of uncertain inputs. iple climate projections as inputs has become a comure for quantifying uncertainties of estimated impacts omi et al., 2009; Garrido et al., 2011; Zhang et al., et al., 2012; Osborne et al., 2013; Graux et al., 2013; al., 2013; Deryng et al., 2014; Fuhrer et al., 2014). In pact model uncertainties are increasingly considered ding author. Tel.: +41 58 468 75 16. ress: (A. Holzkämper). through ensembles including multiple crop model parameterizations and/or multiple impact models (e.g. Aggarwal and Mall, 2002;

Challinor et al., 2005; Tao et al., 2009; Semenov and Stratonovitch, 2010; Ceglar and Kajfez-Bogataj, 2012; Tao and Zhang, 2013;

Asseng et al., 2013). Recent ensemble studies showed that variation among crop models can have large effects on uncertainty in estimated climate change impacts (Asseng et al., 2013; Bassu et al., 2014). In fact, variation among crop models (i.e. structural impact model uncertainty) can contribute even more to uncertainty than variation among downscaled GCMs (Asseng et al., 2013).

Such structural impact model uncertainty has only been quantified so far in mechanistic crop model ensembles where it is represented by variation in the description of functional relationships and parameter values. As an alternative approach, statistical crop models have been used to assess impacts of climate change on crop yields (e.g. Lobell et al., 2006; Iglesias et al., 2010; Schlenker and Lobell, 2010). Few ensembles have been run with statistical crop models using replicates of statistical models based on bootstrap resampling (e.g. Lobell et al., 2006; Tebaldi and Lobell, 2008).

However, agro-climate ensembles involving different crop modelling approaches have never been applied so far.

In this study, for the first time three fundamentally different modelling approaches are applied in an impact assessment ensemble: a statistical crop model, a process-based crop model and a recently developed hybrid approach for estimating climate rg/10.1016/j.agrformet.2015.08.263 2015 Elsevier B.V. All rights climate change impacts on grain m ifferent crop model approaches mpera,∗, P. Calancaa, M. Hontib, J. Fuhrera stitute for Sustainability Sciences, Climate and Air Pollution Group, Reckenholzstr. 191, ter Research Group, Hungarian Academy of Sciences, H-1111 Muegyetem rkp. 3, Buda e i n f o cember 2014 vised form 20 August 2015 ugust 2015 e impacts nning g a b s t r a c t

Decision making in climate change ad standing of uncertainties in projected change on potential grain maize yiel

Switzerland using – for the first tim a process-based, a statistical and an originating from climate model chains eterization. We find that while estim originating from both climate model c specific climate limitations are less am related to the aggregation of differen mated changes in climate limitationse based on 46 Zurich, Switzerland ungary ion planning depends on the quantification and broad underate impacts. In a case study, we estimated impacts of climate to the time horizon 2036–2065 for three climatic regions in ree fundamentally different impact modelling approaches: ert-based approach. The aim was to quantify uncertainties nscaling weather generator choice, and impact model paramclimate impacts on yields are subject to large uncertainties and impact model approaches, estimates of changes in cropous. We conclude that by subtracting the layer of uncertainty ate influences on yield estimates and by focusing on estie decision-relevant information can be provided to support 220 A. Holzkämper et al. / Agricultural and Forest Meteorology 214–215 (2015) 219–230 suitability of grain maize based on expert knowledge and observational data (Holzkämper et al., 2013). Climate suitability approaches were traditionally applied for regional assessments of preferential cultivation zones, but are also increasingly being applied for

Daccache et 2013; Rami refine the cl climate sui mates, whi both the sta

The thre ity. The pro on a daily t

The statisti ships betw yield predi (Holzkämp logical dev estimates b different ph predictors a from literat within preelling appro assumption mechanisti tion of pre or pre-defin ity approac empirical e specific par