@article{fdi:010082066,
  title = {{T}o bias correct or not to bias correct ? {A}n agricultural impact modelers' perspective on regional climate model data},
  author = {{L}aux, {P}. and {R}otter, {R}. {P}. and {W}ebber, {H}. and {D}ieng, {D}. and {R}ahimi, {J}. and {W}ei, {J}. {H}. and {F}aye, {B}abacar and {S}rivastava, {A}. {K}. and {B}liefernicht, {J}. and {A}deyeri, {O}. and {A}rnault, {J}. and {K}unstmann, {H}.},
  editor = {},
  language = {{ENG}},
  abstract = {{M}any open questions and unresolved issues surround the topic of bias correction ({BC}) in climate change impact studies ({CCIS}). {O}ne question relates to the contribution of downscaling of climate change scenarios on the uncertainties in results obtained using impact models for agriculture. {I}n particular, for large area or regional agricultural impact assessments, the question of bias correction is of high relevance. {R}elatively few studies exist looking at the quantification of the impacts of {BC} methods in general circulation model ({GCM}) and regional climate model ({RCM}) data on results of such impact studies. {H}ere, we quantify the impact of different {BC} methods on temperature ({T}) and precipitation ({P}) from different {CORDEX} {GCM}-{RCM} combinations, and how the debiased {T}&{P} signal may propagate through agricultural impact models. {S}pecifically, we estimate the impact of {BC} on (i) an empirical {P}- and fuzzy rule-based algorithm to estimate the crop planting date, and (ii) a mechanistic {T}&{P}-based approach to quantify the crop suitability index ({CSI}) for the main staple crops in {W}est {A}frica (i.e. groundnut, maize, pearl millet, sorghum). {B}oth approaches serve as a proxy for more complex process-based ecophysiological crop models. {D}epending on the choice of the {BC} method, the uncertainties in the assessment of the {CSI} can be more than twice as high compared to the uncertainties from the {GCM}-{RCM} model selection. {C}omparing the estimated {CSI} values with observed harvest area, it is found that {BC} generally improves the performance for models with low hit rates (< 30-35%), but decreases the performance for models with relatively high hit rates (> 35%). {S}uch consequences can also be expected for process-based crop models, which are developed to operate on field-scale but are driven by coarser scale {RCM}s. {I}t is concluded that such agriculturally oriented climate impact models as investigated here should be interpreted with great caution if applied without {BC} or relying on a single {BC} approach only. {R}ather, we suggest to include different {BC} approaches in the generation of climate projections for {CCIS} and quantify their uncertainties following a super-ensemble probabilistic assessment.},
  keywords = {{B}ias correction ; {C}limate change impact studies ; {C}rop planting date ; {C}rop suitability ; {CORDEX} simulations ; {W}est {A}frica ; {AFRIQUE} {DE} {L}'{OUEST}},
  booktitle = {},
  journal = {{A}gricultural and {F}orest {M}eteorology},
  volume = {304},
  numero = {},
  pages = {108406 [17 p.]},
  ISSN = {0168-1923},
  year = {2021},
  DOI = {10.1016/j.agrformet.2021.108406},
  URL = {https://www.documentation.ird.fr/hor/fdi:010082066},
}