Editing IPCC:AR6/WGI/Chapter-11 (section)

=== 11.5.3 Model Evaluation ===

<div id="h2-36-siblings" class="h2-siblings"></div>

Hydrological models used to simulate floods are structurally diverse ( [[#Dankers--2014|Dankers et al., 2014]] ; [[#Mateo--2017|Mateo et al., 2017]] ; [[#Şen--2018|Şen, 2018]] ), often requiring extensive calibration since sub-grid processes and land-surface properties need to be parametrized, irrespective of the spatial resolutions ( [[#Döll--2016|Döll et al., 2016]] ; [[#Krysanova--2017|Krysanova et al., 2017]] ). The data used to drive and calibrate the models are usually of coarse resolution, necessitating the use of a wide variety of downscaling techniques ( [[#Muerth--2013|Muerth et al., 2013]] ). This adds uncertainty not only to the models but also to the reliability of the calibrations. The quality of the flood simulations also depends on the spatial scale, as flood processes are different for catchments of different sizes. It is more difficult to replicate flood processes for large basins, as water management and water use are often more complex for these basins.

Studies that use different regional hydrological models show a large spread in flood simulations ( [[#Dankers--2014|Dankers et al., 2014]] ; [[#Roudier--2016|Roudier et al., 2016]] ; [[#Trigg--2016|Trigg et al., 2016]] ; [[#Krysanova--2017|Krysanova et al., 2017]] ). Regional models reproduce moderate and high flows reasonably well (0.02–0.1 flow annual exceedance probabilities), but there are large biases for the most extreme flows (0–0.02 annual flow exceedance probability), independent of the climatic and physiographic characteristics of the basins (S. [[#Huang--2017|Huang et al., 2017]] a). Global-scale hydrological models have even more challenges, as they struggle to reproduce the magnitude of the flood hazard ( [[#Trigg--2016|Trigg et al., 2016]] ). Also, the ensemble mean of multiple models does not perform better than individual models ( [[#Zaherpour--2018|Zaherpour et al., 2018]] ).

The use of hydrological models for assessing changes in floods, especially for future projections, adds another dimension of uncertainty on top of uncertainty in the driving climate projections, including emissions scenarios, and in the driving climate models (both RCMs and GCMs) ( [[#Arnell--2016|Arnell and Gosling, 2016]] ; [[#Hundecha--2016|Hundecha et al., 2016]] ; [[#Krysanova--2017|Krysanova et al., 2017]] ). The differences in hydrological models ( [[#Roudier--2016|Roudier et al., 2016]] ; [[#Thober--2018|Thober et al., 2018]] ), as well as post-processing of climate model output for the hydrological models ( [[#Muerth--2013|Muerth et al., 2013]] ; [[#Maier--2018|Maier et al., 2018]] ), add to uncertainty for flood projections.

In summary, there is ''medium confidence'' that simulations for the most extreme flows by regional hydrological models can have large biases. Global-scale hydrological models still struggle with reproducing the magnitude of floods. Projections of future floods are hampered by these difficulties and cascading uncertainties, including uncertainties in emissions scenarios and the climate models that generate inputs.

<div id="11.5.4" class="h2-container"></div>

<span id="detection-and-attribution-event-attribution-2"></span>