ALBERT

Description: Distributed hydrological models can make predictions with much finer spatial resolution than the supporting field data. They will, however, usually not have a predictive capability at model grid scale due to limitations of data availability and uncertainty of model conceptualisations. In previous publications we have introduced the Representative Elementary Scale (RES) concept as the theoretically minimum scale at which a model with a given conceptualisation has a potential for obtaining a predictive accuracy corresponding to a given acceptable accuracy. The new RES concept has similarities to the 25 year old Representative Elementary Area (REA) concept, but it differs in the sense that while REA addresses similarity between subcatchments by sampling within the catchment, RES focuses on effects of data or conceptualisation uncertainty by Monte Carlo simulations followed by a scale analysis. In the present paper we extend and generalise the RES concept to a framework for assessing the minimum scale of potential predictability of a distributed model applicable also for analyses of different model structures and data availabilities. We present three examples with RES analyses and discuss our findings in relation to Beven's alternative blueprint and environmental modelling philosophy from 2002. While Beven here addresses model structural and parameter uncertainties, he does not provide a thorough methodology for assessing to which extent model predictions for variables that are not measured possess opportunities to have meaningful predictive accuracies, or whether this is impossible due to limitations in data and models. This shortcoming is addressed by the RES framework through its analysis of the relationship between aggregation scale of model results and prediction uncertainties and for considering how alternative model structures and alternative data availability affect the results. We suggest that RES analysis should be applied in all modelling studies that aim to use simulation results at spatial scales smaller than the support scale of the calibration data.

Description: Precipitation gauge catch correction is often given very little attention in hydrological modelling compared to model parameter calibration. This is critical because significant precipitation biases often make the calibration exercise pointless, especially when supposedly physically-based models are in play. This study addresses the general importance of appropriate precipitation catch correction through a detailed modelling exercise. An existing precipitation gauge catch correction method addressing solid and liquid precipitation is applied, both as national mean monthly correction factors based on a historic 30 yr record and as gridded daily correction factors based on local daily observations of wind speed and temperature. The two methods, named the historic mean monthly (HMM) and the time–space variable (TSV) correction, resulted in different winter precipitation rates for the period 1990–2010. The resulting precipitation datasets were evaluated through the comprehensive Danish National Water Resources model (DK-Model), revealing major differences in both model performance and optimised model parameter sets. Simulated stream discharge is improved significantly when introducing the TSV correction, whereas the simulated hydraulic heads and multi-annual water balances performed similarly due to recalibration adjusting model parameters to compensate for input biases. The resulting optimised model parameters are much more physically plausible for the model based on the TSV correction of precipitation. A proxy-basin test where calibrated DK-Model parameters were transferred to another region without site specific calibration showed better performance for parameter values based on the TSV correction. Similarly, the performances of the TSV correction method were superior when considering two single years with a much dryer and a much wetter winter, respectively, as compared to the winters in the calibration period (differential split-sample tests). We conclude that TSV precipitation correction should be carried out for studies requiring a sound dynamic description of hydrological processes, and it is of particular importance when using hydrological models to make predictions for future climates when the snow/rain composition will differ from the past climate. This conclusion is expected to be applicable for mid to high latitudes, especially in coastal climates where winter precipitation types (solid/liquid) fluctuate significantly, causing climatological mean correction factors to be inadequate.

Description: The question of which climate model bias correction methods and spatial scales for correction are optimal for both projecting future hydrological changes as well as removing initial model bias has so far received little attention. For 11 climate models (CMs), or GCM/RCM - Global/Regional Climate Model pairing, this paper analyses the relationship between complexity and robustness of three distribution-based scaling (DBS) bias correction methods applied to daily precipitation at various spatial scales. Hydrological simulations are forced by CM inputs to assess the spatial uncertainty of groundwater head and stream discharge given the various DBS methods. A unique metric is devised, which allows for comparison of spatial variability in climate model bias and projected change in precipitation. It is found that the spatial variability in climate model bias is larger than in the climate change signals. The magnitude of spatial bias seen in precipitation inputs does not necessarily correspond to the magnitude of biases seen in hydrological outputs. Variables that integrate basin responses over time and space are more sensitive to mean spatial biases and less so on extremes. Hydrological simulations forced by the least parameterized DBS approach show the highest error in mean and maximum groundwater heads; however, the most highly parameterised DBS approach shows less robustness in future periods compared with the reference period it was trained in. For hydrological impacts studies, choice of bias correction method should depend on the spatial scale at which hydrological impacts variables are required and whether CM initial bias is spatially uniform or spatially varying. © 2015 John Wiley & Sons, Ltd.

Description: Uncertainty in conceptual model structure and in environmental data is of essential interest when dealing with uncertainty in water resources management. To make quantification of uncertainty possible is it necessary to identify and characterise the uncertainty in geological and hydrogeological data. This paper discusses a range of available techniques to describe the uncertainty related to geological model structure and scale of support. Literature examples on uncertainty in hydrogeological variables such as saturated hydraulic conductivity, specific yield, specific storage, effective porosity and dispersivity are given. Field data usually have a spatial and temporal scale of support that is different from the one on which numerical models for water resources management operate. Uncertainty in hydrogeological data variables is characterised and assessed within the methodological framework of the HarmoniRiB classification.

Description: An existing rain gauge catch correction method addressing solid and liquid precipitation was applied both as monthly mean correction factors based on a 30 yr climatology (standard correction) and as daily correction factors based on daily observations of wind speed and temperature (dynamic correction). The two methods resulted in different winter precipitation rates for the period 1990–2010. The resulting precipitation data sets were evaluated through the comprehensive Danish National Water Resources model (DK-Model) revealing major differences in both model performance and optimized model parameter sets. Simulated stream discharge is improved significantly when introducing a dynamic precipitation correction, whereas the simulated hydraulic heads and multi-annual water balances performed similarly due to recalibration adjusting model parameters to compensate for input biases. The resulting optimized model parameters are much more physically plausible for the model based on dynamic correction of precipitation. A proxy-basin test where calibrated DK-Model parameters were transferred to another region without site specific calibration showed better performance for parameter values based on the dynamic correction. Similarly, the performances of the dynamic correction method were superior when considering two single years with a much dryer and a much wetter winter, respectively, as compared to the winters in the calibration period (differential split-sample tests). We conclude that dynamic precipitation correction should be carried out for studies requiring a sound dynamic description of hydrological processes and it is of particular importance when using hydrological models to make predictions for future climates when the snow/rain composition will differ from the past climate. This conclusion is expected to be applicable for mid to high latitudes especially in coastal climates where winter precipitation type (solid/liquid) fluctuate significantly causing climatological mean correction factors to be inadequate.

Description: Uncertainty in conceptual model structure and in environmental data is of essential interest when dealing with uncertainty in water resources management. To make quantification of uncertainty possible it is necessary to identify and characterise the uncertainty in geological and hydrogeological data. This paper discusses a range of available techniques to describe the uncertainty related to geological model structure and scale of support. Literature examples on uncertainty in hydrogeological variables such as saturated hydraulic conductivity, specific yield, specific storage, effective porosity and dispersivity are given. Field data usually have a spatial and temporal scale of support that is different from the one on which numerical models for water resources management operate. Uncertainty in hydrogeological data variables is characterised and assessed within the methodological framework of the HarmoniRiB classification.