Use of expert elicitation to assign weights to climate and hydrological models in climate impact studies

Various methods are available for assessing uncertainties in climate impact studies. Among such methods, model weighting by expert elicitation is a practical way to provide a weighted ensemble of models for specific real-world impacts. The aim is to decrease the influence of improbable models in the results and easing the decision-making process. In this study both climate and hydrological models are analyzed and the result of a research experiment is presented using model weighting with the participation of 6 climate model experts and 6 hydrological model experts. For the experiment, seven climate models are a-priori selected from a larger Euro-CORDEX ensemble of climate models and three different hydrological models are chosen for each of the three European river basins. The model weighting is based on qualitative evaluation by the experts for each of the selected models based on a training material that describes the overall model structure and literature about climate models and the performance of hydrological models for the present period. The expert elicitation process follows a three-stage approach, with two individual elicitations of probabilities and a final group consensus, where the experts are separated into two different community groups: a climate and a hydrological modeller group. The dialogue reveals that under the conditions of the study, most climate modellers prefer the equal weighting of ensemble members, whereas hydrological impact modellers in general are more open for assigning weights to different models in a multi model ensemble, based on model performance and model structure. Climate experts are more open to exclude models, if obviously flawed, than to put weights on selected models in a relatively small ensemble. The study shows that expert elicitation can be an efficient way to assign weights to different hydrological models, and thereby reduce the uncertainty in climate impact. However, for the climate model ensemble, comprising seven models, the elicitation in the format of this study could only reestablish a uniform weight between climate models.

Estimating Regionalized Hydrological Impacts of Climate Change Over Europe by Performance-Based Weighting of CORDEX Projections

Ensemble projections of future changes in discharge over Europe show large variation. Several methods for performance-based weighting exist that have the potential to increase the robustness of the change signal. Here we use future projections of an ensemble of three hydrological models forced with climate datasets from the Coordinated Downscaling Experiment - European Domain (EURO-CORDEX). The experiment is set-up for nine river basins spread over Europe that hold different climate and catchment characteristics. We evaluate the ensemble consistency and apply two weighting approaches; the Climate model Weighting by Independence and Performance (ClimWIP) that focuses on meteorological variables and the Reliability Ensemble Averaging (REA) in our study applied to discharge statistics per basin. For basins with a strong climate signal, in Southern and Northern Europe, the consistency in the set of projections is large. For rivers in Central Europe the differences between models become more pronounced. Both weighting approaches assign high weights to single General Circulation Models (GCMs). The ClimWIP method results in ensemble mean weighted changes that differ only slightly from the non-weighted mean. The REA method influences the weighted mean more, but the weights highly vary from basin to basin. We see that high weights obtained through past good performance can provide deviating projections for the future. It is not apparent that the GCM signal dominates the overall change signal, i.e., there is no strong intra GCM consistency. However, both weighting methods favored projections from the same GCM.

Alleviating the independence assumptions of averaged one-dependence estimators by model weighting

Of numerous proposals to refine naive Bayes by weakening its attribute independence assumption, averaged one-dependence estimators (AODE) has been shown to be able to achieve significantly higher classification accuracy at a moderate cost in classification efficiency. However, all one-dependence estimators (ODEs) in AODE have the same weights and are treated equally. To address this issue, model weighting, which assigns discriminate weights to ODEs and then linearly combine their probability estimates, has been proved to be an efficient and effective approach. Most information-theoretic weighting metrics, including mutual information, Kullback-Leibler measure and the information gain, place more emphasis on the correlation between root attribute (value) and class variable. We argue that the topology of each ODE can be divided into a set of local directed acyclic graphs (DAGs) based on the independence assumption, and multivariate mutual information is introduced to measure the extent to which the DAGs fit data. Based on this premise, in this study we propose a novel weighted AODE algorithm, called AWODE, that adaptively selects weights to alleviate the independence assumption and make the learned probability distribution fit the instance. The proposed approach is validated on 40 benchmark datasets from UCI machine learning repository. The experimental results reveal that, AWODE achieves bias-variance trade-off and is a competitive alternative to single-model Bayesian learners (such as TAN and KDB) and other weighted AODEs (such as WAODE).

Improving the Accuracy of Water Storage Anomaly Trends Based on a New Statistical Correction Hydrological Model Weighting Method

The Gravity Recovery and Climate Experiment (GRACE) satellite solutions have been considerably applied to assess the reliability of hydrological models on a global scale. However, no single hydrological model can be suitable for all regions. Here, a New Statistical Correction Hydrological Model Weighting (NSCHMW) method is developed based on the root mean square error and correlation coefficient between hydrological models and GRACE mass concentration (mascon) data. The NSCHMW method can highlight the advantages of good models compared with the previous average method. Additionally, to verify the effect of the NSCHMW method, taking the Haihe River Basin (HRB) as an example, the spatiotemporal patterns of Terrestrial Water Storage Anomalies (TWSA) in HRB are analyzed through a comprehensive comparison of decadal trends (2003–2014) from GRACE and different hydrological models (Noah from GLDAS-2.1, VIC from GLDAS-2.1, CLSM from GLDAS-2.1, CLSM from GLDAS-2.0, WGHM, PCR-GLOBWB, and CLM-4.5). Besides, the NSCHMW method is applied to estimate TWSA trends in the HRB. Results demonstrate that (1) the NSCHMW method can improve the accuracy of TWSA estimation by hydrological models; (2) the TWSA trends continue to decrease through the study period at a rate of 15.7 mm/year; (3) the WGHM and PCR-GLOBWB have positive reliability with respect to GRACE with r > 0.9, while all the other models underestimate TWSA trends; (4) the NSCHMW method can effectively improve RMSE, NES, and r with 3–96%, 35–282%, 1–255%, respectively, by weighting the WGHM and PCR-GLOBWB. Indeed, groundwater depletion in HRB also proves the necessity of the South-North Water Diversion Project, which has already contributed to groundwater recovery.

Changes in mean and extreme temperature and precipitation events from different weighted multi-model ensembles over the northern half of Morocco

Internal variability, multiple emission scenarios, and different model responses to anthropogenic forcing are ultimately behind a wide range of uncertainties that arise in climate change projections. Model weighting approaches are generally used to reduce the uncertainty related to the choice of the climate model. This study compares three multi-model combination approaches: a simple arithmetic mean and two recently developed weighting-based alternatives. One method takes into account models’ performance only and the other accounts for models’ performance and independence. The effect of these three multi-model approaches is assessed for projected changes of mean precipitation and temperature as well as four extreme indices over northern Morocco. We analyze different widely used high-resolution ensembles issued from statistical (NEXGDDP) and dynamical (Euro-CORDEX and bias-adjusted Euro-CORDEX) downscaling. For the latter, we also investigate the potential added value that bias adjustment may have over the raw dynamical simulations. Results show that model weighting can significantly reduce the spread of the future projections increasing their reliability. Nearly all model ensembles project a significant warming over the studied region (more intense inland than near the coasts), together with longer and more severe dry periods. In most cases, the different weighting methods lead to almost identical spatial patterns of climate change, indicating that the uncertainty due to the choice of multi-model combination strategy is nearly negligible.

The Mediterranean climate change hotspot in the CMIP5 and CMIP6 projections

The increased warming trend and precipitation decline in the Mediterranean region makes it a climate change hotspot. We compare projections of multiple CMIP5 and CMIP6 historical and scenario simulations to quantify the impacts of the already changing climate in the region. In particular, we investigate changes in temperature and precipitation during the 21st century following scenarios RCP2.6, SSP1-2.6, RCP4.5, SSP2-4.5, RCP8.5 and SSP5-8.5, as well as the HighResMIP high resolution experiments. A model weighting scheme is applied to obtain constrained estimates of projected changes, which accounts for historical model performance and inter-independence of the multi-model ensembles, using an observational ensemble as reference. Results indicate a robust and significant warming over the Mediterranean region along the 21st century over all seasons, ensembles and experiments. The Mediterranean amplified warming with respect to the global mean is mainly found during summer. The temperature changes vary between CMIPs, being CMIP6 the ensemble that projects a stronger warming. Contrarily to temperature projections, precipitation changes show greater uncertainties and spatial heterogeneity. However, a robust and significant precipitation decline is projected over large parts of the region during summer for the high emission scenario. While there is less disagreement in projected precipitation between CMIP5 and CMIP6, the latter shows larger precipitation declines in some regions. Results obtained from the model weighting scheme indicate increases in CMIP5 and reductions in CMIP6 warming trends, thereby reducing the distance between both multi-model ensembles.

The Estimation of Magnetite Prospective Resources Based on Aeromagnetic Data: A Case Study of Qihe Area, Shandong Province, China

In the Qihe area, the magnetic anomalies caused by deep and concealed magnetite are weak and compared with ground surveys, airborne surveys further weaken the signals. Moreover, the magnetite in the Qihe area belongs to a contact-metasomatic deposit, and the magnetic anomalies caused by the magnetite and its mother rock overlap and interweave. Therefore, it is difficult to directly delineate the target areas of magnetite according to the measured aeromagnetic maps in Qihe or similar areas, let alone estimate prospective magnetite resources. This study tried to extract magnetite-caused anomalies from aeromagnetic data by using high-pass filtering. Then, a preliminary estimation of magnetite prospective resources was realized by the 3D inversion of the extracted anomalies. In order to improve the resolution and accuracy of the inversion results, a combined model-weighting function was proposed for the inversion. Meanwhile, the upper and lower bounds and positive and negative constraints were imposed on the model parameters to further improve the rationality of the inversion results. A theoretical model with deep and concealed magnetite was established. It demonstrated the feasibility of magnetite-caused anomaly extraction and magnetite prospective resource estimation. Finally, the magnetite-caused anomalies were extracted from the measured aeromagnetic data and were consistent with known drilling information. The distribution of underground magnetic bodies was obtained by the 3D inversion of extracted anomalies, and the existing drilling data were used to delineate the volume of magnetite. In this way, the prospective resources of magnetite in Qihe area were estimated.

Reduced global warming from CMIP6 projections when weighting models by performance and independence

<p><span>To extract reliable estimates of future warming and related uncertainties from multi model ensembles such as CMIP6, the spread in their projections is often translated into probabilistic estimates such as the mean and likely range. Here, we use a model weighting approach, which accounts for the CMIP6 models' historical performance as well as their interdependence, to calculate constrained distributions of global mean temperature change.</span></p><p><span>We investigate the skill of our approach in a perfect model test framework, where we use previous-generation CMIP5 models as pseudo-observations in the historical period. The performance of the distribution weighted in the abovementioned manner with respect to matching the pseudo-observations in the future is then evaluated, and we find a mean increase in skill of about 17 % compared with the unweighted distribution. In addition, we show that our independence metric correctly clusters models known to be similar based on a CMIP6 “family tree”, which enables the application of a weighting based on the degree of inter-model dependence.</span></p><p><span>We then apply the weighting approach, based on two observational estimates, to constrain CMIP6 projections. Our results show a reduction in the projected mean warmingbecause some CMIP6 models with high future warming receive systematically lower performance weights. The mean of end-of-century warming (2081–2100 relative to 1995–2014) for SSP5-8.5 with weighting is 3.7°C, compared with 4.1°C without weighting; the likely (66%) uncertainty range is 3.1 to 4.6°C, which equates to a 13 % decrease in spread.</span></p><p><br><br></p>