DCG-MIP: The Debris-Covered Glacier melt Model Intercomparison exPeriment

As glaciers respond to climate change, the scientific community has dedicated increasing attention to the development of melt models for debris-covered glaciers. Here, we present an intercomparison aimed at advancing our understanding of the skills of models of different complexity to simulate ice melt under a debris layer. We compare 14 models with different degrees of complexity at nine sites in the European Alps, Caucasus, Chilean Andes, Nepalese Himalaya and the Southern Alps of New Zealand, over one melt season. We run the models with meteorological data from automatic weather stations and estimated or measured debris properties. Model performance is evaluated using on-site measurements of sub-debris melt (for all models) and surface temperature (for models based on the surface energy balance) at each site. We find that the two main categories of models considered, physically-based energy balance (EB) models and empirical temperature index (TI) models perform in a distinct manner. Temperature index models are reliably accurate when they are recalibrated, and show a range of results when parameters are uncalibrated. The most accurate energy balance models are those with the highest degree of complexity at the atmosphere-debris interface. However, we also find that additional complexity within the debris and at the debris-ice interface does not improve performance, which results from a lack of data to accurately force the models to represent these processes. The difficulty to properly estimate the physical properties of debris layers and within-the-debris processes are a likely cause. One of our main conclusions is thus that sophisticated models need high quality input data. An important data gap emerged from our experiment: the poor performance of all models at three sites was related to poor knowledge of debris properties; specifically, of thermal conductivity. Since debris properties are a major control on melt model simulations, we demonstrate that consistent data acquisition efforts are required to more rigorously evaluate existing models and support new model developments. Future work should seek to advance models by improving how they account for processes such as debris-snow interactions, moisture in the debris and refreezing. We suggest that a systematic effort of model development using a single model framework could be carried out in phase II of the Working Group.