| 28 May 2026
Inferring on-glacier meteorology from physical modeling and remote sensing
Abstract. Local meteorology is crucial to understanding the response of mountain glaciers to climate change, yet it remains one of the largest sources of uncertainty in glacier modeling due to limited observations and complex glacier–atmosphere interactions. Recent advances in high-resolution, globally available remote sensing observations provide new opportunities to exploit observed glacier changes in order to infer high-mountain meteorology from climate reanalysis data. Here, we present a Bayesian framework combining physical energy and glacier mass balance modeling with remote sensing data to infer spatial bias corrections for on-glacier air temperature and precipitation. Our method performs a spatially-distributed inference using a physically-based land-surface model forced with statistically downscaled ERA5-Land reanalysis and an ensemble of bias-correction factors, with satellite-derived glacier surface albedo and surface mass balance as targets. The method is tested and evaluated at four benchmark glaciers in the European Alps and High Mountain Asia, with available independent in-situ observations. Results demonstrate that 1) Leveraging physical modeling with quantitative and multitemporal albedo observations can substantially reduce parameter equifinality in the inferred meteorological bias corrections; 2) Spatially-variable bias corrections improve the consistency between the model and the distributed satellite observations; 3) Compared to statistical downscaling, multi-year average air temperature and precipitation inferred with our framework show improved agreement with nearby station observations and more realistic spatial patterns over glaciers; 4) Our framework provides promising annual and seasonal mass balance with RMSEs relative to in-situ measurements of <1.5 m w.e. and <1 m w.e. respectively, corresponding to improvements of 40-50% and 20-60% compared to without inference. These results make this framework a promising avenue to derive spatial patterns of air temperature and precipitation as well as temporally-resolved glacier mass balance at the regional scale.
Review to Ren et al.: Inferring on-glacier meteorology from physical modeling and remote sensing, submitted to the journal The Cryosphere in 2026
The authors present a study inferring on-glacier fields of air temperature and precipitation correction factors from a Bayesian framework building on a physical, process-based model forced by reanalysis data. The concept was applied on four glaciers with diverse climatological settings, different density and quality of in-situ and remotely sensed observations and (as shown in the discussion) incoherent representation in reanalysis data. From this, the authors conclude that their approach gives reasonable results, but (as usual) relies on the quality of input data for forcing the model and constraining the range of priors for the Bayesian framework.
The manuscript is sound and clear and the idea of the study is very clever, but the title suggests more than what is really delivered. Rating glacier mass balance models in process resolution, index models show the least and process orientated models the highest performance (Hock, 2005). The temperature index model simulates ablation as melting only using a factor (the degree day factor) to express all energy fluxes through air temperature and accumulation through solid precipitation. This setup based on two climate variables is prone to parameter equifinality, especially when forcing data is coarse and/or in-situ data for calibration is sparse (and the authors of the presented manuscript are aware of this). Here, a process orientated framework is used to minimize parameter equifinality and to find the best temperature-precipitation combination with a complex model chain that serves as degree day factor. Hence, the study does not result in inferred meteorology (see also my comment terminology below), the results give a best temperature-precipitation combination to fulfil the given degree day factor, which is here a Bayesian framework building on a physical, process-based model, which is not calibrated to the individual glaciers but applied in a fixed setup optimized in previous studies. This best parameter combination reflects the effective fields of air temperature and precipitation correction factor fitted by the method and is not necessarily true meteorology and thus, we must not use this term. However, this concept is very innovative and worth publishing after addressing my comments.
Terminology: The authors state that they infer on-glacier meteorology. This expression is rather jargon and the term meteorology encompasses much more than just the two elements focused on here. I'd suggest using a more precise wording of e.g., on-glacier fields of effective air temperature and precipitation correction factors, especially in the title, headings and captures and use the term meteorology in floating text only to keep sentences easier to read.
Lapse rate (L 429 and frequently in the manuscript): The temperature gradient in the inferred temperature field is not a lapse rate. It is a temperature gradient along the near glacier surface air layer. Lapse rates are defined thermodynamically and have a maximum of -9.8 K/km in the dry adiabatic case. The lapse rates derived for Abramov glacier (Figure 6) indicate a violation of the lapse rate concept, and generally the near glacier surface air layer is not adiabatic.
Glacier boundary layer context: The method applied in the study is a downscaling of reanalysis data into the glacier boundary layer focusing on air temperature and precipitation. However, I miss a discussion of relevant processes in the glacier boundary layer that this method fails to resolve or is able to parameterize. To constrain the short-wave energy flux the authors apply remotely sensed albedo, for the long-wave radiation a Stefan-Boltzmann approach is used. The latent heat flux is not discussed at all and to constrain the sensible heat flux the model solves for best combinations of precipitation correction factors and air temperature fields. For the latter two fluxes (also called turbulent heat fluxes) there is not enough explanation of their performance, thinking e.g., of wind regimes and atmospheric stability, not meeting the meteorological intention of the study.
Sensitivity to mass balance and testing equifinality: The authors state a resulting RMSE of <1.5 m w.e. of their mass balance modelling. Would it be possible to show a scatter plot similar to Fig. 4 but plotting the Ta/kP combinations against their mass balance RMSE? This way one could imagine how many parameter combinations are below a certain RSME threshold.
Specific comments:
L 37: Include also the role of glacier dynamics in your explanation of glacier responses to climate.
First paragraph of the introduction: The role of glacier winds and turbulence are missing.
L 192: The spatial resolution is not clear. Here you say the 100 m are aggregated to the T&C model resolution, but in L 167 above you say the T&C model resolution is 500 m.
L 236: The method of Liston and Elder (2006) requires a digital elevation model. Describe sources and resolution to which wind speed was downscaled.
L 304: Explain the 2 after the interval in brackets.
L 335, eq 12: For which temperature range and over which surface is the equation valid?
L 339, eq 14: Why is the adjusted incoming LW calculated with an unchanged emissivity? To ensure physical consistency (L 330) also the emissivity must be adjusted to changing moisture.
L598: What are potential reasons why lapse rates (see terminology) become shallower?
Figure 4 (and S5): Does the colour code refer to the to the probability or to the Monte Carlo simulations? If to the latter, what to the non-black points show?
Figure S1: Could you also colour code the years of the acquisition of a scene?
Â
Bibliography:
Hock, R.: Glacier melt: a review of processes and their modelling, Progress in Physical Geography: Earth and Environment, 29, 362–391, https://doi.org/10.1191/0309133305pp453ra, 2005.
Liston, G. E. and Elder, K.: A Meteorological Distribution System for High-Resolution Terrestrial Modeling (MicroMet), J. Hydrometeorol., 7, 217–234, https://doi.org/10.1175/JHM486.1, 2006.
Â