| 06 Oct 2026
Sunlit and Unlit: The Limitations of Albedo Prediction in Energy Balance Models
Abstract. Energy balance models (EBMs) comprise a simplified physical approach to climate modeling relative to more advanced general circulation model (GCM) counterparts. In particular, EBMs simplify numerous radiative features of Earth's surface-atmosphere system through a singular parameterization of albedo which defines the proportion of reflected to total incoming solar radiation. The literature provides some basic albedo models. However, these models are known to be limited in their real-world application. This paper takes a model identification approach to predicting the rate of change in top-of-atmosphere (TOA) albedo given real climate data. First, we design and implement several regression models for the prediction of the rate of change in albedo. Namely, we consider a variety of polynomial and kernel ridge regression (KRR) solutions, comparing these predictors to benchmark functions from the EBM literature. Second, we evaluate these predictors through a series of case studies in which we justify the selection of simulated temporal resolutions. Third, we quantify the performance of individual albedo predictors and test the impact of coupled temperature and albedo updates. Finally, this work finds that KRR with a radial basis function (RBF) kernel can provide a robust predictor of albedo, but requires careful consideration of tuning, spatial resolution, and input feature selection. In short, the paper delivers a thorough assessment of the limitations of albedo model identification and prediction.
This paper was an interesting read, but I unfortunately must recommend rejection. My reason for this is that, at least by the standards of my field, there isn’t a whole lot of new contribution to climate science. I think this could make a valuable contribution to a different journal – I discuss more below.
EBMs involve only a small portion of the physics contained in GCMs, so calling them simplified GCMs is missing the point. I think what the authors mean to refer to is Reduced Complexity Models, which have quite a bit of associated research and a model intercomparison project devoted to them (RCMIP). I know what uses RCMs can have for understanding basic climate response. But EBMs are so limited and so abstracted from GCMs that the specific purpose for using them needs to be stated, as well as a proposed way of parameterizing all of the processes that you’re not explicitly representing. Geoffroy et al. (2015)’s land/sea contrast model is a good example of doing this. Without such physical grounding, you’re left with a paper whose point is, essentially determining what top-of-atmosphere albedo “should” be for given climate data, even though we are able to both measure and model it quite well. So essentially the premise the authors claim to be pursuing isn’t all that interesting or useful. Your conclusion, that there are limits to albedo model identification and prediction, is a strawman argument – I don’t know of anyone who would think that pure energy balance models _should_ do a great job of reproducing real climate data beyond a very coarse aggregate. I will point out that the introduction does a decent job of reviewing work on EBMs. That in itself could be an interesting contribution.
I think ultimately this is the wrong journal for this sort of work. GMD is primarily focused on developing tools for new geoscience discovery, and I don’t think this paper provides or enables new climate or Earth system science in any convincing capacity. Nor do I think that, given the framework the authors have set up, it can do that. This is an interesting formulation for using simple, physically-based models and constraining them with real-world data for, essentially, optimization. That strikes me as a better fit for a SIAM journal or something like that. And I suspect that if the authors submitted to a SIAM journal, they would have a more friendly review process that is focused on their methods, rather than their contributions to climate science.
On a minor note, there are lots of places where you want to use \citep instead of \cite.