Combining Parametric Land Surface Models with Machine Learning

Abstract

A hybrid machine learning and process-based-modeling (PBM) approach is proposed and evaluated at a handful of AmeriFlux sites to simulate the top-layer soil moisture state. The Hybrid-PBM (HPBM) employed here uses the Noah land-surface model integrated with Gaussian Processes. It is designed to correct the model only in climatological situations similar to the training data else it reverts to the PBM. In this way, our approach avoids bad predictions in scenarios where similar training data is not available and incorporates our physical understanding of the system. Here we assume an autoregressive model and obtain out-of-sample results with upwards of a 3-fold reduction in the RMSE using a one-year leave-one-out cross-validation at each of the selected sites. A path is outlined for using hybrid modeling to build global land-surface models with the potential to significantly outperform the current state-of-the-art.

What Role Does Hydrological Science Play in the Age of Machine Learning?

“…10.1029/2020WR028091 (e.g., Pelissier et al, 2020;Rackauckas et al, 2020), and (iii) architecturally constrained neural networks (e.g., Beucler et al, 2019;Daw et al, 2020). This would allow for hypothesis testing against a backdrop, or null hypothesis, that is derived solely from data (Nearing et al, 2020;Nearing & Gupta, 2015).…”

“…The challenge is to learn the g * (•) function given that we cannot expect to have direct observation pairs (X t , X t − 1 ) of all system states to use for supervised learning. As an example, Nearing and Gupta (2015) applied the data assimilation approach by Ghahramani and Roweis (1999) to the HyMod conceptual rainfall runoff model, and Pelissier et al (2020) applied a similar technique to the Noah-MP land surface model for soil moisture accounting.…”

A Review of Machine Learning Applications in Land Surface Modeling

“…Soil moisture, turbulent momentum, and heat flux are some of the critical variables of land-atmosphere interactions apart from ET. Pelissier et al [61] combined Noah LSM and an ML technique to improve the prediction of top-layer soil moisture at nine AmeriFlux tower sites. They were able to obtain a 3-fold decrease in error metric.…”

Post processing the U.S. National Water Model with a Long Short-Term Memory network

“…Several other methods of combining the benefits of machine learning (predictability) with the benefits of physically realistic hydrologic theory (robustness) are in development. For example, Pelissier et al (2019) use Gaussian Processes to predict error between modeled and observed soil moisture, which allows ML to be used dynamically within a land surface model to correct the soil moisture state at each timestep of a simulation. Another example is using physical principles to constrain the…”

“…One potential problem with ML, however, is that it lacks a physical basis. While there are emerging efforts in hydrology to merge physical understanding with machine learning (Karpatne et al, 2017a;Daw et al, 2020;Pelissier et al, 2019;Chadalawada et al, 2020;Tartakovsky et al, 2020), theory informed machine learning (Karpatne et al, 2017b) is still relatively immature in hydrology.…”