Process-based modelling offers interpretability and physical consistency in many domains of geosciences but struggles to leverage large datasets efficiently. Machine-learning methods, especially deep networks, have strong predictive skills yet are unable to answer specific scientific questions. In this Perspective, we explore differentiable modelling as a pathway to dissolve the perceived barrier between process-based modelling and machine learning in the geosciences and demonstrate its potential with examples from hydrological modelling. ‘Differentiable’ refers to accurately and efficiently calculating gradients with respect to model variables or parameters, enabling the discovery of high-dimensional unknown relationships. Differentiable modelling involves connecting (flexible amounts of) prior physical knowledge to neural networks, pushing the boundary of physics-informed machine learning. It offers better interpretability, generalizability, and extrapolation capabilities than purely data-driven machine learning, achieving a similar level of accuracy while requiring less training data. Additionally, the performance and efficiency of differentiable models scale well with increasing data volumes. Under data-scarce scenarios, differentiable models have outperformed machine-learning models in producing short-term dynamics and decadal-scale trends owing to the imposed physical constraints. Differentiable modelling approaches are primed to enable geoscientists to ask questions, test hypotheses, and discover unrecognized physical relationships. Future work should address computational challenges, reduce uncertainty, and verify the physical significance of outputs. Differentiable modelling is an approach that flexibly integrates the learning capability of machine learning with the interpretability of process-based models. This Perspective highlights the potential of differentiable modelling to improve the representation of processes, parameter estimation, and predictive accuracy in the geosciences.

Process-based modelling offers interpretability and physical consistency in many domains of geosciences but struggles to leverage large datasets efficiently. Machine-learning methods, especially deep networks, have strong predictive skills yet are unable to answer specific scientific questions. In this Perspective, we explore differentiable modelling as a pathway to dissolve the perceived barrier between process-based modelling and machine learning in the geosciences and demonstrate its potential with examples from hydrological modelling. ‘Differentiable’ refers to accurately and efficiently calculating gradients with respect to model variables or parameters, enabling the discovery of high-dimensional unknown relationships. Differentiable modelling involves connecting (flexible amounts of) prior physical knowledge to neural networks, pushing the boundary of physics-informed machine learning. It offers better interpretability, generalizability, and extrapolation capabilities than purely data-driven machine learning, achieving a similar level of accuracy while requiring less training data. Additionally, the performance and efficiency of differentiable models scale well with increasing data volumes. Under data-scarce scenarios, differentiable models have outperformed machine-learning models in producing short-term dynamics and decadal-scale trends owing to the imposed physical constraints. Differentiable modelling approaches are primed to enable geoscientists to ask questions, test hypotheses, and discover unrecognized physical relationships. Future work should address computational challenges, reduce uncertainty, and verify the physical significance of outputs. Differentiable modelling is an approach that flexibly integrates the learning capability of machine learning with the interpretability of process-based models. This Perspective highlights the potential of differentiable modelling to improve the representation of processes, parameter estimation, and predictive accuracy in the geosciences.

The Community Land Model (CLM) is the land component of the Community Earth System Model (CESM) and is used in several global and regional modeling systems. In this paper, we introduce model developments included in CLM version 5 (CLM5), which is the default land component for CESM2. We assess an ensemble of simulations, including prescribed and prognostic vegetation state, multiple forcing data sets, and CLM4, CLM4.5, and CLM5, against a range of metrics including from the International Land Model Benchmarking (ILAMBv2) package. CLM5 includes new and updated processes and parameterizations: (1) dynamic land units, (2) updated parameterizations and structure for hydrology and snow (spatially explicit soil depth, dry surface layer, revised groundwater scheme, revised canopy interception and canopy snow processes, updated fresh snow density, simple firn model, and Model for Scale Adaptive River Transport), (3) plant hydraulics and hydraulic redistribution, (4) revised nitrogen cycling (flexible leaf stoichiometry, leaf N optimization for photosynthesis, and carbon costs for plant nitrogen uptake), (5) global crop model with six crop types and time‐evolving irrigated areas and fertilization rates, (6) updated urban building energy, (7) carbon isotopes, and (8) updated stomatal physiology. New optional features include demographically structured dynamic vegetation model (Functionally Assembled Terrestrial Ecosystem Simulator), ozone damage to plants, and fire trace gas emissions coupling to the atmosphere. Conclusive establishment of improvement or degradation of individual variables or metrics is challenged by forcing uncertainty, parametric uncertainty, and model structural complexity, but the multivariate metrics presented here suggest a general broad improvement from CLM4 to CLM5.