Search NASA⌕ Search

Engineering topics

Harman, Ciaran J.

Publications and source records attributed to Harman, Ciaran J..

Differentiable modelling to unify machine learning and physical models for 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. Here, 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.

58 GEOSCIENCES↗

Transit Time Estimation in Catchments: Recent Developments and Future Directions

Abstract Water transit time is now a standard measure in catchment hydrological and ecohydrological research. The last comprehensive review of transit time modeling approaches was published 15+ years ago. But since then the field has largely expanded with new data, theory and applications. Here, we review these new developments with focus on water‐age‐balance approaches and data‐based approaches. We discuss and compare methods including StorAge‐Selection functions, well/partially mixed compartments, water age tracking through spatially distributed models, direct transit time estimates from controlled experiments, young water fractions, and ensemble hydrograph separation. We unify some of the heterogeneity in the literature that has crept in with these many new approaches, in an attempt to clarify the key differences and similarities among them. Finally, we point to open questions in transit time research, including what we still need from theory, models, field work, and community practice.

54 ENVIRONMENTAL SCIENCES↗

Solute Transport and Transformation in an Intermittent, Headwater Mountain Stream with Diurnal Discharge Fluctuations

Time-variable discharge is known to control both transport and transformation of solutes in the river corridor. Still, few studies consider the interactions of transport and transformation together. Here, we consider how diurnal discharge fluctuations in an intermittent, headwater stream control reach-scale solute transport and transformation as measured with conservative and reactive tracers during a period of no precipitation. One common conceptual model is that extended contact times with hyporheic zones during low discharge conditions allows for increased transformation of reactive solutes. Instead, we found tracer timescales within the reach were related to discharge, described by a single discharge-variable StorAge Selection function. We found that Resazurin to Resorufin (Raz-to-Rru) transformation is static in time, and apparent differences in reactive tracer were due to interactions with different ages of storage, not with time-variable reactivity. Overall we found reactivity was highest in youngest storage locations, with minimal Raz-to-Rru conversion in waters older than about 20 h of storage in our study reach. Therefore, not all storage in the study reach has the same potential biogeochemical function and increasing residence time of solute storage does not necessarily increase reaction potential of that solute, contrary to prevailing expectations.

54 ENVIRONMENTAL SCIENCES↗