Search for: All records

In many environmental applications, recurrent neural networks (RNNs) are often used to model physical variables with long temporal dependencies. However, due to minibatch training, temporal relationships between training segments within the batch (intra-batch) as well as between batches (inter-batch) are not considered, which can lead to limited performance. Stateful RNNs aim to address this issue by passing hidden states between batches. Since Stateful RNNs ignore intra-batch temporal dependency, there exists a trade-off between training stability and capturing temporal dependency. In this paper, we provide a quantitative comparison of different Stateful RNN modeling strategies, and propose two strategies to enforce both intra- and inter-batch temporal dependency. First, we extend Stateful RNNs by defining a batch as a temporally ordered set of training segments, which enables intra-batch sharing of temporal information. While this approach significantly improves the performance, it leads to much larger training times due to highly sequential training. To address this issue, we further propose a new strategy which augments a training segment with an initial value of the target variable from the timestep right before the starting of the training segment. In other words, we provide an initial value of the target variable as additional input so that the network can focus on learning changes relative to that initial value. By using this strategy, samples can be passed in any order (mini-batch training) which significantly reduces the training time while maintaining the performance. In demonstrating the utility of our approach in hydrological modeling, we observe that the most significant gains in predictive accuracy occur when these methods are applied to state variables whose values change more slowly, such as soil water and snowpack, rather than continuously moving flux variables such as streamflow. 

There is a growing consensus that solutions to complex science and engineering problems require novel methodologies that are able to integrate traditional physics-based modeling approaches with state-of-the-art machine learning (ML) techniques. This paper provides a structured overview of such techniques. Application-centric objective areas for which these approaches have been applied are summarized, and then classes of methodologies used to construct physics-guided ML models and hybrid physics-ML frameworks are described. We then provide a taxonomy of these existing techniques, which uncovers knowledge gaps and potential crossovers of methods between disciplines that can serve as ideas for future research. 

Abstract. Agricultural nitrous oxide (N2O) emission accounts for a non-trivialfraction of global greenhouse gas (GHG) budget. To date, estimatingN2O fluxes from cropland remains a challenging task because the relatedmicrobial processes (e.g., nitrification and denitrification) are controlledby complex interactions among climate, soil, plant and human activities.Existing approaches such as process-based (PB) models have well-knownlimitations due to insufficient representations of the processes oruncertainties of model parameters, and due to leverage recent advances inmachine learning (ML) a new method is needed to unlock the “black box” toovercome its limitations such as low interpretability, out-of-sample failureand massive data demand. In this study, we developed a first-of-its-kindknowledge-guided machine learning model for agroecosystems (KGML-ag) byincorporating biogeophysical and chemical domain knowledge from an advanced PBmodel, ecosys, and tested it by comparing simulating daily N2O fluxes withreal observed data from mesocosm experiments. The gated recurrent unit (GRU)was used as the basis to build the model structure. To optimize the modelperformance, we have investigated a range of ideas, including (1) usinginitial values of intermediate variables (IMVs) instead of time series asmodel input to reduce data demand; (2) building hierarchical structures toexplicitly estimate IMVs for further N2O prediction; (3) using multi-tasklearning to balance the simultaneous training on multiple variables; and (4)pre-training with millions of synthetic data generated from ecosys and fine-tuningwith mesocosm observations. Six other pure ML models were developed usingthe same mesocosm data to serve as the benchmark for the KGML-ag model.Results show that KGML-ag did an excellent job in reproducing the mesocosmN2O fluxes (overall r2=0.81, and RMSE=3.6 mgNm-2d-1from cross validation). Importantly, KGML-ag always outperformsthe PB model and ML models in predicting N2O fluxes, especially forcomplex temporal dynamics and emission peaks. Besides, KGML-ag goes beyondthe pure ML models by providing more interpretable predictions as well aspinpointing desired new knowledge and data to further empower the currentKGML-ag. We believe the KGML-ag development in this study will stimulate anew body of research on interpretable ML for biogeochemistry and otherrelated geoscience processes. 

Streamflow prediction is a long‐standing hydrologic problem. Development of models for streamflow prediction often requires incorporation of catchment physical descriptors to characterize the associated complex hydrological processes. Across different scales of catchments, these physical descriptors also allow models to extrapolate hydrologic information from one catchment to others, a process referred to as “regionalization”. Recently, in gauged basin scenarios, deep learning models have been shown to achieve state of the art regionalization performance by building a global hydrologic model. These models predict streamflow given catchment physical descriptors and weather forcing data. However, these physical descriptors are by their nature uncertain, sometimes incomplete, or even unavailable in certain cases, which limits the applicability of this approach. In this paper, we show that by assigning a vector of random values as a surrogate for catchment physical descriptors, we can achieve robust regionalization performance under a gauged prediction scenario. Our results show that the deep learning model using our proposed random vector approach achieves a predictive performance comparable to that of the model using actual physical descriptors. The random vector approach yields robust performance under different data sparsity scenarios and deep learning model selections. Furthermore, based on the use of random vectors, high‐dimensional characterization improves regionalization performance in gauged basin scenario when physical descriptors are uncertain, or insufficient.