Transferring Hydrologic Data Across Continents – Leveraging Data‐Rich Regions to Improve Hydrologic Prediction in Data‐Sparse Regions

Abstract

There is a great deal of geographic imbalance in global hydrologic data sets. Outside of the US and parts of Europe, there are many parts of the world that have only sparsely available streamflow gauge networks with only a few years' worth of data (Do et al., 2017;Fekete & Vörösmarty, 2007). Besides streamflow gauges, these regions also lack data on physiographic attributes such as geology and soil depth. Nevertheless, climate change is stressing these parts of the world, and accurate hydrologic simulations are needed for these regions just as much, or even more than for data-rich regions.Catchments across the world are often perceived as being unique from each other, requiring customized model development for each basin (Teutschbein & Seibert, 2012). As a rule of thumb, when we create process-based hydrologic models, our development effort scales roughly linearly to the modeled area, computational effort scales linearly at best, and accuracy is unrelated to the number of basins modeled. It is typically difficult to apply knowledge gained from one basin to another, as parameters or experiences do not transfer easily. As a result, although there have been calls for hydrologic studies to transcend the uniqueness

Deep learning approaches for improving prediction of daily stream temperature in data‐scarce, unmonitored, and dammed basins

“…Even after removing seasonality, the median NSE of the residuals was 0.95. These results echo with strong performance metrics reported for LSTM in prediction of soil moisture (Fang et al, 2017;, streamflow (Feng et al, 2020;Kratzert et al, 2019;Xiang et al, 2020) and dissolved oxygen (Zhi et al, 2021), even in spatially data sparse regions (Feng et al, 2021;K. Ma et al, 2021).…”

“…Even after removing seasonality, the median NSE of the residuals was 0.95. These results echo with strong performance metrics reported for LSTM in prediction of soil moisture (Fang et al, 2017; Fang & Shen, 2020), streamflow (Feng et al, 2020; Kratzert et al, 2019; Xiang et al, 2020) and dissolved oxygen (Zhi et al, 2021), even in spatially data sparse regions (Feng et al, 2021; K. Ma et al, 2021). However, as a data‐driven model's quality largely depends on the quality and quantity of the training data, it is unclear how effective such models can be if the sampling frequency is limited, for example, only about 10% of the days are sampled and sampling may be concentrated in time.…”

Improving hydrologic models for predictions and process understanding using Neural ODEs

“…Machine learning methods provide great versatility (Shen, 2018;Shen et al, 2018;Reichstein et al, 2019) and have demonstrated unprecedented accuracy in various modelling tasks like predictions in un-gauged basins (PUB, e.g. Kratzert et al, 2019b;Prieto et al, 2019), in transfer learning to data-scarce regions (Ma et al, 2021) or flood forecasting (Frame et al, 2021;Nevo et al, 2021).…”

Editorial: Broadening the Use of Machine Learning in Hydrology

“…Outside of this Research Topic, machine learning has been applied to soil moisture (Fang et al, 2019), soil data extraction (Chaney et al, 2019), hydrology-influenced water quality variables including in-stream water temperature (Rahmani et al, 2020) and dissolved oxygen (Zhi et al, 2021), human water management through reservoirs (Yang et al, 2019;Ouyang et al, 2021), subsurface reactive transport (Laloy and Jacques, 2019;He et al, 2020), and vadose zone hydrology (Bandai and Ghezzehei, 2021), among others. ML is not only applicable in data-rich regions but can also be leveraged by data-scarce regions (Feng et al, 2021;Ma et al, 2021). DLnative methods for uncertainty quantification have also emerged (Zhu et al, 2019;Fang et al, 2020).…”