Reach Scale Hydrology

Home >> Data Records >> GRADES-hydroDL

Global Reach-scale A priori Discharge Estimates for SWOT

GRADES-hydroDL

The GRADES-hydroDL (Yang et al., 2023) is a major upgrade to the original GRADES (Global Reach-scale A priori Discharge Estimates for SWOT) model-derived daily discharge database for millions of vector river reaches from 1980-present. The major upgrades to GRADES include:

The VIC land surface model (0.25°, daily) that is used to derive gridded runoff fields is replaced with the Long Short-Term Memory (LSTM) model developed by the hydroDL project following Feng et al., 2020.
A different version of MERIT-Basins (MERIT_Hydro_v07_Basins_v01) hydrography is used for routing, and the RAPID river routing model with updated parameters is used.
The precipitation forcing input is taken from a newer version of MSWEP (version 2.8) and the other meteorological fields from the ERA5.
A near real-time (NRT) data stream (upon request) is available following the release of ERA5 every month, i.e., 1-2 months behind real time.

Summary

In recent years, data driven machine learning models, particularly the Long Short-Term Memory (LSTM) model, have shown significant promise in estimating discharge. Despite this, the applicability of LSTM models for global river discharge estimation remains largely unexplored. Here we diverge from the conventional basin-lumped LSTM modeling in limited basins. For the first time, we apply an LSTM on a global 0.25° grid, coupling it with a river routing model to estimate river discharge for every river reach worldwide. We rigorously evaluate the performance over 5332 evaluation gauges globally for the period 2000-2020, separate from the training basins and period. The grid-scale LSTM model effectively captures the rainfall-runoff behavior, reproducing global river discharge with high accuracy and achieving a median Kling-Gupta Efficiency (KGE) of 0.563. It outperforms an extensively bias-corrected and calibrated benchmark simulation based on the Variable Infiltration Capacity (VIC) model, which achieved a median KGE of 0.466. Using the global grid-scale LSTM model, we develop an improved global reach-level daily discharge dataset spanning 1980 to 2020, named GRADES-hydroDL.

See Yang et al., 2023 for more details.

Inputs

Dynamic inputs:

For the precipitation forcing, a recently published global 0.1° and 3‐hourly precipitation dataset MSWEP version 2.8 that optimally merges a range of gauge‐, reanalysis‐, and satellite‐based precipitation (Beck et al., 2019) is used. Other forcing variables (including min/max 2‐m air temperatures and 10‐m wind speed) are obtained from the ERA5. The monthly leaf area index (LAI) is from PROBAV VITO.

Static inputs:

10 sensitive attributes including climate, topography, and soil attributes.

Runoff Simulation - LSTM Model

River Routing

For the river network routing, the Routing Application for Parallel computatIon of Discharge (RAPID; David et al., 2011; David, 2019) is used due to its flexibility in dealing with vector river networks in a range of regional‐ to continental‐scale applications. Global vector river flowlines in MERIT-Basins version 1.0 (MERIT_Hydro_v07_Basins_v01) are used for RAPID routing (~2.94 million, covering 60°S to 90°N).

Validation

A rigorous evaluation is conducted over 5332 (daily) / 5331 (monthly) evaluation gauges globally for the period 2000-2020, separate from the training basins and period to test both the temporal and spatial generalization ability.

Download (Google Drive Linux instructions)

Discharge: preliminary data available from this [Globus collection] (you will need a Globus account if you don't have one already with your institution and the signup is free). If the Globus collection is not reachable, you can also try this [Google Drive folder].

Note: companion paper still in review. If you want to use this pre-publication data for your research, please let us know first: Yuan Yang yuy068@ucsd.edu and Ming Pan m3pan@ucsd.edu.

Underlying hydrography (version 1.0 of MERIT-Basins): [Google Drive] | [TPDC (users in China)]

Reference

Please refer to the following paper(s) for the details of the description of this global discharge database:

Yang, Y., D. Feng, H. E. Beck, W. Hu, A. Sengupta, L. Delle Monache, R. H. Hartman, C. Shen, and M. Pan, 2023: Global Daily Discharge Estimation Based on Grid-Scale Long Short-Term Memory (LSTM) Model and River Routing. Water Resources Research, in review, preprint on ESS Open Archive.

Feng, D., Fang, K., & Shen, C. (2020). Enhancing streamflow forecast and extracting insights using long‐short term memory networks with data integration at continental scales. Water Resources Research, 56(9), e2019WR026793. https://doi.org/10.1029/2019WR026793

Lin, P., M. Pan, H. E., Beck, Y. Yang, D. Yamazaki, R. Frasson, C. H. David, M. Durand, T. M. Pavelsky, G. H. Allen, C. J. Gleason, and E. F. Wood, 2019: Global reconstruction of naturalized river flows at 2.94 million reaches. Water Resources Research, https://doi.org/10.1029/2019WR025287.