Hydrostreamer v1.0 - Improved streamflow predictions for local applications from an ensemble of downscaled global runoff products

Marko Kallio*, Joseph H.A. Guillaume, Vili Virkki, Matti Kummu, Kirsi Virrantaus

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

14 Downloads (Pure)

Abstract

An increasing number of different types of hydrological, land surface, and rainfall-runoff models exist to estimate streamflow in river networks. Results from various model runs from global to local scales are readily available online. However, the usability of these products is often limited, as they often come aggregated in spatial units which are not compatible with the desired analysis purpose. We present here an R package, a software library Hydrostreamer v1.0, which aims to improve the usability of existing runoff products by addressing the modifiable area unit problem and allows non-experts with little knowledge of hydrology-specific modelling issues and methods to use them for their analyses. Hydrostreamer workflow includes (1) interpolation from source zones to target zones, (2) river routing, and (3) data assimilation via model averaging, given multiple input runoff and observation data. The software implements advanced areal interpolation methods and area-to-line interpolation not available in other products and is the first R package to provide vector-based routing. Hydrostreamer is kept as simple as possible - intuitive with minimal data requirements - and minimises the need for calibration. We tested the performance of Hydrostreamer by downscaling freely available coarse-resolution global runoff products from the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) in an application in 3S Basin in Southeast Asia. Results are compared to observed discharges as well as two benchmark streamflow data products, finding comparable or improved performance. Hydrostreamer v1.0 is open source and is available from http://github.com/mkkallio/hydrostreamer/ (last access: 5 May 2021) under the MIT licence.

Original languageEnglish
Pages (from-to)5155-5181
Number of pages27
JournalGEOSCIENTIFIC MODEL DEVELOPMENT
Volume14
Issue number8
DOIs
Publication statusPublished - 18 Aug 2021
MoE publication typeA1 Journal article-refereed

Fingerprint

Dive into the research topics of 'Hydrostreamer v1.0 - Improved streamflow predictions for local applications from an ensemble of downscaled global runoff products'. Together they form a unique fingerprint.

Cite this