Botter et al. (2026) Headwater streams control the non-perennial fraction of the global river network

Identification

• Journal: Nature Water
• Year: 2026
• Date: 2026-01-06
• Authors: Gianluca Botter, Francesca Barone, Nicola Durighetto
• DOI: 10.1038/s44221-025-00549-x

Research Groups

• Department of Civil, Environmental and Architectural Engineering, University of Padova, Padova, Italy.

Short Summary

This study provides new global estimates of non-perennial stream fractions by combining global simulations and detailed field observations, revealing that non-perennial streams are far more prevalent than previously recognized, with the global fraction rising above 0.7 when small headwater streams are comprehensively accounted for.

Objective

• To what extent is the fraction of non-perennial streams in large basins controlled by the headwaters?
• What changes can we expect in current estimates of the proportion of non-perennial streams across different basins and regions when the headwaters are comprehensively accounted for?

Study Configuration

• Spatial Scale: Global (entire world, disaggregated into 60 river basins), regional (Italy, eastern USA), and 12 headwater catchments (Italy, USA) with areas ranging from 0.22 hectares to 23 square kilometers.
• Temporal Scale: Streamflow intermittency is defined as drying out for at least one day per year on average. Field observations of network dynamics were collected multiple times (at least seven) under highly variable hydrological conditions.

Methodology and Data

• Models used:
  • Dual scaling model with analytical expressions for network-scale temporariness (ϕN(A)) and catchment-scale temporariness (ϕC(A)).
  • Power-law scaling relationships for network length (LN(A) = a A−b) and perennial length (Lp,N(A) = cA−d).
  • Informal Generalized Likelihood Uncertainty Estimation for parameter identification.
  • Random forest model (used in the Global Intermittent Rivers and Ephemeral Streams (GIRES) dataset to calculate intermittence probability).
• Data sources:
  • Global Intermittent Rivers and Ephemeral Streams (GIRES) dataset (Messager et al., 2021) for network-scale temporariness across large domains.
  • Empirical field observations from 12 headwater catchments in Italy and the eastern USA, including local persistency maps and digital terrain models (1 to 5 meter resolution).
  • A compiled dataset of 344 empirical drainage density (Dd) values from literature for global Dd estimation.
  • Global map of drainage density by Lin et al. (2021) for spatially variable Dd in disaggregated scenarios.
  • RiverATLAS and HydroATLAS datasets, and a digital terrain model (15 arc-seconds, approximately 500 meters at the Equator) for the GIRES network construction.
  • Spatially distributed discharge data from 5,615 streamflow gauging stations worldwide for training and cross-validation of the GIRES model.

Main Results

• The global fraction of non-perennial channels (Tf) rises above 0.7, reaching up to 0.784 ± 0.020 (aggregated estimate) or 0.728 ± 0.011 to 0.736 ± 0.010 (disaggregated estimate) when small headwater streams are comprehensively accounted for.
• Regional estimates of Tf in relatively humid regions such as Italy and the eastern USA exceed 0.5 (Italy: 0.59 ± 0.01; eastern USA: 0.62 ± 0.02).
• Headwaters consistently control the temporary fraction across all catchment sizes, indicating that the signature of channel wetting and drying persists in larger basins due to the abundance of low-order streams.
• Headwaters (channels with a contributing area smaller than Amax) account for at least 84.4% of Tf in US basins and 83.5% in Italian basins.
• The fraction of non-perennial streams exhibits a strong climatic signature, with higher Tf values associated with more arid regions (wetness index P/PET < 2.5), which accounts for approximately 90% of the global land surface.
• The global mean drainage density is estimated at 10.48 ± 1.13 kilometers per square kilometer, implying a total global river network length exceeding 1.2 billion kilometers (excluding deserts and permanently glaciated regions).

Contributions

• Provides significantly improved global and regional estimates of non-perennial stream prevalence by integrating low-resolution global simulations with detailed field observations, specifically addressing the underrepresentation of headwater streams in previous assessments.
• Introduces a novel dual scaling model that allows for robust extrapolation of network and catchment temporariness across different spatial scales, reconciling the scale mismatch between local observations and large-scale models.
• Quantifies the pivotal role of headwaters in controlling the overall temporary fraction of river networks, demonstrating that flow intermittency is not confined to upland streams but influences larger basins.
• Challenges the perception of river networks as static entities and calls for a paradigm shift in water science to adequately consider channel network dynamics in hydrological, ecological, and societal service assessments.
• Offers a robust and objective framework for quantifying the relative contributions of headwaters and larger streams to the total temporary fraction.

Funding

• ERC project ‘DyNET’ (H2020 European Research Council, grant no. 770999, G.B.).

Citation

@article{Botter2026Headwater,
  author = {Botter, Gianluca and Barone, Francesca and Durighetto, Nicola},
  title = {Headwater streams control the non-perennial fraction of the global river network},
  journal = {Nature Water},
  year = {2026},
  doi = {10.1038/s44221-025-00549-x},
  url = {https://doi.org/10.1038/s44221-025-00549-x}
}