ABSTRACT:

Data and code for Jiang et al., "Global Analysis of Hack's Law: Decomposing the Geometric Origins of h > 0.5 Across Scales and Sampling Designs" (Water Resources Research, manuscript 2026WR044573, revised version of October 2026). This is version 1.1; it replaces version 1.0 (https://doi.org/10.4211/hs.d29b097e15524e13afc61cd229819b96), which the submitted manuscript cites. CHANGES.md lists what changed.

The tables cover all 340,991 drainage basins of at least 1 km² on MERIT Hydro (3 arc-seconds, about 90 m), 331,299 of which drain to the coast. For each basin they give per-basin Hack's law fits (along-stem and tributary-junction), shape descriptors, outlet type and 22 environmental predictors. The package also contains the along-stem and junction samples behind the pooled fits, the repetition of the analysis on HydroSHEDS v2 at 1 arc-second (about 30 m) in South and North America, the channel-head experiment, the list of junction samples excluded as flow lines on lake surfaces, and basin outlines at 0.05° for the maps.

The code consists of the Python programs that build the tables from the input products, and the analysis and figure scripts that reproduce the numbers and figures of the paper. README.md describes the files and how to run them; DATA_SOURCES.md lists the input products and their licenses. Code: MIT license. Tables: CC BY 4.0.

ABSTRACT:

Data and code repository for "Global Analysis of Hack's Law: Decomposing the Geometric Origins of h > 0.5 Across Scales and Sampling Paradigms" (Jiang et al., 2026, submitted to Water Resources Research).

This dataset contains 340,991 globally extracted drainage basins derived from MERIT Hydro (Yamazaki et al., 2019, ~90 m resolution), covering 180°W–180°E, 60°S–85°N. For each basin we provide per-basin Hack's law parameters (exponent h, prefactor
C, R²) computed under five sampling paradigms—along-stem and tributary-junction at both pooled and per-basin scopes, plus independent-outlet fitting—along with 23 environmental covariates spanning terrain, lithology and soil, landform, climate, ectonics, and basin-type categories, and four dimensionless shape descriptors (sinuosity, Schumm elongation ratio, convexity, Gravelius compactness coefficient).

The repository also includes the complete Python pipeline (17 scripts) that reproduces every figure in the manuscript (Fig 1–12 main + Fig A1–A5 appendix), the cached random forest feature-importance results (rf_importance_results_v1.pkl), and a
robustness-check JSON (rf_robustness_results.json) covering cross-fold Spearman ρ, threshold sensitivity, and gradient-boosting cross-validation.

ABSTRACT:

This study enhances the understanding of the predictability of subseasonal-to-seasonal (S2S) scale streamflow and flooding and their forecasting skills, while tackling long-time challenges introduced by low-skill climate forcings (SCFs). Utilizing six numerical weather prediction (NWP) models and a calibrated hydrological model, we assess the precipitation, streamflow, and flood predictability at 24 river stations in the rainfall-dominant Pearl River Basin, South China. Various model configurations are conducted to address uncertainties in initial hydrological conditions (IHCs) and SCFs, with the conventional ensemble streamflow prediction (ESP) serving as the benchmark. The findings are: (1) NWP driven hydrological forecasts demonstrate enhanced streamflow forecasting proficiency, achieving KGE >0.5 over 44 days, significantly outperforming its precipitation forecast skill with KGE >0.2 up to 10 days; (2) NWP-based deterministic streamflow predictions also outperform ESP, with KGE exceeding 0.6 for 20 days versus ESP's 5 days, and the Critical Success Index (CSI) for flood event detection over 0.3 for three weeks compared to only one week for ESP; (3) Bias correction of NWP precipitation further improves deterministic streamflow forecasts, with KGE > 0.6 for 30 days while having less effects on probilistic forecasting; (4) The performance enhancement from precipitation to streamflow forecast is mainly due to the persistent and slow response of streamflow to subsurface flow, while it is dominated by baseflow condition. Overall, IHCs initially dominate S2S forecasting, while SCF's impact accumulates, overtaking IHCs at a certain lead time, which indicates a future pathway to shift the S2S "predictability desert" narrative by enhancing both IHC and SCF.