Reproducibility data and code for evaluating drainage topology in hydropower runoff assignment with non-contributing benchmarks
Description
This dataset contains the analytical inputs, benchmark assignments, computational scripts, expected results and validation records supporting the manuscript “Evaluating drainage topology in hydropower runoff assignment with non-contributing benchmarks”. The study evaluates how comparisons with geometrically matched and contiguous non-contributing supports change the interpretation of runoff indicators associated with Brazilian hydropower generation. The package combines the historical hydropower analysis with subsequent methodological extensions A1–A6. It includes the reconciled entity-month generation panel, ERA5-Land runoff inputs, HydroBASINS network mappings, six historical benchmark families, ENA and naturalized-inflow evaluations, uniform-reference and complete-support-area sensitivities, a common-sample comparison, and an independent monthly observed-streamflow evaluation at 19 CAMELS-BR gauges. A hydropower entity register reconciles the 84 input entities and 52 targets with the 82 entities and 50 targets retained by the primary PPML model. A single Python entry point runs the supplied-data analysis and compares new estimates with archived references. The full mode refits all benchmark distributions used by the packaged model stages. Fixed monthly grid-cell extracts, spatial weights and bootstrap draws support reconstruction of the observed-streamflow results. Code and numerical sources are provided for five main figures and eight supplementary figures, with 600-dpi raster and vector output. SHA-256 manifests, software versions, source attribution and execution reports document the release. The automated workflow begins with frozen analytical inputs. It does not reacquire and rebuild every national raw source file. English editions of internal protocols and original-source hashes preserve auditability; historical timestamps do not constitute public prospective registration. Benchmark proportions and descriptive confidence intervals should not be interpreted as causal effects of river connectivity. This is an expanded, consolidated companion release. It reuses and cites the historical model-stage dataset at https://doi.org/10.17632/3b3g53ph65.1 and adds the subsequent analyses and reproducibility infrastructure. It should be assigned its own DOI by the new Mendeley Data record.
Files
Steps to reproduce
1. Download Hydropower_JOH_English_1of3.zip, Hydropower_JOH_English_2of3.zip and Hydropower_JOH_English_3of3.zip, together with SHA256SUMS.txt, the standalone Run_in_Colab.ipynb and README.md. All three ZIP archives are required. 2. For Google Colab execution, place the three ZIP archives and SHA256SUMS.txt together in a folder in your own Google Drive. Keep the original filenames. Manual extraction is unnecessary for this route. 3. Open the standalone Run_in_Colab.ipynb in Google Colab. Select a CPU runtime and retain the default runtime version. Before running all cells, edit ARCHIVE_DIR to match your own Drive folder, for example /content/drive/MyDrive/Hydropower_reproduction. Keep MODE = 'full'. 4. Run all cells and authorize access to your Drive when prompted. The notebook verifies the archive checksums, extracts the complementary files into one package, creates an isolated Python 3.12.14 environment and installs the specified dependencies. Internet access is required for environment setup. 5. The workflow verifies 1,140 release files and executes all twelve computational stages, including model estimation, benchmark comparisons, supporting analyses, streamflow evaluation, sample accounting and figure generation. Newly calculated results are compared with the supplied numerical references. 6. Confirm that the final VALIDATION_REPORT.json reports status PASS, mode full and PASS for all twelve stages. The notebook saves the reproduced outputs, execution report and results checksum in a new REPRODUCED_RESULTS_<timestamp> subfolder within ARCHIVE_DIR. Alternatively, follow the local execution instructions in README.md using Python 3.12. Extract all three archives into the same parent directory so that they populate a single Hydropower_Runoff_Reproducibility folder. Install requirements.txt in a clean environment and run: python reproduce.py --mode full --output runs/full. The workflow reproduces the analyses from the supplied frozen analytical inputs. It does not reacquire all original source datasets. See README.md, VALIDATION_REPORT.md and COLAB_VALIDATION_REPORT.json for scope and execution evidence.
Institutions
- Universidade de São PauloSão Paulo, São Paulo