GRIDF-BR: Code, Datasets, and Results for Gridded Intensity–Duration–Frequency Curves for Brazil

DOI: 10.5281/zenodo.20262939 Resource type: DatasetDataset contents: Source code, gridded rainfall products, fitted IDF parameter rasters, intermediate outputs, diagnostics, and reproducible workflow filesProject: GRIDF-BR — Gridded Intensity–Duration–Frequency Curves for BrazilAssociated code repository: https://github.com/marcusnobrega-eng/GRIDF Interactive web application: GRIDF-BR Google Earth Engine App Version: v1.0 Overview This Zenodo record contains the code, datasets, intermediate outputs, final results, and diagnostic products associated with GRIDF-BR, a national-scale framework for generating gridded rainfall Intensity–Duration–Frequency (IDF) curves for Brazil. GRIDF-BR stands for Gridded Intensity–Duration–Frequency Curves for Brazil. The framework combines long-term gridded precipitation products, locally derived sub-daily disaggregation coefficients, bias correction of rainfall extremes, extreme-value frequency analysis, and pixelwise Sherman IDF equation fitting. The archive supports the manuscript A Nationally Consistent Assessment of Intensity–Duration–Frequency Curves for Brazil Using Long-Term Gridded and Satellite Rainfall Data. The manuscript addresses a central limitation in Brazilian design-rainfall practice: existing IDF equations are often spatially fragmented and were developed using heterogeneous records, fitting methods, disaggregation assumptions, and temporal periods. GRIDF-BR was developed to provide a transparent, reproducible, and spatially continuous framework for estimating and comparing design rainfall curves over the Brazilian territory. The dataset is organized to support both reproducibility and practical use. Users who want to inspect or reproduce the complete analysis can download the full working archive. Users who only need final data products can download selected results-only archives containing fitted IDF rasters, annual maximum precipitation rasters, disaggregation coefficients, bias-correction outputs, figures, and diagnostics. Scientific Motivation and Contribution IDF curves are fundamental for hydraulic design, urban drainage planning, stormwater infrastructure, flood-risk screening, and hydrologic engineering. In Brazil, many operational IDF equations remain based on local station analyses, fixed disaggregation coefficients, and regionally heterogeneous methods. This creates difficulties for comparing design rainfall estimates across municipalities, basins, rainfall products, and hydroclimatic regions. GRIDF-BR contributes a nationally consistent workflow and data archive that enables users to evaluate gridded IDF estimates across Brazil using multiple rainfall products and alternative methodological choices. The framework is not intended to replace all local engineering judgment or site-specific studies. Instead, it provides a transparent national baseline, reproducible outputs, and diagnostic layers that help users evaluate rainfall-product behavior, bias correction, sub-daily disaggregation, statistical fitting, and spatial variability. Methodological Scope GRIDF-BR combines rainfall observations, gridded precipitation datasets, geospatial processing, and statistical frequency analysis. The main workflow components are: Daily gridded rainfall products: BR-DWGD, CHIRPS, IMERG, and PERSIANN-CDR. Annual maximum rainfall extraction: yearly maximum daily rainfall rasters are generated for each rainfall product. Bias correction of extremes: multiplicative correction factors are estimated from upper-tail gauge–product rainfall pairs. Sub-daily disaggregation: daily design rainfall depths are converted to sub-daily durations using locally derived disaggregation coefficients from ANA telemetric rainfall stations. Spatial interpolation: station-derived correction factors and disaggregation ratios are interpolated to gridded surfaces when required. Extreme-value analysis: the Gumbel distribution is used as the baseline model, and the Generalized Extreme Value (GEV) distribution is included as a sensitivity formulation. Sherman IDF fitting: gridded IDF curves are summarized through pixelwise four-parameter Sherman equation parameters. Diagnostics: the archive includes fit-quality rasters, Kolmogorov–Smirnov statistics, rejection masks, parity plots, percentile-threshold sensitivity outputs, GEV–Gumbel sensitivity outputs, and benchmark comparison figures. Interactive access: selected GRIDF-BR products are available through the Google Earth Engine web application. Archive Organization The Zenodo record is organized into one full working archive and several results-only archives. This structure allows users to choose between complete reproducibility and targeted download of specific output products. Recommended use Full reproducibility: download GRIDF_BR_full_working_archive_v1.zip. Final IDF products only: download GRIDF_BR_results_idf_parameters_v1.zip. Frequency-analysis inputs: download GRIDF_BR_results_annual_maximum_precipitation_v1.zip. Bias-correction products: download GRIDF_BR_results_bias_correction_outputs_v1.zip. Sub-daily disaggregation products: download GRIDF_BR_results_disaggregation_coefficients_v1.zip. Figures, diagnostics, and sensitivity analyses: download GRIDF_BR_results_figures_and_diagnostics_v1.zip. Files Included in This Zenodo Record 1. GRIDF_BR_full_working_archive_v1.zip This file contains the complete GRIDF-BR working folder used during the manuscript workflow. It preserves the original project organization so that the relationship between scripts, data inputs, intermediate products, final outputs, and figures can be inspected directly. This archive is recommended for users who want the most complete version of the analysis environment. Expected top-level contents include: Bias_Correction/ Bias_Correction_Pipeline/ BrazilShapefiles/ Disag_Coefficients/ Existing_IDFs/ Figures/ GEE_Interface/ IDF_Fitting/ IDF_Parameters_Bias_Corrected/ IDF_Parameters_Raw_Data/ Annual_Maximum_Precipitation/ Zonal_Disaggregation_Coefficients_FINAL/ Misc/ Shapefile/ README.md LICENSE Main contents include: Python scripts used for rainfall-product processing, bias correction, IDF fitting, disaggregation, diagnostics, and plotting. Configuration files used by the modular bias-correction pipeline. Annual maximum daily precipitation rasters. Raw and bias-corrected IDF parameter rasters. Daily-to-sub-daily disaggregation coefficient products. Comparison files for existing Brazilian IDF equations. Manuscript and supplementary figure outputs. Google Earth Engine interface files. Supporting shapefiles, metadata, and diagnostic outputs. Important note: the full working archive preserves the analysis folder structure used during development. Some scripts may contain local path assumptions and may require path updates before execution on another computer. 2. GRIDF_BR_results_idf_parameters_v1.zip This file contains the primary gridded IDF results. It is the recommended archive for users who want to extract IDF curves, map fitted IDF parameters, or compare raw and bias-corrected estimates without rerunning the complete workflow. Expected folders include: IDF_Parameters_Bias_Corrected/ IDF_Parameters_Raw_Data/ These folders contain product-specific outputs for rainfall datasets such as BR-DWGD, CHIRPS, IMERG, and PERSIANN-CDR. Depending on the product and disaggregation option, subfolders may include: CETESB/ RASTER/ STATION/ Typical files include: IDF_k.tif — Sherman equation scale parameter. IDF_a.tif — Sherman return-period exponent. IDF_b.tif — Sherman duration offset parameter. IDF_c.tif — Sherman duration exponent. IDF_params_stack.tif — multiband raster stack containing fitted IDF parameters. IDF_params_stack_named.vrt — virtual raster with named parameter bands. IDF_RMSE.tif — root mean square error of the Sherman fit. IDF_MSE.tif — mean square error of the Sherman fit. IDF_R2.tif — coefficient of determination for the Sherman fit. IDF_KS_D.tif — Kolmogorov–Smirnov D statistic for the daily annual-maximum distribution fit. IDF_KS_p.tif — Kolmogorov–Smirnov p-value. IDF_KS_reject.tif — distribution-fit rejection mask. QC_problem_mask.tif — quality-control mask for disaggregation or interpolation issues. QC_disagg_violation_count.tif — count of disaggregation consistency violations. QC_disagg_violation_magnitude.tif — magnitude of disaggregation consistency violations. QC_nearest_distance_px.tif — distance to the nearest valid source pixel used in repair procedures. QC_nearest_src_row.tif and QC_nearest_src_col.tif — row and column indices of nearest valid source pixels. Recommended uses: Extract local IDF parameters at any raster cell. Generate IDF curves for selected locations. Compare raw and bias-corrected rainfall products. Compare CETESB, station-based, and raster-based disaggregation options. Screen locations using fit-quality and distributional diagnostics. 3. GRIDF_BR_results_annual_maximum_precipitation_v1.zip This file contains annual maximum daily precipitation rasters used as input to the frequency-analysis workflow. Expected contents include: Annual_Maximum_Precipitation/ Product-specific subfolders may include: BR-DWGD/ CHIRPS_Max/ IMERG_V06_Max/ IMERG_V07_Max/ PERSIANN_CDR_Max/ These rasters represent yearly maximum daily precipitation depths for each rainfall product. File names generally identify the rainfall product, grid or processing convention, and year. Recommended uses: Reproduce or modify the annual maximum frequency-analysis step. Fit alternative distributions to annual maximum rainfall series. Compare rainfall products before IDF fitting. Evaluate the spatial structure of daily rainfall extremes. Generate additional return-level products. 4. GRIDF_BR_results_bias_correction_outputs_v1.zip This file contains outputs associated with the rainfall-product bias-correction workflow. Expected folders include: Bias_Correction/ Bias_Correction_Pipeline/ The bias-correction workflow estimates multiplicative correction factors using upper-tail gauge–product rainfall pairs. The correction follows the general relationship: zeta = gauge rainfall / product rainfall corrected rainfall = raw product rainfall × zeta Expected contents may include: Product-specific bias-correction scripts. Modular bias-correction pipeline code. YAML configuration files. Gauge inventory and rainfall-station metadata. Gauge–product event-pair tables. Station-level correction-factor tables. Interpolated correction-factor rasters. Corrected annual maximum precipitation rasters. Bias-correction diagnostic figures. Parity plots comparing gauge rainfall and product rainfall. Percentile-threshold sensitivity outputs. Mean-versus-median estimator sensitivity outputs. Product-specific diagnostic summaries. Recommended uses: Inspect how bias-correction factors were estimated. Compare corrected and uncorrected rainfall products. Evaluate sensitivity to the percentile threshold used for upper-tail event selection. Reproduce or modify the gridded bias-correction workflow. Methodological note: the bias-correction diagnostics are calibration diagnostics based on retained upper-tail event pairs. They should not be interpreted as independent validation using withheld stations or withheld years. 5. GRIDF_BR_results_disaggregation_coefficients_v1.zip This file contains the daily-to-sub-daily rainfall disaggregation products used to convert daily design rainfall depths into sub-daily design rainfall depths before fitting IDF curves. Expected folders include: Zonal_Disaggregation_Coefficients_FINAL/ Disag_Coefficients/ The archive may include both raster and tabular products, such as: relative_to_daily/ relative_to_subdaily/ GeoPackage/ CSV/ zonal summaries coefficient maps diagnostic outputs Two main coefficient families may be present: Relative-to-daily coefficients: ratios expressing sub-daily rainfall depths relative to daily or 24-hour rainfall. Relative-to-subdaily coefficients: chained ratios between durations, such as 5 min / 30 min, 30 min / 1 h, or 1 h / 24 h. Typical raster naming components include: IDW — inverse-distance weighting interpolation. P5m_Pday — rainfall depth ratio between 5 minutes and daily rainfall. R_5m_30m — chained ratio between 5 minutes and 30 minutes. res0.100 — approximate grid resolution in degrees. k10 — interpolation using 10 nearest neighbors. p2.0 — IDW power parameter equal to 2. Recommended uses: Convert daily design rainfall depths to sub-daily design rainfall depths. Compare local disaggregation coefficients with fixed CETESB ratios. Analyze regional variation in rainfall disaggregation across Brazil. Support generation of IDF curves from daily rainfall products. Methodological note: the shortest durations, especially 5 and 10 minutes, may involve modeled extensions where direct station observations are unavailable. These durations should be interpreted with greater caution than directly observed hourly or daily quantities. 6. GRIDF_BR_results_figures_and_diagnostics_v1.zip This file contains manuscript figures, supplementary figures, diagnostics, sensitivity analyses, and benchmark comparison outputs. Expected folders include: Figures/ Existing_IDFs/ IDF_Fitting/ GEV_Gumbel_RE_outputs/ GEV_Gumbel_SA/ Expected contents may include: Manuscript figures and supplementary figures. Bias-correction parity plots and correction-factor maps. Distribution-fit diagnostics for Gumbel and GEV models. GEV–Gumbel return-level sensitivity analyses. Percentile-threshold sensitivity outputs. Disaggregation coefficient maps and fallback-ratio diagnostics. Benchmark comparisons with existing Brazilian IDF equations. Product-specific summary tables and figures. Recommended uses: Reproduce or inspect manuscript figures. Review supplementary diagnostic analyses. Evaluate sensitivity to distributional assumptions. Evaluate sensitivity to bias-correction thresholds. Understand how GRIDF-BR estimates compare with existing Brazilian IDF equations. Accessing the Data Users can access GRIDF-BR products through Zenodo, GitHub, and the Google Earth Engine application. Zenodo This Zenodo record provides downloadable archives for the full working folder and selected results-only product groups. Zenodo is the recommended access point for users who need local copies of GeoTIFFs, CSV files, diagnostic outputs, or manuscript figure products. GitHub The maintained source-code repository is available at https://github.com/marcusnobrega-eng/GRIDF . GitHub should be used for the most up-to-date code, repository documentation, and issue tracking. Google Earth Engine Application The interactive GRIDF-BR application is available at https://gridf-470516.projects.earthengine.app/view/gridf-br . The application allows users to visualize gridded IDF products, query a location, compare rainfall products and disaggregation methods, and export selected outputs without downloading the complete archive. Suggested Local Setup For users who only need final gridded IDF products Download and extract: GRIDF_BR_results_idf_parameters_v1.zip Then open the GeoTIFF files in QGIS, ArcGIS, Python, R, Google Earth Engine, or another geospatial tool. For users who want to reproduce the full workflow Download and extract: GRIDF_BR_full_working_archive_v1.zip Then inspect the workflow folders: Bias_Correction_Pipeline/ IDF_Fitting/ Disag_Coefficients/ Existing_IDFs/ GEE_Interface/ Users may need to update local paths in Python scripts or YAML configuration files before rerunning the workflow. Suggested Python Environment A minimal environment can be created with: conda create -n gridf python=3.10 conda activate gridf Install core Python packages: pip install numpy pandas geopandas rasterio rioxarray xarray scipy matplotlib scikit-learn pyyaml tqdm earthengine-api Optional packages for extended plotting, mapping, or tabular workflows include: pip install contextily cartopy seaborn openpyxl For Google Earth Engine workflows, authenticate using: earthengine authenticate Methodological Notes and Interpretation Bias correction Bias correction is based on multiplicative correction factors derived from gauge–product upper-tail rainfall pairs. These diagnostics should be interpreted as calibration diagnostics, not as fully independent validation using withheld stations or withheld years. Sub-daily disaggregation Sub-daily rainfall depths are obtained using duration-specific disaggregation coefficients. These coefficients include directly observed, station-derived, interpolated, and modeled components. The shortest durations have greater uncertainty where direct observations are limited. Frequency analysis The baseline extreme-value model is Gumbel. The Generalized Extreme Value distribution is included as a sensitivity model. Users interested in long return periods should inspect the distributional diagnostics and GEV–Gumbel sensitivity outputs. Sherman IDF fitting GRIDF-BR summarizes IDF curves using pixelwise Sherman equation parameters. These parameters are stored as GeoTIFF rasters and can be used to reconstruct IDF curves for selected locations. Existing IDF comparisons Comparisons with existing Brazilian IDF equations should be interpreted as contextual benchmarks because the reference equations differ in record length, observational support, disaggregation assumptions, fitting procedures, and temporal periods. Known Limitations Some scripts preserve local path assumptions from the original development environment. The full working archive preserves the analysis folder structure but is not a fully containerized software package. Some outputs may be duplicated between the full working archive and the results-only archives. Very short rainfall durations, especially 5 and 10 minutes, have greater uncertainty than hourly or daily durations. Sparse station coverage may reduce interpolation reliability in some regions. Bias-correction diagnostics should not be interpreted as independent validation. High-return-period rainfall estimates remain sensitive to distribution choice and record length. Comparisons with legacy IDF equations are affected by heterogeneity in existing datasets and methods. Recommended Citation Please cite both the associated manuscript and this Zenodo archive when using GRIDF-BR data, outputs, or code. Suggested dataset citation: Gomes Jr., M. N. (2026). GRIDF-BR: Code, Datasets, and Results for Gridded Intensity–Duration–Frequency Curves for Brazil. Zenodo. https://doi.org/10.5281/zenodo.20262939 . Suggested software citation: Gomes Jr., M. N. (2026). GRIDF-BR: Source Code for Gridded Intensity–Duration–Frequency Curves for Brazil. GitHub. https://github.com/marcusnobrega-eng/GRIDF . BibTeX placeholder: @dataset{gomes_gridf_br_zenodo_2026, author = {Gomes Jr., Marcus N.}, title = {GRIDF-BR: Code, Datasets, and Results for Gridded Intensity--Duration--Frequency Curves for Brazil}, year = {2026}, publisher = {Zenodo}, doi = {10.5281/zenodo.20262939}, url = {https://doi.org/10.5281/zenodo.20262939} } Keywords Intensity–Duration–Frequency; IDF curves; design rainfall; Brazil; rainfall extremes; gridded precipitation; bias correction; sub-daily disaggregation; BR-DWGD; IMERG; CHIRPS; PERSIANN-CDR; Gumbel distribution; GEV distribution; Sherman equation; Google Earth Engine; flood risk; urban drainage; hydrology. Contact Marcus N. Gomes Jr., PhDPostdoctoral Researcher, Stanford UniversityGitHub: https://github.com/marcusnobrega-eng ORCID: https://orcid.org/0000-0002-8250-8195 Project repository: https://github.com/marcusnobrega-eng/GRIDF