We implemented automated workflows using Jupyter notebooks for each state. The GIS processing, crucial for merging, extracting, and projecting GeoTIFF data, was performed using ArcPy—a Python package for geographic data analysis, conversion, and management within ArcGIS (Toms, 2015). After generating state-scale LES (large extent spatial) datasets in GeoTIFF format, we utilized the xarray and rioxarray Python packages to convert GeoTIFF to NetCDF. Xarray is a Python package to work with multi-dimensional arrays and rioxarray is rasterio xarray extension. Rasterio is a Python library to read and write GeoTIFF and other raster formats. Xarray facilitated data manipulation and metadata addition in the NetCDF file, while rioxarray was used to save GeoTIFF as NetCDF. These procedures resulted in the creation of three HydroShare resources (HS 3, HS 4 and HS 5) for sharing state-scale LES datasets. Notably, due to licensing constraints with ArcGIS Pro, a commercial GIS software, the Jupyter notebook development was undertaken on a Windows OS.

The images have been collected from sentinel-hub apps.sentinel-hub.com, and captured by Sentinel-2 mission satellites with the MSI(Multispectral Instrument), bands 1,2,3,4,8,12 have been merged as a single 6 band GEOTIFF file, which is further cut into tiles of 512x512 resolution. Access the GEOTIFF images using rasterio library in python incase your machine doesn't have the said library; run this on command line/shell pip install rasterio then import the library with import rasterio in your code. For further help on using rasterio, read official documentation https://rasterio.readthedocs.io/en/stable/

World Heightmaps 360px

This is a dataset of 360x360 Earth heightmaps generated from SRTM 1 Arc-Second Global. Each heightmap is labelled according to its latitude and longitude. There are 573,995 samples.

  Method

Convert GeoTIFFs into PNGs with Python and Rasterio.

import rasterio import matplotlib.pyplot as plt import os

input_directory = '...' output_directory = '...' file_list = os.listdir(input_directory)

for i in range(len(file_list)): image =… See the full description on the dataset page: https://huggingface.co/datasets/novaia/world-heightmaps-360px.

This dataset provides a 10-meter resolution Digital Elevation Model (DEM) of Rwanda, offering detailed topographic data suitable for various geospatial applications. The dataset is valuable for:

✅ Terrain Analysis – Slope, aspect, and elevation profiling ✅ Hydrological Modeling – Watershed delineation, runoff simulation, and flood risk assessment ✅ Environmental & Land Use Planning – Soil erosion modeling, agriculture planning, and infrastructure development ✅ Disaster Management – Landslide susceptibility and flood risk mapping ✅ Urban & Civil Engineering – Road network planning and site selection

The DEM is derived from high-quality satellite data, ensuring accurate elevation representation across Rwanda. It can be used with GIS software such as QGIS, ArcGIS, and Google Earth Engine for spatial analysis.

📌 Resolution: 10 meters 📌 Format: GeoTIFF (.tif) 📌 Coverage: Rwanda (Nationwide)

Usage Instructions:

Load into QGIS or ArcGIS for visualization and analysis Use with Python libraries like rasterio, geopandas, and GDAL for geospatial processing Overlay with vector data (e.g., administrative boundaries, rivers, roads) for enhanced analysis

Database including cloud-free Sentinel-2 optical imagery cropped for the Tatra Transboundary Biosphere Reserve area. The data was obtained from the Copernicus Data Space Ecosystem - CDSE service, which provides data for the Earth observation programme Copernicus managed by the European Commission and the European Space Agency. Data acquired by the MSI optical multispectral instrument. Data were preprocessed at the L2A processing level, i.e. including atmospheric and geometric correction.The area included two Sentinel-2 scenes (granules): 34UCV and 34UDV, located on orbits 079, 036. Spectral channels (12 bands) were resampled to a common resolution of 10m and the scenes mosaicked with each other. The raster data were then cropped to the extent of the Tatra Transboundary Biosphere Reserve. For each year, one image from the September-October period was selected to allow spectral coherence of the images for analysis. The open-source library GDAL, rasterio and the Python language were used for data processing.Raster data characteristics:Compression: LZW EPSG code: 32634Number of channels: 12Channels order:'B01' - Coastal aerosol (443 nm, 60m resolution) 'B02' - Blue (490 nm, 10m resolution) 'B03' - Green (560 nm, 10m resolution) 'B04' - Red (665 nm, 10m resolution) 'B05' - Vegetation red edge (705 nm, 20m resolution) 'B06' - Vegetation red edge (740 nm, 20m resolution) 'B07' - Vegetation red edge (783 nm, 20m resolution) 'B08' - Near-infrared (NIR) (842 nm, 10m resolution) 'B8A' - Narrow NIR (865 nm, 20m resolution) 'B09' - Water vapor (945 nm, 60m resolution) 'B11' - Shortwave infrared (SWIR) (1610 nm, 20m resolution) 'B12' - Shortwave infrared (SWIR) (2190 nm, 20m resolution) Research funded by the National Science Centre (NCN), under the project Preludium 22, grant no. 2023/49/N/ST10/00517, entitled: ‘Spruce Forest Damage Assessment Using Machine Learning on Sentinel-2 Time Series in the Tatra Mountains’.

The perceived wealth and physical disorder scores prediction dataset consists of three files, each corresponding to perception data from one of the three analytical levels. Based on the perception image annotation dataset labeled by Chinese urban planners (https://figshare.com/s/a942f102cd07f4a73515), these perception scores are predicted through model training and inference across urban China.The first file (point version) contains perception scores for 36,262,700 street view images. This CSV-formatted data includes the longitude and latitude of each image’s shooting location, the time the image was taken, and the perceived wealth and physical disorder scores. This data can be reprocessed by researchers for their specific analytical need. The second file (grid version) and third file (community version) are stored as GeoTIFF (.tif) files with the Albers conic equal area projection. The mean perceived wealth score and mean perceived physical disorder score from 2013 to 2022 are aggregated using 500m×500m grids and community administrative areas as the analytical units, respectively. They can be processed using GIS software such as ArcGIS and QGIS, as well as Python programming language packages such as Rasterio. We also publish corresponding simplified tables in CSV format showing the mean perceived wealth and physical disorder scores in each grid and community. These tables include the community’s name, the latitude and longitude of the grid (or community) centroid, and the names of the county-, prefecture-, and province-level areas in which the grid (or community) is located.Source publication: Zhang, Y., You, Y., Chen, S., & Cai, L. (2025). Geospatial dataset on human perceptions of wealth and physical disorder in urban China using street view imagery and deep learning. Data in Brief, 112116.

This data set was generated from dual generative adversarial networks that generates building footprints fraction and average building heights using land cover data as the primary conditional input. Model training used c2015 land cover data and building footprints and heights data. Building footprints and heights were sourced from the Model America data set (version 1) accessible via DOI: 10.15485/2283980. Annual 2015 land cover data was based on the National Land Cover Database (NLCD) Collection 1, accessible via DOI: 10.5066/P94UXNTS. The dataset contains 192 files in the GeoTiff format at 30 m spatial resolution containing urban morphology projections for the year 2100 for: two building attributes, i.e., average building heights (BH) and building footprint fractions (BF) (2), two SSP scenarios (SSP3 and SSP5) (2), two population (original and updated) scenarios (2), two developed land intensification scenarios (intensification and weighted average) (2), four intensification levels (low, med, high, and very high) (4), and for each of these scenarios three generative adversarial network outputs (3). File naming convention encodes this metadata directly within the filename to facilitate easy identification, sorting, and retrieval. Each filename follows the following structure: [Building Attribute]-[SSP Scenario]-[Population Scenario]-[Intensification Scenario]-[Intensification Levels]-[Year]-[GANOutputID].tif. Each GeoTiff file has a size of 4280, 3632more » [width, height], and has a projected Albers Equal Area Coordinate Reference System (CRS) with values for BF ranging from 0 to 1 and for BH from 0 to 75 meters. Pixels values are stored with a thirty two bit floating point data type. The dual generative adversarial network architecture used to generate this dataset is based on the original methodology proposed in Goodfellow et al. (2020). Each of the GeoTiff file was read back in python and R programming languages, using rasterio and raster packages, respectively, and metadata outputs were confirmed against expectations.« less

Software presented here was developed using python v3.7, and requires installation the following packages: Rasterio Pandas Numpy Geopandas Jupyter Copy Matplotlib