Spatiotemporal fusion datasets of remote sensing images (Wadi al-Dawasir and Poyang Lake)

For use of the datasets, please cite the following paper: Jingan Wu, Shuang Luo, Tongwen Li, Shuheng Zhao, Liupeng Lin, Boosting MODIS-Landsat spatiotemporal fusion accuracy through multi-scale learning with 250-m MODIS observations, 2025, under review Contact information: [e-mail] ----------------------------------------------------------- We provided two datasets made of MODIS and Landsat concurrent observations to assess spatiotemporal fusion algorithms. Two surface change scenarios were examined: non-shape and shape changes. Non-shape change describes spectral variations with constant spatial extent of surface features, such as seasonal vegetation dynamics. In contrast, shape change involves shifts in both spectral signals and spatial coverage, typically associated with land cover transformations. Table 1 summarizes the key characteristics the two datasets. The 30-m Landsat 8 Level-2 surface reflectance data were acquired from the United States Geological Survey (USGS) EarthExplorer portal (https://earthexplorer.usgs.gov/), while the 250-m (MOD09GQ) and 500-m (MOD09GA) Terra MODIS surface reflectance data were downloaded from NASA’s Level-1 and Atmosphere Archive & Distribution System Distributed Active Archive Center (LAADS DAAC) (https://ladsweb.modaps.eosdis.nasa.gov/). The MODIS data were the optimal observations within a five-day time window corresponding to the Landsat observation dates. These datasets underwent preprocessing including geometric correction, radiometric calibration, and atmospheric correction. The two MODIS products were resampled to 240-m and 480-m resolutions, respectively, ensuring integer multiples of Landsat’s 30-m resolution, and then they were upscaled to 30-m resolution for pixel-by-pixel alignment with the Landsat imagery. Three kinds of data were spatially co-registered to two regions of interest, each covering an area of 2,304 km² and corresponds to 1600×1600 Landsat pixels. Table 1. Descriptions on the two time-series datasets Scenarios Datasets Numbers of image triplets Acquisition dates of Landsat images (YYYY-MM-DD) Non-shape change Wadi al-Dawasir 16 2021-01-23, 2021-02-08, 2021-02-24, 2021-03-12, 2021-04-13, 2021-05-31, 2021-06-16, 2021-07-02, 2021-08-03, 2021-08-19, 2021-09-20, 2021-10-06, 2021-10-22, 2021-11-07, 2021-11-23, 2021-12-09. Shape change Poyang Lake 18 2021-01-12, 2021-06-05, 2021-09-25, 2021-12-06, 2022-02-24, 2022-07-10, 2022-08-11, 2022-08-19, 2022-09-12, 2022-10-14, 2022-12-25, 2023-01-18, 2023-02-27, 2023-04-08, 2023-10-17, 2023-11-02, 2023-11-18, 2023-12-28. (1) Wadi al-Dawasir (WAD) The agricultural system in Wadi al-Dawasir (20°17’27”N, 44°44’45”E), located in Riyadh Province of Saudi Arabia, was employed to examine the non-shape change scenario. This desert farming region exhibits a high density of circular crop fields, primarily cultivating alfalfa with small portions of wheat and various vegetables. The distinctive circular patterns, each measuring less than 1 km in diameter, result from center-pivot irrigation systems that extract groundwater through centrally located wells and distribute water via rotating sprinkler mechanisms. These agricultural circles display considerable variations in color due to differences in crop types, planting densities, and growth stages. The sharp contrast between the vibrant crop circles and the surrounding arid desert environment creates significant spectral mixing effects in coarse MODIS imagery. In this area, temporal variations are confined within the fixed boundaries of individual crop circles, making this an exemplary case for investigating non-shape changes. For this study, we acquired and analyzed 16 triplets of concurrent MODIS and Landsat images throughout 2021 within this agricultural area. (2) Poyang Lake (PYL) Poyang Lake (29°03’18”N,116°25’09”E), China’s largest freshwater lake located in northern Jiangxi Province along the southern bank of the Yangtze River, was selected as our study area to examine shape change dynamics. This ecologically significant waterbody experiences pronounced seasonal water-level fluctuations due to combined influences from upstream tributaries and Yangtze River backflow effects. During summer and autumn flood seasons, water levels rise substantially, expanding the lake surface to approximately 150 km (north-south) by 31 km (east-west). Conversely, winter and spring dry seasons see dramatic water level drops, exposing extensive mudflats and marshlands. These cyclical hydrological variations in this area create dynamic shoreline changes that are particularly suitable for investigating shape transformation processes. To capture these changes, we collected 18 triplets of concurrent MODIS and Landsat satellite images from 2021 to 2023 across the Poyang Lake region.