This dataset adopts a penalty least squares regression method based on three-dimensional discrete cosine transform. Firstly, seamless global daily L-VOD products are generated. Then, non local filtering ideas are used to achieve spatiotemporal fusion of high-resolution and low resolution data. Finally, a global daily seamless 9-kilometer L-VOD product is generated from January 1, 2010 to July 31, 2021. To verify product quality, time series validation and simulated missing area validation were performed on the reconstructed data. The fusion product has been validated in both time and space, and compared numerically with the original 9-kilometer data during the overlap period. The results showed that the coefficient of determination (R ²) of the seamless SMOS (SMAP) dataset under simulated real missing masks was 0.855 (0.947), and the root mean square error (RMSE) was 0.094 (0.073). The spatiotemporal consistency of the reconstructed daily L-VOD product is consistent with the time series distribution of the original effective values. The spatial information of the fusion product and the original 9-kilometer data during the overlapping period is basically consistent (R ²: 0.926-0.958, RMSE: 0.072-0.093, average absolute error MAE: 0.047-0.064). The time changes of the integrated product and the original product are basically synchronized. This dataset can provide timely vegetation information during natural disasters such as floods, droughts, and forest fires, supporting early disaster warning and real-time response.
1. L-VOD data
(1) The SMOS IC L-VOD dataset is released by the European Space Agency (ESA), with a satellite revisit period of 8 days, a spatial resolution of 25 km, and a global spatial coverage. This study used the latest improved version 2 L-VOD data from January 1, 2010 to December 31, 2017, sourced from https://ib.remote-sensing.inrae.fr/index.php/smos-ic-v2-product-documentation/ These data help to construct baseline data and generate 9-kilometer L-VOD data for the target time.
(2) The SMAP MT-DCA L-VOD dataset covers the global surface, with a satellite revisit period of 3 days and a spatial resolution of 9 km. The dataset uses the latest SMAP MT-DCA 5th edition L-VOD data published by Feldman and Entekhabi (2019) from April 1, 2015 to July 31, 2021, sourced from https://doi.org/10.5281/zenodo.5619583 Using it as high-resolution baseline data in spatiotemporal fusion models to provide fine spatial detail information for VOD fusion products.
2. Auxiliary data
(1) Based on MODIS MCD12C1 V061, the time span is from 2001 to 2022, with a spatial resolution of 0.05 °, and annual global land cover type data.
(2) Based on MODIS MYD13C1 V061, with a spatial resolution of 0.05 °, 16 day synthesis, NDVI data from 2010 to 2021.
1. For the selected VOD_smos and VOD_smap datasets, preprocessing steps such as reprojection, exception handling, and resampling need to be performed;
2. Fill in vacancies;
3. Data fusion;
4. Algorithm settings.
Due to the lack of in-situ L-VOD data, three validation strategies were employed to evaluate the data: (1) time series validation, (2) simulation of missing regions validation, and (3) data comparison validation. Through quantitative and qualitative evaluations, we found that the fusion product VOD_st effectively maintains the stable long-term characteristics of VOD_resmos and achieves good spatial consistency. It is very close in numerical value to VOD_resmap, thus alleviating the underestimation problem related to L-VOD products derived from SMOS satellites.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.ZENODO.DB6892.2025
10.5281/zenodo.13334757
