Location: Hydrology and Remote Sensing LaboratoryTitle: Mapping daily leaf area index at 30m resolution over a meadow steppe area by fusing Landsat, Sentinel-2A and MODIS data
|LI, Z. - Chinese Academy Of Agricultural Sciences|
|HUAG, C - University Of Maryland|
|ZHU, Z. - United State Geological Service|
|TANG, H. - Collaborator|
|XIN, X - Chinese Academy Of Agricultural Sciences|
|DING, L. - Chinese Academy Of Agricultural Sciences|
|SHEN, B. - Chinese Academy Of Agricultural Sciences|
|LIU, J - Collaborator|
|CHEN, B. - Chinese Academy Of Agricultural Sciences|
|WANG, X. - Chinese Academy Of Agricultural Sciences|
|YAN, R. - Chinese Academy Of Agricultural Sciences|
Submitted to: International Journal of Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/8/2018
Publication Date: 10/11/2018
Citation: Li, Z., Huag, C., Zhu, Z., Gao, F.N., Tang, H., Xin, X., Ding, L., Shen, B., Liu, J., Chen, B., Wang, X., Yan, R. 2018. Mapping daily leaf area index at 30m resolution over a meadow steppe area by fusing Landsat, Sentinel-2A and MODIS data. International Journal of Remote Sensing. 39(23):9025-9053. https://doi.org/10.1080/01431161.2018.1504342.
Interpretive Summary: Leaf area index (LAI) is a key biophysical parameter used for vegetation growth monitoring and water resource management. Frequent estimates of LAI at fine spatial resolution are essential for capturing rapid changes in vegetation condition. However, current LAI data products derived from satellite remote sensing imageries are either spatially coarse or temporally infrequent. This paper develops an integrated approach that retrieves LAI from the daily Moderate Resolution Imaging Spectroradiometer (MODIS) product at 500m resolution and the 16-day Landsat surface reflectance at 30m resolution. The Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) was applied to merge MODIS and Landsat LAI for generating daily LAI at 30m resolution. Results show good agreements with the ground measurements in northern China under three different pasture management approaches (mowed, grazed and fenced) from 2014-2015. This paper demonstrates a feasible method for producing a daily time-series of LAI from MODIS and Landsat imagery at field scales, which is important for pasture and grazing management.
Technical Abstract: Leaf area index (LAI) is one key parameter of vegetation canopy structure and is closely associated with vegetation photosynthesis, transpiration, and energy balance. Developing a landscape-scale LAI dataset with high temporal resolution (daily) is essential for capturing rapidly vegetation structure variation at field scale and in support of regional biophysical modeling efforts. In this study, a multisensor satellite data fusion algorithm (STARFM), combined with a LAI retrieval radiative transfer model (PROSAIL), were applied to Landsat, Sentinel-2A and MODIS surface reflectance images to generate a daily 30m LAI time series dataset over a meadow steppe site in northern China. The results were assessed using ground measurements and a second LAI dataset fused from Landsat, Sentinel-2A LAI data and MCD15A3H LAI product. Compared to ground measurements, the PROSAIL generated LAI maps from Landsat and MODIS images all showed a high accuracy, the RMSE between Landsat LAI and ground measured LAI is 0.3054 m2/m2, MCD43A4 NBAR (Nadir Bidirectional reflectance distribution function Adjusted Reflectance) data retrieved LAI exhibited a closer seasonal variation with measured and Landsat LAI and had less anomalous points compared to MCD15A3H LAI product. LAI maps generated from Sentinel-2A and Landsat images also showed a good agreement and similar spatial patterns with LAI difference between ±0.5 m2/m2. Estimated using ground measurements, STARFM fused daily LAI estimates from Sentinel-2A, Landsat, and MCD43A4 LAI has an accuracy with RMSE of 0.44 m2/m2 and MAE of 0.34 m2/m2, an improvement from LAI estimated and fused from Landsat, Sentinel-2A LAI data and MCD15A3H LAI product (RMSE of 0.55 m2/m2 and MAE of 0.41 m2/m2), the later dataset also exhibited an anomalous seasonal trajectory with many anomalous points. The use of Sentinel-2A data has effectively increased landscape vegetation observation frequency and provides temporal information about canopy changes occurring between Landsat overpass dates. However, the STARFM algorithm showed a degraded accuracy for small patch landscapes due to the model assumption which was also addressed in existing literature. The scheme developed by this study can be used as a reference for regional vegetation dynamic study and can be applied for larger areas to improve grassland modeling efforts.