Abstract: The microwave vegetation index, known as vegetation optical depth (VOD), is a dimensionless vegetation index.VOD is an important parameter for retrieving soil moisture and surface biomass. Additionally, VOD serves as an important complement to optical vegetation indices in monitoring vegetation growth. However, the spatial resolution of existing passive microwave VOD products is relatively low, which hinders the application of VOD in small regional areas. Therefore, the downscaling of VOD is of significant importance.
The normalized difference vegetation index (NDVI) is a widely used vegetation index. As an optical vegetation index, NDVI is proven to have a strong correlation with VOD. The spatial resolution of NDVI is typically high, with numerous kilometer-scale products being widely used. There are currently two primary approaches for downscaling VOD. One involves integrating active microwave remote sensing, while the other involves integrating optical remote sensing.Currently, there is limited research focused on downscaling VOD, particularly using optical vegetation indices, which is very uncommon.Thus, statistical methods can be employed in conjunction with NDVI to spatially downscale VOD, exploring the feasibility of using optical remote sensing data for VOD downscaling. NDVI with spatial resolution of 1 km is utilized to fuse and obtain VOD at 1 km scale.
Penetration depths of VOD at various frequencies differ, leading to distinct relationships with NDVI. Firstly, it is essential to analyze the relationship between VOD at different frequencies and NDVI. Spatial and temporal similarities and differences of VOD and NDVI in China from 2003 to 2022 are compared, between C-band (electromagnetic wave frequencies of 4-8 GHz), X-band (frequencies of 8-12 GHz), and Ku-band (frequencies of 12-18 GHz), as well as relative value differences across various vegetation types. A downscaling method that utilizes classification residual correction is proposed. The method initially establishes an equation that represents the difference between NDVI and VOD before and after downscaling. It then utilizes similar classification characteristics of NDVI and VOD to allocate differences, thereby facilitating the spatial downscaling of 0.25° VOD to 1 km resolution. The spatial downscaling of Ku-band VOD with NDVI is performed and assessed.
Results indicate that Ku-band VOD exhibits the highest spatial and temporal correlation with NDVI, showing a consistent response across various vegetation types, with a correlation coefficient of 0.8240. The spatial resolution of downscaled VOD is significantly enhanced, preserving spatial distribution characteristics of original data while accurately reflecting the temporal variation features of VOD.
The application of optical vegetation indices are investigated to spatially downscale VOD, offering a reference for further research. 1 km VOD data generated through downscaling facilitates the application of vegetation optical depth at regional scales and plays an important role in monitoring local vegetation changes and estimating related ground parameters with higher spatial resolution.