Analyzing satellite and airborne Ka-band passive microwave observations over land for temperature and vegetation monitoring

Abstract

Passive microwave observations at Ka-band (36-37 GHz) have been widely available for decades, but their full potential for land applications has been hardly exploited. This study analyzed Ka-band observations at different spatial scales. Between September and October 2019, a series of airborne flights carrying L- and Ka-band instruments were conducted at the Yanco study area in southeastern Australia. Complementary satellite-based passive microwave data, including Ka-band observations from the Advanced Microwave Scanning Radiometer 2 (AMSR2), were also collected. These data were compared against LST from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat 8, as well as vegetation indices such as the Normalized Difference Vegetation Index (NDVI). A strong correlation (R² = 0.98) was found between Ka-band vertically polarized brightness temperature from AMSR2 and MODIS LST over a 12-year period. Airborne Ka-band observations similarly showed a strong spatial correlation with Landsat 8 LST (R² = 0.70), but only for areas with dense vegetation (NDVI ≥ 0.6). At lower NDVI values, the observations became more sensitive to soil surface characteristics, particularly soil wetness (soil moisture > 0.3 m³ m⁻³), causing up to 20 K drops in brightness temperature. The Ka-band Vegetation Optical Depth (VOD) was derived using a radiative transfer model for both satellite and airborne data. The satellite-derived Ka-band VOD closely matched published VOD products from other frequencies, and the aircraft-based VOD provided realistic spatial patterns over different landscapes. At the satellite scale, a clear relationship between VOD and NDVI was observed. The aircraft-based VOD signal was noisier and had a weak spatial correlation with NDVI, although it demonstrated similar trends as at the satellite scale. These results highlight the promising capability of Ka-band observations for land applications, and its varying sensitivity across scales, with local variability being more pronounced at higher spatial resolutions.