Submesoscale Sea Surface Temperature Variability from UAV and Satellite Measurements

Abstract

Earlier studies of spatial variability in sea surface temperature (SST) using ship-based radiometric data suggested that variability at scales smaller than 1 km is significant and affects the perceived uncertainty of satellite-derived SSTs. Here, we compare data from the Ball Experimental Sea Surface Temperature (BESST) thermal infrared radiometer flown over the Arctic Ocean against coincident Moderate Resolution Imaging Spectroradiometer (MODIS) measurements to assess the spatial variability of skin SSTs within 1-km pixels. By taking the standard deviation, σ, of the BESST measurements within individual MODIS pixels, we show that significant spatial variability of the skin temperature exists. The distribution of the surface variability measured by BESST shows a peak value of O(0.1) K, with 95% of the pixels showing σ < 0.45 K. Significantly, high-variability pixels are located at density fronts in the marginal ice zone, which are a primary source of submesoscale intermittency near the surface. SST wavenumber spectra indicate a spectral slope of −2, which is consistent with the presence of submesoscale processes at the ocean surface. Furthermore, the BESST wavenumber spectra not only match the energy distribution of MODIS SST spectra at the satellite-resolved wavelengths, they also span the spectral slope of −2 by~3 decades, from wavelengths of 8 km to <0.08 km.

Using Saildrones to Validate Satellite-Derived Sea Surface Salinity and Sea Surface Temperature along the California/Baja Coast

“…Spectral slopes of k -2 to k -3 are consistent with the kinetic energy spectra for mesoscale to submesoscale variability [15,19]. [15] compared spectral slopes from MODIS Terra and Aqua, as well as an Unmanned Surface Vehicle (Ball Experimental Infrared Radiometer). They found spectral slopes of k -2 for wavelengths between 10 m and 100 km, thus encompassing primarily submesoscale variability.…”

“…Spectral slopes were calculated for the SST and SSS spectra (see Table 4). Overlaid on Figures 7 and 8 (as a reference) is the spectral slope of k -2 , which is used as a reference as it is commonly associated with power spectra resolving the mesoscale variability [15]. Table 4 summarizes the spectral slopes, based on the above spectra of the different SST and SSS products.…”

“…Recently, a multi-stage trigonometric interpolation technique has been applied to determine the subpixel variability of satellite SST data, illustrating that this methodology could be used to examine the stripes in MODIS data, as well as possible features associated with cloud contamination [14]. In the Arctic Ocean, measurements from the Ball Experiment SST (BESST) thermal infrared radiometer were compared against MODIS data and revealed significant spatial variability within 1-km pixels that were associated with density fronts in the marginal ice zone [15]. Hence, methodologies must include understanding the subpixel scale variability, especially in areas of high spatial variability such as in upwelling and downwelling regions.…”

Surface Water Temperature Heterogeneity at Subpixel Satellite Scales and Its Effect on the Surface Cooling Estimates of a Large Lake: Airborne Remote Sensing Results From Lake Geneva

“…Past studies have examined subpixel‐scale surface temperature variability by using airborne systems. However, due to the challenges intrinsic to thermal image registration over water (Rahaghi et al, ) or instrumental restrictions, they only reported along‐track point (Mahrt & Khelif, ) or area‐averaged (Castro et al, ) measurements. These studies observed skin temperature variations of >1 °C within ~1‐km distance, which can affect the area‐averaged surface heat flux calculation.…”

An experimental 2DVAR retrieval using AMSR2

“…structures can be brought to bear on the inversion problem as a further constraint. Sub-mesoscale fluctuations in SST fields can exist (Castro et al, 2017), but most oceanic near-surface wind speed variability is synoptically forced and thus spatial decorrelation lengths are significant. Wind retrievals can also benefit from accounting for spatial correlations for ambiguity removal (Vogelzang and Stoffelen, 2018;Lin et al, 2016).…”