You are here

Multi-scale Ultra-high Resolution (MUR) Sea Surface Temperature (SST) Analysis

Background

Figure Caption: MUR SST analysis at various scale-dependent stages of its production.
MUR SST analysis at various scale-dependent stages

Sea Surface Temperature (SST) has been observed by satellite instruments since September 1981 and is one of the longest satellite-based records of any Earth climate variable. Measurements from the different satellite sensors need to be calibrated with each other in order to produce an SST Climate Data Record (CDR) that is consistent both in space and time. The microwave (MW) sensors typically have coarser resolutions on the order of 25 km compared to infrared (IR) sensors that can resolve down to 1-km scales. The Figure below illustrates differences in SST using various spatial resolutions, with greater detail observed at higher resolutions. On the other hand, IR-based measurements are prone to data-voids due to cloud contamination, which does not affect MW sensors nearly as much.  Combination of these datasets can thus be complementary, contributing to accuracy of the blended SST maps. The objective for creating the Multi-scale Ultra-high Resolution (MUR) SST is to develop a coherent and consistent daily map of SST at the highest spatial (horizontal) resolution possible.

Contents
MUR provides global SST data every day at a spatial resolution of 0.01 degrees in longitude-latitude coordinates, roughly at 1 km intervals (Figure above, bottom right). Currently, the dataset spans from June 1, 2002 to present (i.e., roughly the duration covered by the Aqua satellite).
 
Each MUR data file adheres to the international SST data standard known as the GHRSST Data Processing Specification (GDS; see http://www.ghrsst.org) and contains an estimate of SST uncertainty, land-mask flag, and sea ice concentration for each SST value provided, plus the longitude-latitude grid coordinates and time of analysis, along with the analyzed SST values. The analyzed SST value provided by the MUR dataset is an estimate of the "foundation temperature", or the near-surface temperature below the extent of diurnal fluctuation due to surface solar heating.
What makes this dataset unique?
The MUR dataset is among the highest resolution SST analysis datasets currently available. The resolving power, contrast, and fidelity afforded by such a technique could benefit identification of surface features, such as oceanic fronts and rings.
 
To evaluate the resolution of an SST map, one must consider not only the grid resolution but also the internal resolution as determined by such measures as the power spectral density (PSD). Comparison of PSDs associated with well-known SST analysis datasets (Figure below, right panel) shows that MUR has the highest power especially for fine-scale features such as the 20-km wavelength.
Power spectral density (PSD) plots for SST datasets.  Left: un-gridded satellite retrievals (“Level 2”) datasets which would be used for production of gridded SST analyses.  Right: gridded (“Level 4”) SST analysis.Figure Caption: Power spectral density (PSD) plots for SST datasets. Left: un-gridded satellite retrievals (“Level 2”) datasets which would be used for production of gridded SST analyses. Right: gridded (“Level 4”) SST analysis
Ultra-high resolution is achieved for the MUR SST dataset using the Multi-Resolution Variational Analysis (MRVA), which employs wavelet-based multi-scale signal expansion to address the irregularity in measurement locations and scale-dependent interpolation issues that are common in satellite-based surface observations. The unique strengths of the MRVA method are:
  • The spectral power (scale-dependent energy distribution) in the SST measurements is preserved better than the available L4 datasets. There is no statistical synthesis of any kind in MUR SST analysis; all high-resolution SST features are due to the measurements.
  • The original measurement coordinates are preserved, without resorting to a pre-gridding procedure such as nearest grid-point relocation or binning, due to the use of continuous basis functions (so-called "grid-less" analysis). An example of this procedure is shown in the Figure below
Demonstration of a “grid-less” interpolation.  Left: observations (open circles) are typically averaged using nearest grid-points (closed circles).  Middle: gridded averaging (colors inside big circles) can distort spatial information like the orientation of the background colors, due to the nearest-neighbor approximation.  Right: MRVA averaged observations at their original locations, hence typically less distortion, even at a low resolution as shown here.
Figure Caption: Demonstration of a “grid-less” interpolation. Left: observations (open circles) are typically averaged to the nearest grid-points (closed circles). Middle: gridded averaging (colors inside big circles) can distort spatial information like the orientation of the background colors, due to the nearest-neighbor approximation. Right: MRVA averaged observations at their original locations, hence typically less distortion, even at a low resolution as shown here.