Comparing daily multi-resolution SST analysis to satellite track data using varying time windows — Sub-daily analysis may be needed to map the kilometer-scale features
Wednesday, 17 December 2014
Sea surface temperature (SST) has been one of the most extensively observed Earth physical parameters by satellite instruments. Measurements from a diverse set of sensors, including satellite infrared (IR) and microwave (MW) data along with drifter/buoy/ship-based data, provide highly variable coverages of the SST field. For example, the microwave (MW) sensors have typically coarser 25-km resolution than the infra-red (IR) sensors which can resolve down to a 1-km scale. However, the IR-based data are prone to voids due to cloud contamination, which does not affect MW sensors nearly as much. We have combined the coarse MW data with even coarser buoy data along with the high-resolution IR data (from MODIS and AVHRR sensors) into a daily global analysis (synoptic map). Specifically, the Multi-Resolution Variational Analysis (MRVA) algorithm is used to produce the maps at different resolutions: 1km, 2km, 4km, 8km, 16km, 32km, etc. We compare these multi-resolution analyses against independent (not participating in the combination) satellite data sets including the recent 1km NPP-VIIRS and 9km AVHRR tracked (L2) data sets, with a focus on determining how well the higher resolution features are represented in the daily analysis. We have examined the mean differences between the analysis and L2 data as a function of the inherent (feature) scale of the analysis and have observed that the minimum difference (best match) tends to occur near the stated resolution of the L2 data but not at resolution higher than 4km. We also show that the mean difference generally decreases as the match-up time window for the L2 data is limited in duration from a day to its fractions (down to 3 hours) and that the length of the time window affects the finer scales progressively more. These indicate that the resolving capability of the daily analysis combining this particular set of data products is closer to 4km than 1km and that narrowing the analysis time-window can potentially improve the match-up agreement at higher resolution, given the implication that the smaller SST features tend to evolve faster than larger features. A sub-daily analysis using more L2 data sets would then be necessary to map the smaller scale features resolvable by the present-day satellite SST sensors.