The quality of the data products and their accuracy are continually being monitored by the AATSR validation team, led by the validation scientist at the University of Leicester.

The current version of the ATSR archive is Version 3.0. ESA, BEIS (formerlly DECC) and NERC work closely together to maintain the alignment of the mirror archives and also work through the ATSR Quality Working Group to improve the quality of the ATSR data. Details of the quality of the SST products in Verison 3.0 of the ATSR archive can be found in the ATSR V3.0 SST Validation Report available from here.

SST Validation

The formal specifications require that retrieved AATSR dual view SST values achieve an absolute accuracy of better than ± 0.5 K, with ± 0.3 K (one sigma) adopted by the project as the target accuracy. The derived AATSR SST values are estimated from algorithms based on radiative transfer models (see for example Závody et al., 1995, Merchant et al., 1999), which perform a linear regression of SST to simulated brightness temperatures (BTs) with nominal band centres located at 3.7 µm, 11 µm and 12 µm, utilising either the nadir view or a combination of the nadir and forward views offered by the instrument. For well-characterized sensors like AATSR, radiative-transfer based algorithms are now established as effective alternatives to algorithms based on empirical regression (Merchant and Le Borgne, 2004). During the day the 3.7 µm channel is not used due to solar contamination and so there are four possible retrieved SST values, referred to as N2 (nadir two channel), N3 (nadir three channel), D2 (dual view two channel) and D3 (dual view three channel), respectively.

There are two operational Level 2 AATSR SST products that require validation: a 1-km gridded SST product referred to as the ATS_NR__2P product and a spatially averaged SST product (at resolutions of 17 km, 50 km, 10´, and 30´), referred to as either the ATS_AR__2P product or the ATS_MET_2P product; the ATS_MET_2P product is reduced Level 2 product containing only 10´ resolution data for meteorological users. Detailed information on these products and their content can be found in the AATSR product handbook. For continuity with its predecessors, ATSR-1 and ATSR-2, the ATS_NR__2P gridded product is also referred to as the GSST (Gridded Sea Surface Temperature) product and the ATS_AR__2P averaged produced is also referred to as the ASST (averaged Sea Surface Temperature) product.

Activities

Three main methods of SST validation have been used in the analysis to date. First, in situ data from accurate infrared radiometers mounted on ships of opportunity have been used for validation of the AATSR GSST product; this method of validation is the primary means of determining the absolute accuracy of AATSR. Second, global validation of the AATSR ASST product has been performed through comparison with global in situ buoy data; this method of validation offers a means of determining the global accuracy of AATSR. Third, global AATSR ASST data have been compared to data from MODIS, AVHRR, AMSR-E and TMI satellite datasets and HadISST1 SST analyses; this method of validation offers a means of checking the consistency of AATSR against other SST datasets.

Regional Analysis using Radiometers

The AATSR instrument is sensitive to skin SST (the top few µm), which can differ from the bulk SST of the water just a few mm below the skin by some tenths of a degree (see for example Donlon et al., 2002). Therefore, the absolute accuracy of AATSR can only be determined through validation of the AATSR GSST product by comparisons with skin SST measurements retrieved from accurate radiometer measurements collected during in situ validation campaigns. In certain circumstances (i.e. high wind speeds) bulk SST measurements can be used to add to the stock of data available for establishing the performance of AATSR, as it is believed that at wind speeds greater than 6 ms-1 the skin effect becomes predictable (Donlon et al., 2002). However, no bulk SST validation of the GSST product has been carried out so far. Four different radiometers have been used in the validation of the GSST product, the DAR011 operated by CSIRO (Australia), the ISAR operated by SOC (UK), the M-AERI operated by University of Miami (USA) and the SISTeR operated by RAL (UK). A brief description of each in situ instrument is given in Barton et al. (2004). Whatever the method of validation, confidence in the accuracy of the in situ measurements is required. Here, the emphasis is on the need for external, traceable calibration of all measuring equipment to ensure maximum exploitation of data obtained from validation campaigns. All the radiometers used for validating AATSR data so far were recently inter-compared with each other during the 2001 Miami Radiometer Intercomparison (Barton et al., 2004); the outcome being that on average no radiometer differed from any other radiometer by more than 0.15 K.

The ideal scenario is for in situ observations to be made precisely at the time of the satellite overpass and within the instrument swath. Observations outside this coincidence limit will introduce some additional error into the validation data set. For example, tests in the southern Norwegian Sea have shown that spatial separations of about 10 km and time intervals of about 2 hours can introduce r.m.s. differences of 0.2º C into the error budget of a satellite validation dataset (Minnett, 1991), although the exact degree of error introduced by sampling away from exact overpasses will vary according to the local conditions. In frontal regions the variability could be several degrees over just a few km, whereas in more stable regions the variability will be much less. In the case of the Norwegian Sea study (Minnett, 1991), the measurements were made in a highly dynamic area, which will tend to promote large temporal and spatial errors. Validation match-ups observed away from exact coincidence will therefore require justification according to the local conditions before being included in the complete AATSR validation data set. However, at this early stage of the validation programme, the limits on both the spatial difference and the temporal difference between the AATSR data point and the in situ data point are relaxed to include coincidences within 10 km and 3 hours.

Global Analysis using Buoys

Routine validation of the AATSR ASST products against moored and drifting buoys also provides essential information on the performance of the AATSR. Such a method of detecting gross errors in SST at an early stage is advantageous in that it can be carried out on a global and regional scale without field data collection campaigns since the buoy data are provided operationally in an autonomous manner via the Global Telecommunications system (GTS). In addition, it can be used to relate the performance of AATSR to a well established standard that most end users are familiar with and plays a role in linking AATSR to the historical SST record. Indeed, with the increasing emphasis on regional climate change detection and attribution, the importance of global and regional SST field characterisation has become pressing and is now a key part of the AATSR validation plan (DECC, 2004). Since the availability of the near-real-time (NRT) ATS_MET_2P product on the 19th August 2002, the Met Office has been validating it to moored and drifting buoys around the globe, collecting a very large data set of over 10000 match-ups. In addition to the NRT validation, the Met Office have performed a skin to bulk correction using the model of Fairall et al. (1996) to account for the difference between the estimated AATSR skin SST and the bulk SST measured by the buoys at around 1 m depth.

Global Analysis using Other SST Data

The University of Leicester and the Met Office have been comparing AATSR SST data with data from MODIS (http://podaac.jpl.nasa.gov), AVHRR (http://podaac.jpl.nasa.gov), TMI (http://www.remss.com), AMSR-E (http://www.remss.com) and HadISST1 (Rayner et al., 2003) analyses over long time periods. The comparisons between datasets are done using monthly mean SST data at half-degree resolution; monthly mean datasets for AATSR data are calculated as the mean dual view SST from all available data (day and night) from the ASST product. Outputs from the comparisons include global and regional scale difference images, statistics and temperature difference distribution plots.

Validation Documentation

Validation of AATSR is defined as the assessment by independent means of the quality of AATSR data products. Over sea, the primary product of AATSR is SST. Over land, because of the developmental nature of potential land products, namely Land Surface Temperature (LST) and Normalised Difference Vegetation Index (NDVI), the primary product for validation purposes is considered to be top-of-atmosphere (TOA) visible and near infrared reflectances and thermal brightness temperatures.

The AATSR validation documentation is made up of three elements:

This document (part 1 of the validation documentation set) gives an overview of the AATSR validation programme and sets out the principles behind it. Click here to download the document.

This document (part 2 of the validation documentation set) discusses the measurements needed for validation, and recommends the instrumentation and procedures that should be used. Click here to download the document.

This document (part 3 of the validation documentation set) describes the requirement and strategy for obtaining validation data at various stages of the AATSR mission.

Several versions of the VIP exist, covering:

The initial validation of AATSR. Detailed activities that were carried out up to and including the MAVT workshop MERIS and AATSR validation team (MAVT) validation workshop, which was held in October 2003, are described in versions 1 to 3 of the VIP. A key outcome of this workshop was the recommendation that AATSR data should be unconditionally released to all users. Click to download Version1, Version2 or Version3.

Validation during early data exploitation. Validation activities during the first Envisat data exploitation phase are described in version 4 of the VIP. These activities covered the first Envisat Symposium, in September 2004 and conclude with the first AATSR/MERIS user workshop in September 2005. Click here to download the document.

Validation during ongoing exploitation. This document is version 5 of the VIP and describes validation activities to support the ongoing scientific exploitation of AATSR data. Click here to download the document.

All of the documents have been approved by the AATSR Science Advisory Group (SAG).

The AATSR validation programme is part of the ENVISAT validation programme, details of which can be found here in the ENVISAT calibration-validation plan.