Methods InfraRed

NOAA has been producing global Sea Surface Temperature (SST) products from AVHRR onboard NOAA polar satellites since the early 1980s, using multi-channel SST (MCSST) and later non-linear SST (NLSST) regression algorithms. Later in the 1990s, SST from imagers onboard GOES satellites was introduced. Historically, the polar and geo SST retrieval systems were developed independently, and diverged over time, thus complicating their development and maintenance.

In the mid 2000s, NOAA initiated a new “enterprise” SST system, the Advanced Clear-Sky Processor for Oceans (ACSPO), which is capable of processing data from various sensors and platforms, both polar and geo. In addition to the satellite brightness temperatures (BTs), their simulated counterparts are calculated using the fast Community Radiative Transfer Model (CRTM) with first guess SST and atmospheric fields as input. Simulated BTs are used to test improved clear-sky mask and physical (i.e., RTM based) SST retrieval algorithms. ACSPO makes retrievals in a wide range of view zenith angles (in particular, in full swath for polar sensors, up to VZA~70deg; cf. VZA<~55deg limitation for heritage systems), and reports every clear sky ocean pixel with a valid SST (unlike the heritage systems, which heavily sub-sample).

Another priority was to develop a comprehensive, near-real time, online SST monitoring system, including the SST Quality Monitor (SQUAM; www.star.nesdis.noaa.gov/sod/sst/squam/), in situ SST Quality Monitor (iQuam; www.star.nesdis.noaa.gov/sod/sst/iquam/), and Monitoring of IR Clear-sky Radiances over Ocean for SST (MICROS; www.star.nesdis.noaa.gov/sod/sst/micros/). SQUAM monitors global satellite and blended SST products, and validates them against quality controlled in situ SSTs provided by iQuam. The MICROS system monitors corresponding global clear-sky TOA BTs against CRTM simulations, and is instrumental to unscramble artifacts in the SST time series, by linking them to the corresponding BTs. Two other online systems are being developed at NOAA: the Sensor Stability for SST (3S; www.star.nesdis.noaa.gov/sod/sst/3s/; aimed at linking BT anomalies to sensor calibration and orbital configuration), and the ACSPO Regional Monitor for SST (ARMS; www.star.nesdis.noaa.gov/sod/sst/arms/; aimed at regional monitoring of ACSPO products).

In this presentation, status of ACSPO SST products and their evaluation in the NOAA SST Monitoring system are reviewed, and future work discussed.

ACSPO operationally runs with AVHRR data, NOAA GAC (4km) and Metop FRAC (1km global). Reprocessing of AVHRR GAC data from 2002-pr is underway.

In October 2011, the S-NPP satellite, a bridge to the new US Joint Polar Satellite System (JPSS) series, was launched. The Visible Infrared Imager Radiometer Suite (VIIRS) instrument has superior resolution, radiometric accuracy, stability, and image quality. ACSPO VIIRS SST products show unprecedented accuracy, stability and global coverage, and are widely used in weather and climate applications at NOAA and worldwide. Reprocessing of the complete S-NPP VIIRS SST record from January 2012-pr is underway, to improve product stability and consistency. Four more VIIRS instruments are planned for launch onboard J1 to J4 from 2017-2026.

In 2016, NOAA plans to launch GOES-R, the first satellite in the new geostationary series. It will carry a superior Advanced Baseline Imager (ABI) onboard. The Japanese Himawari-8 (H8) satellite launched in October 2014 carries the Advanced Himawari Imager (AHI), a close proxy of ABI. NOAA has developed an experimental Level 2 SST product from H8 AHI (in original swath projection and with 10min sampling). However, the large data volume (45GB/day) makes the L2 product challenging to archive and use in various applications. Work is underway to generate a reduced-volume data product (Level 3, or Level 2 collated), and test it out and improve the processing chain end-to-end, to get ready for GOES-R launch.

Sea-surface temperature (SST) is considered by the Global Climate Observing System as an Essential Climate Variable. It is one of the most important characteristics of an aquatic system, helping to control the physical, biological and chemical environment. SST is also a variable that can be retrieved routinely from Earth Observation (EO), through measurements of radiation in the infrared and microwave portion of the electromagnetic spectrum. The coastal zone is one of the most treasured environments on our planet, but increasingly vulnerable to anthropogenic pressure and climate change. Satellite retrievals of SST at the coastline are complex and the accuracy is unknown, due to insufficient in situ datasets available to evaluate the satellite data. Traditional methods for deploying instrumentation in the coastal zone are expensive and hampered by challenges; for instance, from biofouling and wave breaking. Innovative solutions are required to build the in situ datasets on coastal SST needed to evaluate and improve EO products.

        We present results from a two-year study in which a small group of recreational surfers, based in the south west United Kingdom (UK), were tagged with temperature sensors and Global Positioning Systems (GPS) that they used when surfing. SST data collected by the surfers were compared with estimates from a local oceanographic station (~12km offshore, at the UK Western Channel Observatory) and with satellite data (Advanced Very High Resolution Radiometer). Satellite SST performed well at the local oceanographic station, but was less accurate at the coastline. Despite this, the satellite data was found to capture seasonal variations in SST data collected by the surfers, including sporadic changes in SST caused by tidal mixing, illustrating the potential of EO SST in the coastal zone for marine monitoring. In the UK alone, it has been estimated that surfers have the potential of acquiring 40 million measurements of SST per year. Considering surfing is a worldwide sport, tagging surfers with sensors could offer a cost-effective solution to collecting coastal in situ SST data needed to evaluate and improve current and future (e.g. Sentinel 3, SLSTR) EO products. Furthermore, our approach could be easily expanded to other popular marine recreational activities.

A major challenge in obtaining climate-quality sea surface temperature (SST) during the 1980s and early 1990s is the adverse situation for SST retrieval presented by the presence of stratospheric aerosol, following the massive eruptions of El Chichon (1982) and Mt Pinatubo (1991). We present the degree to which additional prior information about aerosol optical depth at thermal wavelengths can be sourced and the effect of exploiting such prior information in the context of SST retrieval.

Developing methodologies for ensuring good stability of SST observation under conditions of high stratospheric aerosol optical depth matters retrospectively, for the climate record. It will inevitably be relevant again at some point in the future, since eruptions with massive injections of sulfur into the stratosphere occur with a return period of a few decades. There is a high chance, for example, of such conditions prevailing at some point during the currently-planned operations of the Copernicus space segment.

Dual view radiometers such as the Sea and Land Surface Temperature Radiometer (SLSTR) and the earlier Along-Track Scanning Radiometers (ATSRs) have sufficient information (spectral and geometric) to make SST retrieval robust against stratospheric volcanic aerosol, without prior knowledge of stratospheric mass loading. The prior information in this case is limited to understanding of the variation in optical depth between different channels, which can be approximated. Prior to the era of dual view radiometers, SST retrieval relies on the Advanced Very High Resolution Radiometer series (AVHRRs), which are single view. Prior information on the spatio-temporal evolution of infra-red aerosol optical depth (IR AOD) is required. In the absence of IR AOD datasets at suitable resolution, this has been pursued with collaborators in the context of ESA's Climate Change Initiative (CCI) project on SST.

We present the degree to which infra-red optical depth information can be sourced by simple Kalman filtering of aerosol-sensitive estimates of stratospheric aerosol over oceans using clear-sky HIRS data. These observations are compared with stratospheric chemistry modelling results in the case of the Pinatubo eruption. The application of the IR AOD estimates to SST retrieval within an optimal estimation framework will be shown, and the effects in terms of SST bias, precision, stability and degree of prior influence will be presented.

We describe the rationale for our new diurnally adjusted SST analysis, the overall processing methodology, discuss the science, implementation and performance of the diurnal adjustment, and show examples of the output.  The analysis methodology employed is a multi-scale OI scheme which emulates Kalman filter behavior.  A non-stationary prior is mimicked by performing the analysis at three different length scales and interpolating the final result based on local data density.  This has the benefit of allowing high resolution where data coverage permits without introducing excessive noise.  Furthermore, it preserves mathematical rigor in terms of the anomaly correlation lengths.

The infrared satellite data are all processed from radiance to SST at NOAA, and a brief overview of the processing methodologies and error characteristics is given.  We also highlight the nature of each datatype, with particular emphasis on the benefits of geostationary data, which we will show are now of almost the same quality as that furnished by polar-orbiting sensors, especially for the most modern instruments.

The satellite observations themselves are individually adjusted to a foundation (a.k.a. bulk) temperature using a 1-d turbulence model forced with NWP fluxes of heat and momentum.  Non-local mixing is parameterized using operational wave model data in a Stokes' Drift formulation.  Residual bias corrections for each dataset (day and night separately for each platform) are estimated on a stochastic basis.

The diurnal correction parameterization has undergone extensive testing, and led to a number of optimizations in treatment of NWP winds and other model parameters.  Since these are of concern to many users, we will describe our findings and explain choices in some detail.  The other parameterization that we pay particular attention to is the nature of the insolation, and we find that additional information available in NWP fields is beneficial in improving its characterization.  Additionally, ocean colour information is found to be of some benefit in determining the in-water absorption profile.

As indicated above, comparisons of resultant analyses with and without the diurnal correction are made, including validation against in situ observations (buoy and Argo).  Contributions of various error sources will also be discussed.  We also evaluate the impact of sensor specific error statistics (bias and uncertainty) in the GHRSST L2P data product.

Another aspect of the presentation will be a brief discussion regarding SST retrieval algorithm sensitivity to diurnal excursions.  This important topic has recently become prominent within the satellite SST community, and a brief summary of discussions from the last GHRSST Science Team Meeting will be presented.

Climate change is one of the major challenges current society is facing. The scientific community is taking action to confront this challenge by understanding the mechanisms that drive climate change and by providing estimates how the climate will evolve in the future. A main need of the society is to have predictions of the climate in the near-term, at the intraseasonal-to-interannual timescales, to which many public sectors such as agriculture, energy production and human health are vulnerable. Predictions of the near climate involves not only the prediction of climate change but also and more challenging the prediction of its natural variability - commonly embraced under the term climate forecasting. Climate forecasting is a rapidly emerging field that is continuously improving and already starting to serve societal needs via climate services.

A key aspect in climate forecasting, commonly also embraced under the term seasonal-to-decadal forecasting, is the assessment of forecast skill based on retrospective predictions and observations of the past climate. Climate forecasting is only of value if we know how reliable the predictions are. Accurate observations of the past climate are therefore fundamental, yet observations are known to be imperfect. The consideration of observational uncertainty remains weakly explored by the modelling community. Practice of choosing an observational dataset is rarely guided by objective criteria and relies often on data accessibility and institutional proximity. Uncertainty estimates provided by the products are further often ignored, partly because of the lack of formal concepts on how to consider observational uncertainties in model verification.

In this study we explore the role of observational uncertainty in climate forecast verification. We present its relevance with respect to other sources of uncertainty and present how climate models can inform about the quality of observations. The evaluation is carried out using the ESA Climate Change Initiative (CCI) data products on sea-surface temperature (SST) and sea-ice concentration (SIC) and a broad set of climate forecasting systems. Our results demonstrate that forecast skill varies systematically depending on which dataset is selected which allows to determine observational noise from a model perspective. We also find that observational uncertainty is an important source of verification uncertainty that should receive more attention. Further efforts should be put into evaluation of observational error that can be propagated onto different spatio-temporal scales required by the modelling community.