MSU equivalent measures from HadAT

Summary

NEW - Channel 3 equivalents for HadAT2 are now available

In the interests of having MSU-equivalent HadAT timeseries available for immediate comparisons with these satellite-derived data, we have created here a set of MSU equivalent measures under very specific criteria. Although justifiable, many of the decisions we have made are subjective and alternative choices could be made that would appear, at least initially, to be equally plausible. We hope to find the time in future to adequately resolve the sensitivity to these choices and come up with a more robust "how to". This page describes the methodology we have employed and includes final MSU equivalent measures for our monthly HadAT2 timeseries as well as a number of intermediate products required to produce these from the Temperature on Pressure Levels data.

Far more information on the real MSU instruments, their measurements and their climate timeseries are available from the University of Alabama in Huntsville and Remote Sensing Systems (external links). There have been additional efforts by other groups to create climate datasets from MSU but, to date at least, the groups involved have not released their data freely to the research community. As and when this is no longer the case we will also link to the data and websites for these groups from here.

Creating gridbox resolution weighting fields

Before creating fully gridded products we felt it necessary to calculate geographically variant weighting fields taking into account the local surface pressure and characteristics. We utilise the static weighting functions on a 5 hPa resolution provided by the University of Alabama in Huntsville group that have been used in numerous scientific studies to date. These have very slightly different weightings over land and ocean for channel 2 and channel 2LT reflecting the different average emissivity characteristics of these surfaces, and for all channels assume a standard atmospheric profile. Of course, the reality is significantly more nuanced than this and effectively these weighting functions provide investigators with just enough rope to go and hang themselves with. Because we have only temperature data we do not attempt a full radiative reconstruction which would additionally require information on the vertical distribution of at least water vapour, and possibly other trace gases. At least for large-scale averages the simple approach used here has been shown to be adequately similar to a full radiative-transfer code.

1. We used annual mean surface pressure fields from a run of HadAM3 (the Hadley Centre's atmospheric model) run with observed Sea Surface Temperatures to define the local surface pressure on the model resolution (3.75 degrees in longitude, 2.5 degrees in latitude). For our current analysis these fields are set as seasonally invariant, which may introduce a small systematic error if the true seasonal cycle in surface pressure is large, particularly in our MSU channel 2LT equivalents.
2. The model land/sea mask is used to define the surface type. We augment this by using the 30-year climatological values of sea-ice from HadISST and defining any gridbox with greater than 50% sea-ice coverage in any given month as "land" surface type. This leads to small seasonal variations in polar regions.
3. For each model gridbox:
   • The lowest valid HadAT pressure level is defined as the first HadAT pressure level above the HadAM3 derived surface pressure estimate.
   • The surface value is given the sum of weightings from the standard atmosphere surface to half way in pressure coordinates between the "actual" surface pressure and the first valid pressure.
   • The first valid pressure is given the sum of weights from this point to half way in pressure coordinates to the next pressure level.
   • Up to the final 30hPa level the weightings are the sum of weights between mid-points in pressure coordinates (not distance) for consecutive levels.
   • For 30hPa the integral continues to the top of the atmosphere.
4. Finally we bilinearly interpolated these weightings fields to the coarser HadAT (10 degrees in longitude, 5 degrees in latitude) grid and calculated seasonal as well as monthly variant weighting functions.

Calculating MSU equivalent HadAT timeseries

To create gridded timeseries we use collocated observations from HadCRUT3 to provide our surface observations and HadAT2 to provide the upper-air observations. Ideally, we would use skin temperature rather than near-surface temperature as it is the skin temperature that the satellite instrument is sensing. On short timescales these measures can be very different, but on the monthly and longer timescales considered here they should exhibit similar variability and trends in the absence of substantial changes in land cover/land use and sea ice extent. Any MSU channel equivalent measure is calculated only if the sum of the weights that can be applied (non-missing data) is greater than 80%. This means that there are less data for Channel 4 (stratospheric channel) than either 2 or 2LT, particularly early in the record when balloons were more prone to burst. No criteria have been applied as to the maximum number of consecutive levels that can be missing or any other criteria that could be used to refine our estimates. If over time there have been changes in reporting performance then there may be spurious trends introduced even if the HadAT data themselves are homogeneous so investigators must take great care in utilising these data.

As of July 2009 we have included Channel 3 equivalents for HadAT2 in the files below.

Maintained by: Holly Titchner Last updated: 10/05/2010 Hosted by: Met Office Hadley Centre

© Crown Copyright