Cross Validation

Cross validation compares observations against other nearby data. A variety of methods are used to make these comparisons. The most common approach is to perform an optimum interpolation (OI) analysis at the observation location and sampling time using nearby validated data, excluding the datum being checked. The innovations

Fig. 4.2 Geographic coverage charts and histograms of AMSR-E andMETOP LAC retrieved SST minus global (red) and regional (green) analysis and climate (blue) backgrounds. The AMSR-E data cut processed 1,369,870 observations on 28 Dec 2009 at 18Z. The METOP LAC data cut processed 2,281,094 observations on 10 SEP 2009 at 01Z. Daytime retrievals are indicated as blue and nighttime retrievals are indicated as green points in the geographic coverage charts. The histograms are formed using 0.25°C temperature difference bins

Fig. 4.2 Geographic coverage charts and histograms of AMSR-E andMETOP LAC retrieved SST minus global (red) and regional (green) analysis and climate (blue) backgrounds. The AMSR-E data cut processed 1,369,870 observations on 28 Dec 2009 at 18Z. The METOP LAC data cut processed 2,281,094 observations on 10 SEP 2009 at 01Z. Daytime retrievals are indicated as blue and nighttime retrievals are indicated as green points in the geographic coverage charts. The histograms are formed using 0.25°C temperature difference bins for the cross validation are computed from an ocean climatology. It is important to ensure that cross validation checks are data driven and independent of any analysis or forecast model backgrounds. The uncertainty of the analyzed value is computed from the OI analysis error reduction of climate variability. The cross validation analyzed value and its uncertainty are then used as the background and background error values in the background field check described in Sect. 4.1. In the absence of any nearby valid data, the cross validation procedure simply returns climate and climate variability as the analysis and error estimates, and the cross validation check is identical to the background check using climatology. Thus, cross validation is analogous to checking observations against a dynamic, time-dependent climatology.

The background error covariances used in the cross validation procedure can be very simple, such as only including data within some specified distance from the observation being checked, or more complicated, based on the multivariate covariances used in the assimilation procedure itself. Cross validation can be applied to all observation data pairs in the quality control or it can be preceded by other data checks which first detect suspect observations. The cross validation is then performed only on the suspect observations to save on computational time. In data sparse areas the cross validation check will have limited effect. However, the continuing development of the Argo profiling float array generally provides an adequate number of nearby data to allow the cross validation of profile observations to work well in practice. Cross validation is also useful in the quality control of altimeter SSH and SWH observations, since individually those data tend to be rejected along sequential segments of altimeter tracks due to phase errors in the model background fields.

Was this article helpful?

0 0
Guide to Alternative Fuels

Guide to Alternative Fuels

Your Alternative Fuel Solution for Saving Money, Reducing Oil Dependency, and Helping the Planet. Ethanol is an alternative to gasoline. The use of ethanol has been demonstrated to reduce greenhouse emissions slightly as compared to gasoline. Through this ebook, you are going to learn what you will need to know why choosing an alternative fuel may benefit you and your future.

Get My Free Ebook


Post a comment