The extended records (files with an "ext" qualifier) are comprised of smoothed values, and interpolated and extrapolated values defined at each time step of the synchronization period. Those who wish to use extended records in their modeling application must simply add the reference MBL vector (COLUMN 3) to the difference climatology (COLUMN 4), i.e., extended record = REF + diff. Users will notice that S(t) = REF + diff wherever smoothed values (COLUMN 2) exist.
You may also choose to use only the smoothed values (COLUMN 2) from the sites that are synchronized which will have assigned default values where there are no measurements.
PLEASE NOTE: Occasional discontinuities at the transition between smoothed values and extrapolated values may be significant in certain modeling applications. These occur when values derived from data extension techniques (based on average behavior) join actual measurements that depart from average behavior. Discontinuities may occur at either end of the smoothed measurement record.
PLEASE NOTE: Discontinuities within periods of interpolated or extrapolated values may occur when MBL measurement records begin, end, or are interrupted for long periods of time. Some discontinuities may be significant in certain modeling applications. Serious discontinuities are identified below.
|1984.208333||5°N||ESRL sampling program at Christmas Island, Kiribati (CHR) begins|
|1985.354167||28°N||ESRL sampling program at Sand Island, Midway (MID) begins|
|1987.000000||25°S||ESRL shipboard sampling in Pacific Ocean (POC) begins|
|1989.125000||32°N||ESRL sampling program at Bermuda (BME, BMW) begins|
|1998.187500||43°N||ESRL sampling program at Cape Meares, Oregon (CMO) ends|
|2008.833333||32°N||ESRL sampling gap at Bermuda East (BME)|
PLEASE NOTE: The data extension procedure requires at least 2 years of observations.
Relative weighting of each value in an extended record can be important because some points are better determined than others. Confidence in the smoothed values depends on the density of the data, the relative occurrence of rejected data, the "scatter" in the data, the type and number of corrections applied, and the length of the measurement period. Masarie and Tans  describe in detail the relative weighting scheme and provide an example of how extended records and relative weights have been used in a 2-D modeling application. Users may choose to ignore our weighting scheme; sufficient information is included in the weight files so that users may devise their own weighting scheme.