reSeArCH Letter
HadSST3, for example, represents transitions between wooden buck-
ets and less-insulated canvas buckets by using globally uniform, linear
weights^4. Another method assumes that differences between SST and
night-time marine air temperatures remain similar to an average for
1971–2000 (ref.^1 ). Simple assumptions regarding bucket corrections
are commensurate with the limited metadata available for determining
observational characteristics^2 ,^4 ,^7 , but are questionable on the basis that
ICOADS contains observations from a wide variety of ships that sailed
for different purposes and sampled the ocean unevenly^22 (Fig. 1 b–d).
More comprehensive bucket corrections are made possible by calcu-
lating offsets between groups of SST measurements that are nearby in
space and time^10. Specifically, we difference bucket SST measurementsYearNumber of bucket measurements (1910 1915 1920 1925 1930 1935 1940×^105 )024681012a020406080100Percentage of measurements from bucketsGB DCK 184
GB DCK 201
GB DCK 202
GB DCK 203
GB DCK 204
GB DCK 245
GB DCK 249
GB DCK 706
NL DCK 193
NL DCK 706
JP DCK 118
JP DCK 706US DCK 705
US DCK 706
US DCK 707
DE DCK 151
DE DCK 192
DE DCK 215
DE DCK 720
RU DCK 732
RU DCK 735
---- DCK 155
---- DCK 156
Other groups
Non bucketsb
60° N
30° N
0°
30° S
60° S
90° E 180° W 90° W0° 90° E 180° W 90° W0° 90° E 180° W 90° W0°c dFig. 1 | Groups of bucket SST measurements. a, Left-hand y-axis: number
of bucket SST measurements from individual groups identified by country
and deck information in ICOADS3.0. Country name abbreviations are:
DE, Germany; GB, Great Britain; JP, Japan; NL, The Netherlands; RU,
Russia; US, United States; and —, missing. Groups having fewer than
100,000 measurements are labelled as ‘other groups’. Decks 118 and 762
are combined into ‘JP DCK 118’ because they are both Japanese Kobe
Collection decks. Right-hand y-axis: percentage of measurements that
have come from buckets, showing that nearly all observations before 1935
are from buckets (black line). b–d, Maps indicating nations that contribute
the most observations within 5° × 5° grids for the periods 1908–1918 (b),
1919–1928 (c) and 1929–1941 (d). White grid boxes have fewer than three
years of data.---- Dck 197RU Dck 735
GB Dck 230 DE Dck 721NL Dck 705 LR Dck 888DE Dck 792 BE Dck 792DE Dck 555 JP Dck 706
DK Dck 927 DE Dck 927GB Dck 555 NL Dck 706---- Dck 888
GB Dck 221 BE Dck 927DK Dck 888 JP Dck 888NL Dck 193 NL Dck 150---- Dck 156 NL Dck 792
AU Dck 900 IS Dck 926EE Dck 888* DE Dck 720 BE Dck 888
NL Dck 992 JP Dck 555* JP Dck 118 US Dck 116* DE Dck 151 JP Dck 926
CN Dck 706 NL Dck 888DK Dck 926
* NL Dck 927 NL Dck 555* ---- Dck 155 JP Dck 705
* GB Dck 249 CN Dck 792* NL Dck 926 EE Dck 555* GB Dck 245 IS Dck 927
* IE Dck 927
* US Dck 710 * US Dck 704* GB Dck 205 * US Dck 926* BR Dck 926 * IE Dck 926SG Dck 792ZA Dck 927RU Dck 732 PT Dck 927
US Dck 927 SP Dck 706UA Dck 888 DE Dck 192---- Dck 203 ---- Dck 926GB Dck 707 FR Dck 926
GB Dck 792 GB Dck 888GB Dck 202
CA Dck 792 RU Dck 792DN Dck 706 FR Dck 888HO Dck 705 NZ Dck 926
GB Dck 204 ---- Dck 927RU Dck 555 US Dck 792GB Dck 203
GB Dck 705 FR Dck 927GB Dck 201 ---- Dck 992RU Dck 888 BE Dck 926
CA Dck 926 DE Dck 215GB Dck 184 JP Dck 927JP Dck 898CA Dck 888 PA Dck 992
GB Dck 992 HK Dck 927DK Dck 792 GB Dck 927DE Dck 926 SE Dck 926---- Dck 792 DE Dck 992
FR Dck 792
GB Dck 152 US Dck 701CN Dck 781 DE Dck 888
RU Dck 185 GB Dck 926** PM Dck 707**US Dck 281
** US Dck 707 * MK Dck 927** FR Dck 706 * IL Dck 926
** US Dck 706 * PL Dck 927* MY Dck 927 * IN Dck 927* AR Dck 927 * IL Dck 927* MY Dck 792 * ES Dck 927
* HR Dck 926* NO Dck 702 * IN Dck 926
* NO Dck 992 * HO Dck 707** NO Dck 706 * US Dck 705* NO Dck 926 * SG Dck 926* NO Dck 927 * PL Dck 926
* IY Dck 706* ---- Dck 255 * SE Dck 927
* NO Dck 792 * HO Dck 706US Dck 555 ZA Dck 888* BX Dck 706 NO Dck 700
* LR Dck 792 PT Dck 926* DL Dck 706
RU Dck 849 NO Dck 888HK Dck 926 UA Dck 792SG Dck 888 UA Dck 555GB Dck 849 SG Dck 927* GB Dck 706 NZ Dck 927
CA Dck 927
RU Dck 927 PA Dck 792RU Dck 926 CA Dck 992EE Dck 792 JP Dck 7921850 1900 1950 2000 1850 1900 1950 2000 1850 1900 1950 2000SST offsets between groups (°C)–0.5 –0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 0.5 10 102 103 104 105 106
Number of annual SST measurementsFig. 2 | Relative offsets between groups of bucket SSTs. Groups for which
fixed effects differ significantly from zero are indicated by * (P < 0.05).
The 46 out of 162 groups that contribute data between 1908 and 1941 are
indicated in black, and those remaining significant after a Bonferroni
correction (P < 0.05/46) are indicated by **. Shading indicates the sum of
fixed and five-yearly effects (regional effects are not shown). Bar widths
indicate the number of SST measurements contributed by each group for
each year. Abbreviated country names correspond to those in Fig. 1 and
are listed in Supplementary Table 1.394 | NAtUre | VOL 571 | 18 JULY 2019