Translation of the US Maury Collection: Time Adjustments

In the keyed data, there were many problems in the time elements (year,
month, day, and hour), stemming from difficult legibility problems in the
original microfilm records, and some keying or processing errors.  The time
elements were extensively edited, including spot comparisons against the
original microfilm images.  Day and hour were then adjusted to 24-hour
local time as described in Table 1.  The data in the Collection fall into
two form types: (1) once a day, and (2) more than once a day.  Within
those general types, four major logform types have been identified in the
Collection, as shown in Figs. 1-3 (form type 2) and Fig. 4 (form type 1).

Logforms such as Fig. 3 lack any indicator for noon or midnight.  The
rationale for the approach used in Table 1 in this case (pattern 7) is
based on two interrelated assumptions discussed below, with limited support
located thus far in the literature:

a) Assumption 1: The early maritime day ended at noon
Catchpole (1992) described 18th-19th century Hudson Bay Company ships'
logbooks, which at least would appear to resemble very closely the content
and organization of early US Maury logbooks:
     "The routine observations were time of day, speed in knots, depth in
     fathoms, course direction, wind direction, and weather.  The seaman's
     day ended at noon when the errors of navigation by dead reckoning were
     generally corrected by solar observation.  Hence midnight is in the middle
     of the page and noon at the bottom.  Beneath the noon entry, there was a
     space for recording the compass variation and navigational information.
     The latter usually included the course direction and distance run over
     the past 24 hours together with the latitude and longitude based on dead
     reckoning and, when available, celestial observations."
Note:  The above reference to "depth" in fathoms may be incorrect.  As
illustrated in Diaz and Woodruff (1999) (title page figure, from the US Maury
Collection, and p. 70), fathoms was typically used in early ship logbooks
as a "subunit of knots in measuring the amount of line run out with the
log to determine the ship's speed."  Any readings of knots and fathoms,
and 24-hour course and distance information, were not keyed from the US Maury
Collection.

However, relatively recent logforms in the US Maury Collection (e.g., on the
facing page of logforms such as Figs. 1-2) include the note:
     "Whether the day commences at noon or midnight, always call from noon
     to 8 P.M. "First Part."
This would appear to indicate that procedures for commencing and ending day
were mixed, at least by the date of these later logform types, and thereby
cast some doubt on Assumption 1.

b) Assumption 2: Noon-8pm was always "First Part"
Referring apparently to British naval logbooks (c. 18th-early 19th century),
Oliver and Kington (1970) state that "Daily entries were made in the journal
or log-book for the previous twenty-four hours...", further indicating
that the day was divided up as follows (Catchpole, 1992 includes a similar
breakdown of watches):
     Part of day    Watch       Hours
     ------------   -------     ---------------------------
     First          Afternoon   noon     - 4pm      (12-16)
                    Dog         4pm      - 8pm      (16-20)
     Middle         First       8pm      - midnight (20-24)
                    Middle      midnight - 4am      (00-04)
     Latter         Morning     4am      - 8am      (04-08)
                    Forenoon    8am      - noon     (08-12)

The first ("Begins"), middle, and latter parts of the day were written in
clearly by the observer in Fig. 3 roughly in accordance with this scheme,
and similarly in other examples that we have examined.  If it was true that
noon-8pm was always "first part," and similarly for the other parts, then
this positioning would tend to support Assumption 1.

A further question concerning much of the data in which day ends or was assumed
to end at noon, is to which date in the modern calendar system the morning and
afternoon reports should be assigned.  According to the procedure outlined in
Table 1, Fig. 3 dated 15 June has its afternoon (morning) reports assigned to
the 14th (15th).  The alternative would be to assign the afternoon (morning)
reports to the 15th (16th).

Due to all these uncertainties, and the possibility of mixed procedures used
by early mariners, day and hour must be regarded as approximate, with probable
uncertainties of at least +/- one day, and +/- 12 hours.


Table 1.  Adjustment of day and hour to 24-hour local time of US Maury Collection
form type 2 (more than one report per day).  The hours and days in this portion
of the Collection were categorized into seven patterns (pattern 6, not listed
in the table, referred to the 24-hour result).

For each pattern, the following information is given, followed by discussion:
     approximate frequency: percentage of total of form type 2 reports
     (546,138 reports, or 39% of total Collection)
     log:  The hours as represented on the original logforms.  Afternoon
           (pm), morning (am), midnight (Mn), or noon (Nn) are marked for
           reference above these hours; for pattern 7 these are assumed.
     key:  The hours as keyed; asterisk (*) marks the hour of the report
           in which location was keyed.
     dHr:  The adjustment to be made to hour, or blank if no adjustment.
      Hr:  The resultant hour.
     dDy:  The adjustment to be made to day, or blank if no adjustment.

Patterns 1-4 and 7 all arise from logforms in which the day starts (or is
believed to start) after noon, i.e., following the early maritime definition
of day ending at noon.  Pattern 5 appears to arise from the same logform type
as patterns 3-4, but it differs because the day started after midnight in the
digitized data.

Note:  Midnight in the time-adjusted output will be represented as 24 of
day-1.  It should be adjusted to 00 of day (and month, year, if applicable),
before adjustment to UTC.
===============================================================================
Pattern: 1 
  approximate frequency: 48%
        pm pm am am Nn
  log    3  8  4  9 12  (5 reports per day)
  key   03 08 04 09 12*
  dHr  +12+12
   Hr   15 20 04 09 12 
  dDy   -1 -1
Discussion:  Logform example: Fig. 1 (the "Nightingale").  Noon is printed on
this type of logform, with a corresponding line on which to write lat/lon, and
the hours of 8pm, 4am, and noon printed in bold face.  Instead of starting with
the first hour of the maritime (noon-to-noon) day, the page starts (ends) with
the first hour after (before) midnight (i.e., 4am and 8pm).  The data were not
keyed according to the 24-hour clock.  Referring to Fig. 1, the first seven
lines of time/space location data were keyed as follows, with the proposed
mapping of day and hour ("Adj") shown to the right (there was no report keyed
at hour 03 in this example):
     voyage   year  mo  dy  hr lat.  long.    Adj dy  hr   Alt dy
     -------------------------------------    ----------   ------
     4800481  1854  06  15  04                    15  04       14
     4800481  1854  06  15  09                    15  09       14
     4800481  1854  06  15  12 437N  2700W        15  12       14
     4800481  1854  06  16  08                    15  20       14
     4800481  1854  06  16  04                    16  04       15
     4800481  1854  06  16  09                    16  09       15
     4800481  1854  06  16  12 419N  2643W        16  12       15  
Note:  The day in Fig. 1 is written immediately above noon (e.g., "15th").
We have been unable to determine how the noon-to-noon convention relates
to the modern midnight-to-midnight convention.  We are shifting day forward
such that the 15th ends at midnight, rather than noon.  However, we might
be making the wrong assumption if the writing refers to the start of the
day, e.g., the 15th.  The "Alt" column reflects an alternative (day-1 for
morning hours) mapping.  The data might also be a mixture of conventions
if mariners did not follow consistent procedures.
-------------------------------------------------------------------------------
Pattern: 2
  approximate frequency: 25%
        pm pm am am Nn
  log    3  8  4  9 12 (5 reports per day)
  key   15 20 04 09 12*
  dHr    
   Hr   15 20 04 09 12
  dDy   -1 -1
Discussion:  Logform example: Fig. 1 (the "Nightingale").  Same as pattern 1,
except the data were keyed according to the 24-hour clock.
-------------------------------------------------------------------------------
Pattern: 3
  approximate frequency: 4%
        pm pm pm pm pm pm Mn am am am am am am Nn
  log    2  3  4  6  8 10 12  2  4  6  8  9 10 12 (14 reports per day)
  key   02 03 04 06 08 10 12 02 04 06 08 09 10 12*
  dHr  +12+12+12+12+12+12+12
   Hr   14 15 16 18 20 22 24 02 04 06 08 09 10 12
  dDy   -1 -1 -1 -1 -1 -1 -1
Discussion:  Logform example: Fig. 2 (the Warren).  Noon is explicitly labelled
on this logform type, with the hours of 3pm, 8pm, 4am, 9am, and noon printed
in bold face.  Instead of starting with the first hour of the maritime
(noon-to-noon) day, the page starts with the first hour after midnight (2am)
and ends with midnight.  The data were not keyed according to the 24-hour
clock.  Referring to Fig. 2, the first 15 lines of time/space location data
were keyed as follows, with the proposed mapping of day and hour ("Adj") shown
to the right:
     voyage   year  mo  dy  hr lat.  long.    Adj dy  hr   Alt dy
     -------------------------------------    ----------   ------
     5201681  1859  07  24  02                    24  02       23
     5201681  1859  07  24  04                    24  04       23
     5201681  1859  07  24  06                    24  06       23  
     5201681  1859  07  24  08                    24  08       23
     5201681  1859  07  24  09                    24  09       23
     5201681  1859  07  24  10                    24  10       23
     5201681  1859  07  24  12 815N 10954W        24  12       23
     5201681  1859  07  25  02                    24  14       23
     5201681  1859  07  25  03                    24  15       23 
     5201681  1859  07  25  04                    24  16       23
     5201681  1859  07  25  06                    24  18       23
     5201681  1859  07  25  08                    24  20       23
     5201681  1859  07  25  10                    24  22       23
     5201681  1859  07  25  12                    24  24       23
     5201681  1859  07  25  02                    25  02       24
Note:  Similarly to pattern 1, an alternative (day-1 for morning hours) mapping
of day to midnight-to-midnight is shown to the right ("Alt").  However, this
looks unlikely in this example because the date (e.g., "July 24th") is written
to the left of hour at around midnight.
-------------------------------------------------------------------------------
Pattern: 4
  approximate frequency: 5%
        pm pm pm pm pm pm Mn am am am am am am Nn
  log    2  3  4  6  8 10 12  2  4  6  8  9 10 12 (14 reports per day)
  key   14 15 16 18 20 22 24 02 04 06 08 09 10 12*
  dHr  
   Hr   14 15 16 18 20 22 24 02 04 06 08 09 10 12
  dDy   -1 -1 -1 -1 -1 -1 -1
Discussion:  Logform example: Fig. 2 (the "Warren").  Same as pattern 3, except
the data were keyed according to the 24-hour clock.
-------------------------------------------------------------------------------
Pattern: 5
  approximate frequency: 6%
        am am am am am am Nn pm pm pm pm pm pm Mn
  log    2  4  6  8  9 10 12  2  3  4  6  8 10 12 (14 reports per day)
  key   02 03 04 06 08 10 12*02 04 06 08 09 10 12
  dHr                       +12+12+12+12+12+12+12
   Hr   02 03 04 06 08 10 12 14 15 16 18 20 22 24
  dDy
Discussion:  Logform example: Fig. 2 (the "Warren").  Similar to pattern
3, except the data were keyed with day running from midnight-to-midnight
(no day-change needed, but the data must be transformed to the 24-hour
clock).  This voyage of the "Warren" is an example of a voyage in which
mixed procedures were used to key the data (some days keyed according
to pattern 3 as shown above, and others according to pattern 5).  We
fixed some of these inconsistencies during time editing (e.g., in this
voyage of the Warren, all days were changed to pattern 5).
-------------------------------------------------------------------------------
Pattern: 7
  approximate frequency: 11%
  log  (not explicitly associated with hour; filled in at bottom of page)
  key  00*
  dHr +12
   Hr  12
  dDy
       pm pm pm pm pm pm pm pm pm pm pm Mn am am am am am am am am am am am Nn
  log   1  2  3  4  5  6  7  8  9 10 11 12  1  2  3  4  5  6  7  8  9 10 11 12
  key  01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
  dHr +12+12+12+12+12+12+12+12+12+12+12+12-12-12-12-12-12-12-12-12-12-12-12-12
   Hr  13 14 15 16 17 18 19 20 21 22 23 24 01 02 03 04 05 06 07 08 09 10 11 12
  dDy  -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1

Discussion: Logform example: Fig. 3 (the "Restitution").  Noon is not
explicitly labelled on this logform type, but day is assumed to end at noon
for reasons as discussed in the text.  The latitude/longitude and magnetic
declination were keyed from the bottom of each logform into a report appearing
at the beginning of each day-sequence, assigned for unknown reasons to hour 00.
Referring to Fig. 3, the hour 00 report for June 15, and the time information
from the five other reports for that day, were keyed as follows, with the
proposed mapping of day and hour ("Adj") shown to the right:
     voyage   year  mo  dy  hr lat.  long.    Adj dy  hr
     -------------------------------------    ----------
     7500011  1815  06  15  00 3647S 3329E        15  12
     7500011  1815  06  15  01                    14  13
     7500011  1815  06  15  12                    14  24
     7500011  1815  06  15  17                    15   5
     7500011  1815  06  15  20                    15   8
     7500011  1815  06  15  23                    15  11
Note:  As a result of this adjustment, the hour 00 report containing latitude,
longitude, and magnetic declination is out of sort-order.  This could be
important if magnetic declination is needed for example to adjust the wind
directions for the other reports for 15 June.
-------------------------------------------------------------------------------

Figure captions:
Figure 1. Sample Abstract log from Ship "Nightingale" (15-23 June 1854;
voyage number 4800481).

Figure 2. Sample Abstract log from Ship "Warren" (24-26 July 1859; voyage
number 5201681).

Figure 3. Sample logforms from Ship "Restitution" (15-16 June 1815; voyage
number 7500011).

Figure 4. Sample Abstract log from Ship "Kensington" (form type 1;
27 November-23 December 1848; voyage number 2200031).

Figure 5. US Maury Collection reports (with location) before location quality
control (QC).

Figure 6. Same as Fig. 5 but after location QC.

Figure 7. US Maury Collection logform types (colors) in terms of numbers
of reports per day (daily, 5 reports/day, 14 reports/day, and hourly).
The total period of record for the Collection is 1784-1863.

References
Catchpole, A.J.W., 1992: Hudson's Bay Company ships' log-books as sources of
sea ice data, 1751-1870.  Climate Since AD 1500.  R.S. Bradley and P.D. Jones,
Eds., Routledge, London and New York.

Diaz, H.F., and S.D. Woodruff (Eds.), 1999: Proceedings of the International
Workshop on Digitization and Preparation of Historical Surface Marine Data and
Metadata (Toledo, Spain, 15-17 September 1997).  WMO/TD-No.957, MMROA Report
No. 43, World Meteorological Organization, Geneva, 114 pp.

Oliver, J., and J.A. Kington, 1970: The usefulness of ships' log-books in
the synoptic analysis of past climates.  Weather, 25, 520-528.

[Translation information][US Maury Collection]