\documentclass[12pt]{article}

\usepackage{epsfig}
\usepackage{ametsoc}

\setlength{\headheight}{0cm} \setlength{\headsep}{0cm}
\setlength{\textheight}{23.7cm} \setlength{\textwidth}{16cm}
\setlength{\oddsidemargin}{0cm} \setlength{\evensidemargin}{0cm}
\setlength{\topmargin}{0cm} \setlength{\topskip}{0cm}
\setcounter{secnumdepth}{5}

\renewcommand{\baselinestretch}{2.0}

\catcode`\"=\active \let"=\"
\let\3=\ss

\begin{document}

\title{\bf New observations of old weather}
\author{{Philip Brohan 
         \thanks{\textit{Corresponding author address:}
         Philip Brohan, Met Office Hadley Centre for Climate Change, Fitzroy Road, Exeter, UK.
         \newline{E-mail: philip.brohan@metoffice.gov.uk}}}\\
Met Office Hadley Centre for Climate Change, Exeter, UK.
\and
Rob Allan\\
Met Office Hadley Centre for Climate Change
\and
J. Eric Freeman, Sourcecorp/NCDC
\and
Anne M. Waple, STG Inc./NCDC
\and
Dennis Wheeler\\
Department of Geography, University of Sunderland
\and
Clive Wilkinson\\
Climatic Research Unit, University of East Anglia
\and
Scott Woodruff\\
Earth Systems Research Laboratory, NOAA
}

\amstitle

{\bf Capsule summary:} Historical ships' logbooks contain vast numbers of undigitised marine observations, of great potential value for climate research --- as illustrated by the rescue of 1.5 million observations around World War II.

\begin{abstract}

Old weather observations form the basis of our historical records of climate and the manner in which it has changed over time. Such observations are vital for climate change monitoring and prediction. For the world's oceans, there are many meteorological and oceanographic observations available back to the mid-twentieth century, but coverage is limited in earlier periods, and particularly also during the two world wars. In the nineteenth century before 1850 there are currently very few instrumental observations available. Consequently, detailed observational estimates of surface climate change can be made only back to the mid-nineteenth century. To improve and extend this otherwise inconsistent early coverage, scientists need to find more observations from these periods. Fortunately, many such observations exist, in logbooks, reports and other paper records, but their inclusion in the climatic data sets requires that these paper records be abstracted from the world's archives, digitised into an electronic form, and blended into existing climate databases.

As a first step in this direction, selected Royal Navy logbooks from the period 1938 to 1947, kept in the UK National Archives, have been photographed and digitised. These have provided more than 1,500,000 new observations for this period, and a preliminary analysis has shown significant improvements to the record of climate change in the mid-twentieth century.

\end{abstract}

\section{Introduction}

Our understanding of climate change and climate variability depend critically on historical weather observations. Studies of recent changes use the instrumental observations directly (\cite{trenberth07ipcc}), while studies of longer term changes make use of climate proxies which are calibrated to the historical observations (\cite{jansen07ipcc}). The currently-available observations allow some surface climate fields to be reconstructed for the last 150 years (figure \ref{f3}).
\begin{figure}
\begin{center}
  \noindent\includegraphics[width=40pc,angle=0]{./figures/hadsst2_global}\\
  \caption{Global mean sea-surface temperature changes (relative to 1961-90 average) estimated from in-situ observations. (From the HadSST2 dataset: \cite{rayner06hadsst2}).}
  \label{f3}
\end{center}
\end{figure}

Weather observations including measurements of temperature and pressure have been made by ships, in large numbers, back to the late eighteenth century; and are recorded in the ships' logbooks kept during the voyages. Observations containing subjective but reliable (\cite{wheeler05logbook}) estimates of wind force and direction are available even further back. National programs to digitise marine observations from some logbooks (and later specialised meteorological forms) appear to have started around the turn of the twentieth century with the invention of Hollerith punched cards. By the mid-century there were international exchanges of large national collections on card `decks' \citep{woodruff87,woodruff05}, and these developed ultimately into today's international scheme for routine digitisation and exchange of delayed-mode Voluntary Observing Ship (VOS) data. In spite of these early efforts, many as-yet untapped log-books still exist, in various archives around the world.

Unfortunately, some of the early observation keying efforts resulted in incomplete and possibly inaccurate data and metadata, owing to early technological limitations \citep{woodruff05}. This makes new digitisation programs even more valuable: if the missing or deficient observations were digitised and made available to climate researchers, they would offer the possibility of significantly extending our knowledge and understanding of climate change.

Until recently, endeavours to digitise the missing observations in the UK archives  have been confined to a number of case studies using logbooks to reconstruct the weather of naval battles (\cite{wheeler01trafalgar}), or large volcanic eruptions (\cite{chenoweth01}), or to examine the historical dimension of specific phenomena such as the Southern Oscillation (\cite{garcia01}), or in specific areas (\cite{farrington98}). In the latter context the work of \cite{catchpole92} is exemplary and uses the logbooks of ships of the Hudson's Bay Company to study ice cover in the nineteenth century across the seas of northern Canada. More comprehensive have been the recent seminal efforts of the EU-funded CLIWOC (Climatological Database for the Worlds Oceans 1750 to 1850) program. This program was concerned specifically with testing the utility of non-instrumental logbook data (principally wind force and direction) and of developing methods and procedures in which such archaic information could be used in a modern-day scientific context (\cite{wheeler06cliwoc}). A database of some 300,000 quality-controlled and processed data was developed and is available at {\tt http://www.ucm.es/info/cliwoc}.

Subsequent to the CLIWOC program, an international project for RECovery of Logbooks And International Marine data (RECLAIM, {\tt http://icoads.noaa.gov/reclaim}) has recently begun to draw attention to, and try to recover, some of the vast quantity of logbook data that remains untouched in the archives. This project has begun to identify the archival sources where there are data worth extracting; and has obtained funding for, and completed, a project to recover observations from one source: Royal Navy logbooks for the period 1938 to 1947.

\section{UK Observations available for digitisation}

Historical marine observations are archived in many parts of the world, but particularly in the UK, where it is estimated that over 300,000 undigitised logbooks, mostly from Royal Navy vessels but with some from other sources, are available for the period 1670 to the late twentieth century (figure \ref{f1}).
\begin{figure}
\begin{center}
 \noindent\includegraphics[width=40pc,angle=0]{./figures/Graph_of_numbers}\\
  \caption{Estimated number of undigitised logbooks in the main UK archives (The National Archive, the National Maritime Museum, and the British Library).}
  \label{f1}
\end{center}
\end{figure}
 The near-global extent of the British Empire had the inevitable consequence that British ships sailed extensively across the world's oceans, each making its daily logbook record as it went. While some UK marine data have been available in digital form for decades \citep{woodruff05}, it is perhaps remarkable that hitherto there has been no systematic UK or international program for digitising more of these potentially valuable historical observations, and most remain unused. This focus on UK collections should not be interpreted as diminishing the importance of similar collections: in particular in France, the Netherlands, Spain, Sweden and the USA; but the UK appears to have by far the largest such collection \citep{wheeler06cliwoc}.
 
Investigations have indicated that the majority of the many thousands of undigitised UK logbooks are to be found in just three major collections: the National Archives (Kew), the British Library (Central London), and the National Maritime Museum (Greenwich). It must, however, be recognised that this unique collection of logbooks resides in several forms. Until the early nineteenth century each senior ship's officer of the Royal Navy carried the responsibility of keeping his own record, giving rise to captains', masters', and lieutenants' logbooks. The general ship logbook became commonplace only as the century advanced. In addition to Royal Navy records, there also exists an important collection of logbooks of ships in the service of the Honourable East India Company. These, though fewer in number, and dating only from the 1600's to the 1840s (\cite{farrington99}), are important in their valuable coverage, often instrumental (after about 1780), of the southern and Indian Oceans as they made passage to and from India and China via the Cape of Good Hope. Figure \ref{f1} summarises the distribution of these logbooks by category and decade.

Another Significant source of historical observations is the Met Office archives, which contain a large number of meteorological logbooks (``Met logs''), in some cases from the same Royal Navy ships whose ships' logbooks are in the national archives, but apparently more rigorously observed and thoroughly documented. To an limited extent these Met logs have already been digitised and blended into ICOADS, but the digitisation process only included a subset of the observations in the paper logs, and also did not capture the valuable metadata (for example, instrument positions and calibration details) recorded in them.

Elsewhere in the UK only small numbers of logbooks are known to be held, and few historical UK merchant marine logbooks appear to have survived. Most important amongst these are the logbooks of whaling vessels from the late eighteenth and early nineteenth centuries. Whaling activities were concentrated along the ice limits around Greenland. As such, whilst their numbers are small (only about 250), they contain highly valuable climatic information for the high northern latitudes at a time for which no other sources are available. The majority of these documents are to be found in the Kingston-upon-Hull City archives (\cite{wheeler07arctic}). Similarly notable collections are to be found in the archives of New England whaling ports. These have been little explored but thoroughly catalogued (\cite{sherman86}). 

In the age of sail, weather information was an important element in the task of navigation; and, not surprisingly, the vast majority of the logbooks contain useful daily weather observations. Before the mid-nineteenth century, logbooks were of a mostly non-instrumental character concerned with wind force and direction, an example of such a logbook is given in appendix \ref{rattlesnake}. These observations, derived by procedures not far removed from those employed today by participants in the VOS network, can be retrieved and used, as shown by the CLIWOC project (\cite{wheeler06cliwoc}). More directly valuable, however, are the logbooks from later times containing instrumental (thermometer and barometer) observations. Instrumental data became widespread only following the Brussels Maritime Conference of 1853 \citep{maury54}. Before that date the East India Company logbooks, which number about 4000 volumes, provide the only consistent source of marine instrumental data, many of them being daily barometer and thermometer observations going back to before 1800. Appendix \ref{hindostan} gives examples from an East India Company logbook. 

The digitisation of all the UK data would be a monumental undertaking, and no resources are currently available to accomplish this task. Nevertheless, a pilot project has been completed. that pays particular attention to logbooks from the period of the Second World War (WW2) and demonstrates the potential of this source.

\section{A pilot program to retrieve WW2 data}

The period around the Second World War (taken here as 1938 to 1947) is a good candidate for a digitisation program for several reasons: firstly, there is a shortage of digitised observations; secondly, during this period there were big changes in the way marine observations were made; and finally, there is evidence of large and sudden changes in the observed temperature record that require substantiation. From  a more practical point of view, logbooks for this period are easily accessible in the National Archives and are in good condition.

Over the past century, periods of major conflict are marked by deficiencies in the global climate record. Figure \ref{f2} shows the drastic drop in global coverage of the existing International Comprehensive Ocean-Atmosphere Data Set (ICOADS: \cite{worley05,woodruff03,woodruff05,woodruff87}) of digitised marine observations , which falls from 65 per cent in 1937 to 35 per cent in 1941, effectively reducing surface observation coverage to that available from the 1880s. Such shortcomings in the record substantially reduce the precision and reliability of our understanding of regional and global climatic variability and change over this period (\cite{rayner06hadsst2}).  In particular, the 1939-1942 period was influenced by a large and protracted El  Ni\~no episode \citep{allan99enso,Allan03DSRII}, and greater resolution of this natural fluctuation would be invaluable to climate science.
 \begin{figure}
\begin{center}
  \noindent\includegraphics[width=40pc,angle=0]{./figures/coverage+nobs}\\
  \caption{Annual number of SST observations, and fractional coverage. Coverage here is the fraction of ocean grid boxes for which the monthly, 5-degree resolution version of the HadSST2 dataset (\cite{rayner06hadsst2}) has data: approximately the fraction of the ocean with at least one observation nearby (within 5 degrees and one month).}
  \label{f2}
\end{center}
\end{figure}

The period between 1938 and 1942 witnessed some significant changes in how marine observations were made. Sea-surface temperatures (SSTs) went from being measured primarily using canvas buckets, to mostly being taken from engine-room intake thermometers, and these two measurements methods don't give the same temperatures. The SST datasets used for climate research contain adjustments to remove these temperature biases (\cite{rayner06hadsst2,folland95buckets,folland05sstbias}), but uncertainties remain in the size of the biases, and how they change over time. Figure \ref{f3} shows that the current estimates of SST (after adjusting for these instrumental biases) suggest a noticeable increase in the WW2 period. This important feature may, be a result of natural climate variability, but concerns that it might be at least partly an artifact of uncorrected biases in the SST measurements require that as much additional data as possible for the period be gathered in order to resolve the issue.

\subsection{Choosing logbooks for digitisation}

Investigation in the UK National Archives revealed around 30,000 Royal Navy ships' and submarines' logbooks held there for the period in question. Each logbook covered a month, and consisted of about 30 pages, each detailing one day. A typical page contained one or more recordings of the ship's position, and six sets of meteorological observations, including air and sea temperatures and air pressure, taken at different times through the day. Figure \ref{f4} shows a typical example.
\begin{figure}
\begin{center}
\noindent\includegraphics[width=30pc,angle=0]{./figures/warspite_log_19410112}\\
  \caption{Logbook of HMS {\it Warspite} from January $12^{th}$ 1941. Reproduced by permission of the UK National Archives.}
  \label{f4}
\end{center}
\end{figure}
 Funding constraints meant that only about 8000 logbooks could be photographed (about 250,000 pages). Given that existing data for that period are concentrated on European waters and the North Atlantic, logbooks were selected predominantly from ships undertaking long voyages into the southern hemisphere, taking advantage thereby of the extensive geographical coverage offered by this source. Moreover, the existing data shortage is most acute for the years from 1941 to 1946, and attention was concentrated on that period, but not to the exclusion of some sampling from 1938, 1939, and 1947. Selection was done by deployment, rather than by individual logbook --- so if, for example,a ship was posted from Plymouth into the Indian Ocean for a year; all her logbooks for that year were selected. Also all observations from each selected logbook were digitised, not just those from data-sparse areas. This gives a comprehensive set of observations, which is helpful for quality control and allows the widest possible use of the new data, but does mean that not all observations are in places ideally suited for studies of marine climate; more than half of the new observations are from ships in port.

Initial selection of the logbooks was based on the Admiralty Pink Lists. These, then secret, documents were issued every four days and described the then known deployments of the Navy's ships. The selection was later updated by reference to the Admiralty Movement Books. The Pink Lists are held at the National Archives, and the Movement Books are held at the Admiralty Library (Portsmouth) with copies available from the National Maritime Museum (Greenwich). Selected information from these records has been placed in the WW2 Logbook and Movement Directory available on the RECLAIM website.

\subsection{Imaging and digitising the logbooks}

The joint processes of imaging and subsequent digitisation (the latter here taken as the manual transcription of data into electronic form, automated optical reading not being practical) required careful cooperation between the UK and US partners. The National Archives were contracted by the Met Office Hadley Centre to produce the images. Each page of every selected logbook was photographed, and the images were then stored on DVD and shipped to NOAA's National Climatic Data Center (NCDC) which administers the Climate Database Modernization Program (CDMP, \cite{DupignyGiroux07cdmp}). The 267,874 images are stored at NCDC, and are available over the Internet to researchers through their Web Search, Store, Retrieve, Display visual archive system (WSSRD: http://www.ncdc.noaa.gov/oa/climate/cdmp/wssrd.html). Each image was given a unique index key in the visual archive for identification and retrieval purposes.

Upon completion of the indexing, a production keying process was designed and implemented using CDMP's Pre-keying Inventory Comments and Summary (PICS) process.
A small subset of the images were initially keyed as a test, and experience from this test was used to design the final keying output format of the data and to estimate the total cost of the keying process. The test identified a few necessary changes for the arrangement of images and indices within WSSRD and also provided clarification of the instructions to the data keyers. The bulk of the keying of the data was then completed with a minimum 99 percent accuracy rate, verified by blind double keying. The keyed data were randomly sampled to verify consistency with the output format. The keyed information includes the several sets of weather observations that appear on each page (see figure \ref{f4}). The date and time of these observations is usually evident, but corresponding information on the ship's location, be it at sea or in port, is not consistently recorded. Each set of observations is thus reliably fixed by date and time, but location is more problematic and, self-evidently for ships at sea, the location will differ day-by-day. Such reservations notwithstanding, by these means a total of about 1,500,000 observations were made available for addition to the data sets.

A sample logbook page and an illustration of the digitised observations for one Royal Navy ship is given in appendix \ref{warspite}.

\subsection{Observation processing and quality control}

For scientific use, and inclusion in the ICOADS\footnote{ICOADS data and products can all be accessed via the ICOADS web portal (http://icoads.noaa.gov/). This includes links to the National Center for Atmospheric Research (NCAR), from where the WW2 data can be accessed as an ICOADS auxiliary dataset (http://dss.ucar.edu/datasets/ds530.0/); and to NCDC's Climate Data Online (CDO) (http://cdo.ncdc.noaa.gov/CDO/CDOMarineSelect.jsp), from where ICOADS data are also available.}, the observations had to be converted into International Maritime Meteorological Archive (IMMA) format (\cite{woodruff07imma}). As well as reformatting the data, the opportunity was taken to perform some basic quality control on the observations. Location proved to be a particular problem, and, for instance, many longitudes were significantly in error, or assigned to the wrong hemisphere.

The principal limitation with the digitised observations is that many observations could not be fixed by position. When at sea, a typical logbook page contains many more weather observations than position records (figure \ref{f4}): many ships recorded positions less often than three times a day. When in port, the logbooks only record the port name, not its latitude and longitude, so assigning a position requires a conversion table between port names and positions.

The digitised observations included more than 1700 port names, but the majority of those are rarely used. A conversion table was produced for the 123 ports the names of which appeared more than 500 times in the digitised observations. This provided locations for 828,000 (87 per cent) of the 952,000 observations from ships in named ports. Missing positions where the ship was at sea were estimated by linear interpolation between known positions. Interpolation was only performed where adjacent digitised positions were separated by less than 5 degrees in both latitude and longitude, and by less than 48 hours in time. Of the digitised observations where the ship was at sea, 178,000 (29 per cent) had positions given in the logbooks, interpolation raised the number of usable observations at sea to 482,000 (79 per cent of the 612,000 at-sea observations). In all, 85 per cent of the new observations were assigned positions, and so are presently usable for climate analysis. In the future we (or other researchers) may be able to expand the amount of readily usable data through the association of additional less frequently used port locations with latitudes and longitudes, or by improved methods for quality control and interpolation of positions.

\subsection{Assessing uncertainty and bias in observations of sea-level pressure (SLP)}

One virtue of having many observations from ships in port is that several ships are likely to use the same harbour at once, and so provide sets of observations that are co-located in time and space. Such sets of observations are invaluable for inter-comparison of their observations, from which estimates can be made of the systematic and random errors in each ship's measurements. During August and September 1938, 11 of the ships whose logbooks were digitised spent some time in Portsmouth harbour. Figure \ref{f14} shows the SLP data recorded by each ship, compared to observations from a nearby land station (at Calshot - data provided by Rob Allan).
\begin{figure}
\begin{center}
 \noindent\includegraphics[width=30pc,angle=0]{./figures/portsmouth_ts_193808-09}\\
  \caption{Observations of SLP from ships in Portsmouth harbour, and comparison with a nearby land weather station at Calshot.}
  \label{f14}
\end{center}
\end{figure}
The comparison suggests that both systematic biases and random errors in the data are small. The systematic differences between the ships range up to about 4hPa, with the random differences typically being even smaller.

\subsection{Using the new observations in climate analysis}

A principal use of historical marine observations is to contribute to gridded datasets showing the changes in SST through time. One such dataset is HadSST2, produced by the Met Office Hadley Centre (\cite{rayner06hadsst2}). The new observations have been used to make two new and improved versions of this dataset: one using only the new observations, and one made from a blend of the new observations with HadSST2. In both cases the observations have been subjected to the quality control, gridding, and bias adjustment methods used in HadSST2 and described in \cite{rayner06hadsst2}.

\subsubsection{Coverage changes}

The new observations have provided a significant increase in the total number of observations available for the period, and an equally important increase in the global coverage of observations (figure \ref{f10}). Much of the improvement in coverage is due to the judicious selection of logbooks from ships crossing oceanic areas that have hitherto been poorly represented.
\begin{figure}
\begin{center}
 \noindent\includegraphics[width=30pc,angle=0]{./figures/n+c_sst}\\
  \caption{Effect of the new observations on global coverage and number of observations.}
  \label{f10}
\end{center}
\end{figure}

This improvement is most noticeable for the Indian Ocean. The Royal Navy was particularly active in this region during the study period, providing an extensive legacy of observations with which to improve the climatic record. The upper half of figure \ref{f11} is a copy of figure 3.5 from \cite{trenberth07ipcc}, and illustrates the problems caused by missing observations in the 1940s; with the extent and timing of the temperature maximum being almost impossible to estimate. The lower half of the figure is produced in the same way, but with the new observations added. The improved coverage for the 1940s needs little elaboration and is now sufficient to characterise the mid-century temperature maximum with greater precision.
\begin{figure}
\begin{center}
\noindent\includegraphics[width=45pc,angle=90]{./figures/ipcc_indian_figure}\\
  \caption{Latitude time sections of zonal-mean Indian-Ocean SST anomalies ($^\circ$C) for 1900 to 2005, relative to the 1961-90 mean (after figure 3.5 of \cite{trenberth07ipcc}). Upper panel: HadSST2 (\cite{rayner06hadsst2}). Lower panel: HadSST2 with the newly digitised observations.}
  \label{f11}
\end{center}
\end{figure}

\subsubsection{SST biases}

Observed SSTs reveal two large changes over the WW2 period and these may not be wholly climatic in character. Firstly, observational procedures changed from measuring SST by taking samples in canvas buckets, to using measurements of engine-room intake temperatures. This is known to have produced a systematic change in the temperature observations ({\cite{folland95buckets}) leading to a significant increase in measured temperatures between 1939 and 1941 (\cite{rayner06hadsst2,folland95buckets}). Even after correcting for these biases, measured SST shows a distinctive if attenuated peak in the mid century, followed by an abrupt fall in 1946 (figure \ref{f3}). Largely because of the war, there are also big changes in the list of ships contributing observations over this period. Different ships, particularly if from different nations and fleets, had different observing practises, and it is possible that there are some additional, as yet unknown, biases in the record. The new Royal Navy observations provide a useful check on this and, because all the ships presumably operated to the same set of meteorological observing practises, they should provide a consistent time-series prepared to consistent observational standards. As a result comparison between the Royal Navy only data and the older observations from a mixture of sources should draw attention to any effects caused by bias and non-climatic signals in the latter.

Comparison of SST time-series from the old (multi-source) observations and from the new Royal Navy observations shows good agreement in the overall trend over the period (figure \ref{f12}). There are fewer new than old observations, a feature that may account for the higher variability of this series. Data shortage is a yet more probable explanation for the departures in the record for 1939 and in 1947. Such differences notwithstanding, the two series agree well over the 1938 to 1945 period; and both then show a fall in 1946 and 1947, though the shape of the curve over these two years is quite different in the two datasets. This suggests that the abrupt fall in SST seen in the older observations in 1946 may be an artifact of changes in measurement methods, but further analysis will be required before this hypothesis can be offered with complete confidence.
\begin{figure}
\begin{center}
 \noindent\includegraphics[width=30pc,angle=0]{./figures/sst_old+new}\\
  \caption{Global mean SST from older observations (HadSST2) and the new Royal Navy observations.}
  \label{f12}
\end{center}
\end{figure}

\subsubsection{Future uses}

These logbook data have a wide application. More extensive work on the SST measurements from the logbook observations is expected to result in new versions of the HadSST2 dataset (\cite{rayner06hadsst2}). The observations also include dry and wet-bulb air temperature measurements, which will contribute to any future versions of the nighttime marine air temperature dataset HadMAT (\cite{rayner03HadISST1}) and the humidity dataset HadCRUH (\cite{willett07phd}). The pressure observations will be used not only in any future versions of gridded pressure datasets like EMSLP and HadSLP (\cite{ansell06emulate,allan06meansealevelpressure}), but also will be assimilated in general circulation model reanalyses covering the period: notably the Twentieth Century Reanalysis Project (\cite{compo06reanalysis}).

It should not be forgotten that only a small proportion of available logbooks have been included in this exercise, and many data remain untapped. A succession of reanalyses using only surface observations, including and building from the Twentieth Century Reanalysis Project are being planned under the Atmospheric Circulation Reconstructions over the Earth (ACRE) initiative (Rob Allan, personal communication). These analyses will require the recovery and digitisation of the bulk of the outstanding ship logbook observations held as hard-copy volumes in UK repositories.

\section{Conclusions}

Our understanding of large-scale climate change and variability is based largely on a database of digitised historical weather observations. There are very few such observations available from before about 1850, and there are also significant shortages of observations in the late 19th century, and in the 20th century during the two world wars (figure \ref{f2}).
Observations exist which could be used to improve the weak points of the database, and to extend the provision of data back to 1800 or before, but these observations are only available on paper - in logbooks and reports in libraries and archives, and are currently unavailable to the climate research community. The Joint World Meteorological Organization(WMO)--Intergovernmental Oceanographic Commission (WMO) Technical Commission for Oceanography and Marine Meteorology (JCOMM) Workshop on Advances in Marine Climatology (CLIMAR, \cite{parker04climar2}) series, and the International Workshop on Advances in the Use of Historical Marine Climate Data (MARCDAT-II, \cite{kent07marcdat}) series, have both emphasized the importance of digitising such observations and making them widely available to researchers.

This campaign to digitise meteorological observations in Royal Navy logbooks for the period between 1938 and 1947 has substantially improved the available observations coverage for the Second World War. It has also confirmed the benefits of utilising this data source. Similar undertakings could provide similar improvements for the First World War and the late nineteenth century, and could eventually enable instrumental estimates to be made of large scale climate variability and change for the whole of the last 200 years.

\begin{acknowledgment} 
Funding for this project was provided by the Climate Database Modernization Program of the US National Climatic Data Center, and by the Joint Defra and MoD Programme, (Defra) GA01101 (MoD) CBC/2B/0417\_Annex C5. The logbooks used are in the collection of the UK National Archives, who also made the digital images.
\end{acknowledgment}
\clearpage

\appendix
\section{Logbook of HMS {\it Rattlesnake}, 1796}
\label{rattlesnake}

Royal Navy officers began to maintain logbooks in their currently recognisable form in the 1670s (\cite{wheeler04usl}). From that point until the mid-nineteenth century the presentational style of these documents was broadly consistent and fell into two categories, the more commonly used of which is exemplified in figures \ref{f5} and \ref{f6}. In this style, logbooks consisted of facing pages. The left hand page (figure \ref{f5}) contained essential navigational information (latitude, longitude and bearings to land marks), the date, and notes on wind direction, the latter commonly on a 32-point compass. The facing page (figure \ref{f6}) was more narrative in style but each day's entry opened with a note of the wind force and the weather at noon of that nautical day. In the example shown from November 1796 HMS {\it Rattlesnake} lies some 52 leagues (approximately 150 miles) west of Table Bay, and the winds on the 3$^{rd}$ November are South by East backing to South later, and are described as ``fresh gales and squally'' (Beaufort force eight according to \cite{cliwoc03dictionary}). The alternative style popular in the East India Company (EIC) is illustrated in appendix \ref{hindostan}. Both contain the same types of information although the EIC logbooks in latter years often also contained daily instrumental observations.
\begin{figure}
\begin{center}
 \noindent\includegraphics[width=25pc,angle=0]{./figures/Rattlesnake_7_}\\
  \caption{Left-hand facing page from the logbook of HMS {\it Rattlesnake} for the period 3rd to 13th November 1797. The entries for the first day shown (3rd November) read as follows: ``Winds SbE, South. Course S54W. Distance covered 86 (nm). Latitude 34.06 (S). Longitude 14.58 (E). Bearings Table Bay N86E Distance 52 leagues (150 miles)''. Reproduced by permission of the National Maritime Museum, Greenwich}
  \label{f5}
\end{center}
\end{figure}

\begin{figure}
\begin{center}
\noindent\includegraphics[width=25pc,angle=0]{./figures/Rattlesnake_8_}\\
  \caption{Right-hand facing page from the logbook of HMS {\it Rattlesnake} for the period 3rd to 13th November 1797. The entries for the first day shown (3rd November) read as follows: ``First part fresh gales and squally. Latter part more moderate. Made sail as necessary. Read Articles of War and mustered the Ship's Company''. Reproduced by permission of the National Maritime Museum, Greenwich.}
  \label{f6}
\end{center}
\end{figure}

\section{Logbook and observations from East India Company ships {\it Castle Huntly} and {\it Hindostan}, 1799 and 1820}
\label{hindostan}

From about 1780, Alexander Dalyrmple, chief hydrographer to the British East-India Company, set up a system for his officers to make daily records of air pressure and temperature on their voyages to and from India and China. Instruments were provided, and observations were made at midday and recorded in the logbooks, which had pre-printed sheets containing sections for these data (figure \ref{f13}). Most of these uniquely valuable East India Company logbooks are archived in the British library.

Some insight into the methods and procedures that Dalrymple expected of his crews is helpfully provided when he describes one such journey in detail and includes also the data to which it gave rise (\cite{dalrymple78}). But this voyage was by no means unique: Figure \ref{f7} shows temperature and pressure records recovered, by Dennis Wheeler, from a further example: that of the {\it Hindostan}, which left London on June 20th 1799 and travelled to China via the Cape of Good Hope and the Indian Ocean, arriving in Canton on January 10th 1800. The graphs summarise, for this one voyage alone, the quantity of data that are available in addition to the more standard wind force and direction described above. Thus far no comprehensive attempt has been made to recover these data, which are of particular value as they include passage through the southern oceans where even today data are scarce. It is estimated that as many as 800 EIC logbooks contain data of this character, embracing the period from the 1780s to 1840s.

\begin{figure}
\begin{center}
\noindent\includegraphics[width=30pc,angle=0]{./figures/Castle_Huntley_page_2_}\\
  \caption{A typical page from the logbook of a ship in the service of the East India Company, the {\it Castle Huntly} en route to Bengal in April 1820, at which time she was sailing in the Indian Ocean. This layout was also used by some Royal Navy officers. Note should be taken of the greater volume of sub-daily wind and weather observations and of boxes for inclusion of daily barometer and thermometer readings. Reproduced by permission of the National Maritime Museum, Greenwich.}
  \label{f13}
\end{center}
\end{figure}

\begin{figure}
\begin{center}
\noindent\includegraphics[width=30pc,angle=0]{./figures/hindustan}\\
  \caption{Temperature and pressure recorded on board the {\it Hindostan}, travelling from London to Canton, in 1799--1800}
  \label{f7}
\end{center}
\end{figure}

\section{Observations from HMS {\it Warspite}, 1941--3}
\label{warspite}

HMS {\it Warspite} (figure \ref{f9}) was one of the most famous ships of the Royal Navy. She was launched in 1913, and served with distinction through both world wars. Her logbooks are in the National Archives, and those for the period 1941 to 1943, when she was serving outside UK home waters, have been digitised as part of the project. Figure \ref{f8} shows the route of the {\it Warspite} during this period, as obtained from newly digitised observations from her logbooks.
\begin{figure}
\begin{center}
\noindent\includegraphics[width=40pc,angle=0]{./figures/route_warspite}\\
  \caption{The route of HMS {\it Warspite} between 1941 and 1943. {\it Warspite} served as flagship of the Mediterranean Fleet from May 1940. She was damaged by air attack in May 1941 during the battle of Crete, and, after temporary repairs at Alexandria, steamed via Singapore to Bremerton, in the U.S. Pacific North West, for repairs. She left Bremerton in January 1942, by which time Japan had entered the war, so she steamed to her new station in the Indian Ocean (as part of the British Eastern Fleet) via the South Pacific and Australia. She served in the Indian Ocean until June 1943, when she was recalled to the Mediterranean; and in September 1943 she was once again damaged in an air attack, this time by armour-piercing glider bombs (during the Salerno landings). Returning to Rosyth in March 1944, {\it Warspite} was never fully repaired, but took part in support of the Normandy landings in June 1944 \citep{coward86}. }
  \label{f8}
\end{center}
\end{figure}

Operational requirements took the {\it Warspite} over a wide area of the globe. On all these journeys weather observations were made, and recorded in the logbook, several times a day. An example of such a page from the ship is shown in figure \ref{f4}. As with the earlier logbooks, this one contains navigational data and a column of general remarks. But now there are instrumental weather observations included: pressure as well as dry-bulb, wet-bulb and sea temperatures. There are also precise positions given.

\begin{figure}
\begin{center}
\noindent\includegraphics[width=30pc,angle=0]{./figures/HMS_Warspite_Indian_Ocean_1942}\\
  \caption{HMS {\it Warspite} underway in the Indian Ocean (image from Wikimedia Commons)}
  \label{f9}
\end{center}
\end{figure}

% Create a bibliography directory and place your .bib file there.
\clearpage
\bibliographystyle{./ametsoc}
\bibliography{database,digitisation_extra}


\end{document}