Historical weather journals from across Canada, spanning 1768-1884, have been transcribed from handwritten records into machine readable formats. The NORTHERN (Nineteenth-century Overseas Records Transcribed for Historical Environmental Reconstruction in the North) project transcribed nearly 2 million weather observations from 46 locations. The original documents are in archives outside Canada. The two principal archives investigated for historical Canadian weather are the United States' National Administration and Records Archives (NARA) and the United Kingdom's Meteorological Office (UKMO) Library and Archives. Some observations were also located in the United States' National Centers for Environmental Information (NCEI) "Forts" dataset. Observers recorded from three to twenty weather variables, in most cases two or three times daily. Validation procedures are carried out with export files produced in both the original format and in modern units. Observations of pressure, temperature, precipitation, snow depth, cloud cover, cloud type, wind direction and wind force are transcribed along with detailed descriptions of events including fires, floods, ice formation and break up, storms and other weather phenomena. The value of these data lies in their detailed observations of sub-daily weather together with descriptive observation of disruptive or extreme weather events. These data will be used to expand knowledge of Canada's climate variability and extreme values for three centuries and to improve global reanalysis data products. 8 ( a ) Wind force measurements taken by the Royal Engineers at Halifax, in pounds per square inch, measured in pounds, ounces, and fractions of ounces; ( b ) wind force measurements by the Royal Engineers in Kingston, in pounds and decimal tenths of pounds, ( c ) original “wind force” measurements for Winnipeg, where an abrupt transition takes place from a one to ten scale (Smithsonian scale) to values up to the high 30 s. Most conversions were applied using the standard functions contained in the lmrlib.py routine produced by the International Comprehensive Ocean-Atmosphere Data Set (ICOADS) from NOAA 41 . Similarly, some observers did not record the leading digits before the decimal for barometric pressure. In most cases either the transcribers added the leading digits based on previous barometric pressure values, or they were added at the first validation stage, but these values needed to be corrected more often than completed written observation in the third validation stage. Common issues and solutions are listed in Table 8 . Table 8 Transcription validation issues. Variable Range (original units) SI units flag Barometric pressure <27 inHg 914.2 hPa out of range <28.5 inHg 965.0 hPa low value >30.6 inHg 1036.1 hPa High value >32 inHg 1083.5 hPa out of range Vapour pressure <0 inHg 0 hPa out of range >2 inHg 6.8 hPa out of range Temperature <−50°F −45.5 °C out of range <−45°F −42.7 °C low value >100°F 37.8 °C High value >120°F 48.9 °C out of range Relative humidity <0% out of range >100% out of range Wind force <0 out of range >10 out of range Cloud velocity <0 out of range >10 out of range Cloud amount <0 <0 out of range >10 >8 out of range Precipitation <0 out of range >30 in 36.2 cm out of range There is the potential for bias in values such as cloud cover, wind or precipitation where the observers did not always record “zero” when there was no cloud, wind or precipitation, but instead left the entry blank. On the other hand, when an observer made an error which is easily verified and corrected, such as writing “39.91” for a barometer entry rather than “29.91” when the sequence of barometric pressure shows a clearing falling pressure trend from 30 inches to 29 inches, the values are updated directly in the database. A data audit feature in the app tracks all changes made to the data post-transcription. The final action in step 2 is the transformation to SI units and the production of the SEF files. Step 3 The third validation check is to verify the data produced in the SEF files. 9 Overview of the dataset ( a ) by variable type and ( b ) by region and station. As can be seen in Fig. 9a , the observations that can by judged by eye, without the need for an instrument, are the most common, with wind direction and force, cloud type, direction of movement, and velocity forming a large part of the dataset. Thermometer observations are the next most common, as the thermometer is a reliable and robust instrument. Barometric pressure observations are nearly as plentiful as thermometer observations, while humidity measurements, mainly derived from dry and wet bulb thermometer observations, are also widely reported. Observers also regularly noted problems with instruments due to external factors. The humidity readings at York Factory were on occasion unable to be recorded as the temperatures were too low to enable readings. The rain gauge on the campus of Acadia College at Wolfville was reported broken on several occasions. The instruments used to measure humidity are of unknown quality and work remains to be done to investigate historical humidity measurements. Other known issues include the pressure values for Winnipeg being too low for credibility from January to July 1869. Wolfville pressure is also lower The relative humidity values for Wolfville between January and May 1856 are also suspect; most of them are too high if considered as accurate recordings but too low if divided by 1000, assuming the observers did not write in the decimal point. The mean sea level pressure values for Mount Forest are sometimes out of range, being higher than could be reasonably expected. The pressure values for Halifax Royal Engineers (RE) suggests two different sites; one from August 1852 to December 1856, and a different site from September 1858 to March 1862. Pressure values for Halifax Dockyards (DY) are higher than physically possible for July and August 1860.