Based on the original paper data sheets (see data sheet PDFs), we determined that there were five sampling periods for which the data collection methodologies were relatively consistent: One: 1936-1939, Two: 1945-1959, Three: 1960-1987, Four: 1988-1995, and Five: 1996-1997. Below we describe, to the best of our ability, the methods that were used to collect the data during each of these periods. We also describe how we managed the data (including how we entered them, did quality control checks on them, and handled missing values), and how we converted values to metric.
Data collection and management:
Period One (1936-1939):
Only precipitation amounts were recorded during this period, which extended from 1936-11-11 to 1939-12-31. Precipitation amounts were recorded irregularly on the data sheets, generally one or more times each week. According to notes written on the data sheets, measurements were made with a Ferguson Rain Gauge that was installed on 1936-11-09, and measurement values (depths in inches) were read directly from the gauge charts.
We manually entered the data from the data sheets. Because measurement values were taken from gauge charts, we assumed that they reflected daily values; that is, we assumed that ”-” and blank measurements for a given day indicated that no precipitation fell. We further assumed that “T” measurements indicated that trace amounts of precipitation fell. Based on these assumptions, we set all “-”s and blanks to 0.000 mm and all “T”s to 0.001 mm in this time period.
We did quality control checks by comparing the entered data back to the data sheets. As a way of testing whether measurements were made daily, we examined the distribution of the count of precipitation events (>0.000 mm of precipitation) across days of the week, by year. Specifically, we conducted a chi-squared test of uniformity with an alpha of 0.10. Results indicated that precipitation events were uniformly distributed, and not biased toward certain days of the week, such as weekdays.
Period Two (1945-1959):
Period two extended from 1945-01-01 to 1959-12-31. Precipitation amounts, precipitation types, current temperatures, maximum temperatures, minimum temperatures, aboveground wind speeds, ground wind speeds, and wind directions were each recorded over at least a portion of this period. Measurements appear to have generally been made daily, typically at or around 8 AM local time. Prior to 1948-09-01, precipitation amounts were always recorded in depths (inches). After 1948-09-01, precipitation amounts were recorded in depths (inches) in summer months (generally May to September) and in both depths (inches) and weights (pounds and ounces) in non-summer months. This suggests that during non-summer months, weights were measured and converted to depths. The data sheets support this interpretation; at one point, they contain a note saying “precip equiv in inches based on change in weight of std gage equipped with modified alter shield.” Also based on this note, precipitation amounts were likely captured with a National Weather Service 8-inch non-recording standard rain gauge, though this was never specified. The instruments used to capture current, maximum, and minimum temperatures (degrees Fahrenheit) were likewise not specified but were likely standard, maximum, and minimum thermometers, respectively. Aboveground wind speeds (continuous distances in miles) were measured with an anemometer placed at 2.1 meters (7.0 feet). Ground wind speeds (continuous distances in miles) were measured with an anemometer placed at 0.5 meters (1.5 feet). The anemometer models were not specified. Wind directions were recorded as the cardinal, intercardinal, or secondary intercardinal direction the wind was blowing from at the time of measurement. The instruments used to capture wind direction were not specified.
We manually entered the data from the data sheets. For precipitation amounts, we assumed that “T” and “TR” indicated that trace amounts of precipitation fell and we set values to 0.001 mm. If precipitation amounts were “-” or were blank, but other data were recorded for that day, we assumed that no precipitation fell and set values to 0.000 mm. Exceptions to this are from 1945-01-01 to 1945-05-31 and from 1946-01-01 to 1946-05-19; during these time frames, it does not appear that precipitation amounts were being recorded, and so we set values to NA (not available). We also set precipitation amount values to NA if they were “-” or were blank and all other entries for the day were also “-” or were blank. For precipitation types, we set values to NA if they were “-” or were blank. For current temperatures, maximum temperatures, and minimum temperatures, we set values to NA if they were “-” or were blank, or if they were written in red, circled, or otherwise flagged on the data sheets; flagged values appeared to be interpolations as they equaled the means of the values recorded just before and just after them. Finally, for aboveground and ground wind speeds and for wind direction, we set values to NA if they were “-” or were blank.
Aboveground and ground wind speeds were subjected to additional manipulations to convert continuous distances in miles to average daily miles per hour. The continuous distances increased from day to day until they reached 999, at which point they cycled over to 000. We did the conversion by subtracting the previous day’s value from the current day’s value (adding 1000 for days in which the value cycled over from 999 to 000) and dividing by 24 hours. We set to NA all average daily miles per hour values that could not be calculated due to NA input values.
We quality controlled the entered data by comparing it to the data sheets. As described above, we also tested whether precipitation amount measurements were made daily by examining the distribution of the count of precipitation events across days of the week, by year. Results indicated that precipitation events were uniformly distributed and not biased toward certain days of the week.
Period Three (1960-1987):
Period three ran from 1960-01-01 to 1987-12-31. Current temperatures, maximum temperatures, and minimum temperatures appear to have been generally recorded weekly at or around 8 AM local time on Mondays. Precipitation amounts, and occasionally precipitation types, were generally recorded weekly concurrent with the other data. They were also generally recorded irregularly one or more times each week, presumably the day following precipitation events. Precipitation amounts were recorded in both depths (inches) and weights (pounds and ounces) for the entire period. This suggests that weights were measured and converted to depths. Past personnel recollect that a National Weather Service 8-inch non-recording standard rain gauge was used to capture precipitation amounts. The instruments used to capture current, maximum, and minimum temperatures (degrees Fahrenheit) were never specified but were probably standard, maximum, and minimum thermometers, respectively.
We manually entered the data from the data sheets. For precipitation amounts, we assumed that “T” indicated that trace amounts of precipitation fell and we set values to 0.001 mm. If precipitation amounts were “-” or were blank, but other data were recorded for that day, we assumed that no precipitation fell and set values to 0.000 mm. Otherwise, we set “-” and blank precipitation amounts to NA because some measurements seemed to potentially capture more than one precipitation event. If precipitation types, current temperatures, maximum temperatures, and minimum temperatures were “-” or were blank we also set values to NA.
As a quality control check, we compared the entered data back to the data sheets. As described above, we also tested whether precipitation amounts were recorded daily following precipitation events by examining the distribution of the count of precipitation events across days of the week, by year. Results indicated that while precipitation events were uniformly distributed across days of the week for several years, for other years, precipitation events were much less likely to be captured on weekends than on weekdays.
Period Four (1988-1995):
Only precipitation amounts, and occasionally precipitation types, were recorded during this period, which ran from 1988-01-01 to 1995-12-31. The timing of measurements is irregular, with one or more measurements recorded in most weeks. We think they were generally made weekly on Tuesdays, as well as daily following precipitation events. Precipitation amounts were recorded in both depths (inches) and weights (pounds and ounces) from the beginning of the period until 1994-08-14, after which they were only recorded in depths (inches). When both weights and depths were recorded, we suspect that they were measured in the former and converted to the latter. Past personnel and data sheet notes (e.g., “take precip depth at seam of 8” can”) indicate that a National Weather Service 8-inch non-recording standard rain gauge was used to capture precipitation amounts.
We manually entered the data from the data sheets. For precipitation amounts, we assumed that “T”, “TR”, and “Trace” indicated that trace amounts of precipitation fell and we set values to 0.001 mm. We set blank precipitation amount values to NA as opposed to 0.000 mm because some measurements seemed to potentially capture more than one precipitation event. We also set blank precipitation type values to NA.
We did quality control checks by comparing the entered data back to the data sheets. As previously described, we also tested whether precipitation amounts were recorded daily following precipitation events by examining the distribution of the count of precipitation events across days of the week, by year. Results suggested that precipitation events were uniformly distributed across days of the week for several years, but for other years, precipitation events were much less likely to be captured on weekends than on weekdays.
Period Five (1996-1997):
The 1996-1997 sampling period extended from 1996-01-01 to 1997-06-24. During this period, precipitation amounts and current temperatures, maximum temperatures, and minimum temperatures were recorded weekly, generally at or around 8 AM local time on Tuesdays. Precipitation amounts were measured in depths (inches), and according to past personnel, measurements were made with a National Weather Service 8-inch non-recording standard rain gauge. Current, maximum, and minimum temperature (degrees Fahrenheit) measurements were likely made with standard, maximum, and minimum thermometers, respectively.
We manually entered the data from the data sheets. We set all blank values to NA. Note that precipitation amounts were not being recorded from 1996-07-16 to 1996-12-31.
We did quality control checks by comparing the entered data back to the data sheets. We did not conduct a chi-squared test of uniformity as data were clearly recorded weekly.
