What
There are two types of analysis available. One is for a 24 hour period which ended at 6 in the winter or 7 in the summer on the morning of the day indicated. The other is for a range of days and is referred to as the "climate" surface. The 24 hour surface is discussed separately. The climate surface gives the cumulative precipitation for the CHARM area, number of precipitation events in the period, which stations reported during the time period and how frequently those stations reported.
When
At the end of each month a summary is made of all precipitation events within the CHARM area.
How
When the software analyzes the CHARM database for a 24 hour period, in addition to the information made available on the web site, a daily archive is retained on the analysis server. Part of each archive is a table of stations accepted for the analysis. Also, on those days where maps of the precipitation were made a table of the (X,Y,Z) values defining the surface is stored. The climate surface is built by summing the (X,Y,Z) tables for the required time period. By reading the tables of stations the total number of stations which reported and were accepted even once is determined, as well as the number of times each reported. The stations are then divided into quartiles based on the frequency of reporting.
Other Products Currently Available
The analysis server maintains an archive of each climate analysis. In addition to the climate map each archive contains the summation (X,Y,Z) table and table of reporting stations with their frequency of reporting. These are available to qualified researchers upon request.
Climatic summaries can also be made for other time intervals quite readily.
Remarks
Automatically producing an objective climate surface directly from the raw CHARM data is a daunting task. Ignoring the detection of input errors and anomalous readings there are major problems. First, over any period measured in weeks there are very few stations that will report and be accepted every day. For example in June of 2002, 115 stations reported and were accepted at least once. Only 40 reported more than 75% of the time, and only 9 reported every day. It is hard to make much of a surface with only 9 points. Second, how should the code handle those reports that span a series of days which go beyond the limits of the time period?
Simply summing the daily surfaces for the period avoids all of these problems. This approach has the advantages that it is quick to obtain, requires no manual examination of the input data and is fully objective as the logic used to make the daily surfaces. It also implicitly takes advantage of the fact that the average is probably a better estimate than any single measurement.
The method does have one potentially serious disadvantage. There is no provision in the current algorithm to directly measure how "good" the final surface is at any one point. This is especially significant near the edges of the final surface where there are few gage