SOURCES AND METHODOLOGY

 

This section describes where I got the information that I used to develop those professional opinions and long-term forecasts that make up my Sunshine Guides.  It also explains the long and tortuous road from the raw data to the finished product.  If you have little or no interest in such matters, please feel free to skip this subject entirely.  You won’t hurt my feelings a bit. 

 

 

SOURCES

The proximate source of my forecasts is the rather extensive personal library of climatic data and information that I have compiled in over fifty years of more or less intensive collection of such things.  This collection is better then that in many university librariesCand much of it was copied in many university libraries.  I have spent way too much time in various libraries, copying sources laboriously by hand, photographing documents, and feeding an endless supply of coins into library copying machines. 

 

The most useful of these libraries have been the Library of Congress in Washington (DC), the National Climatic Data Center (NCDC) library in Asheville (NC), and the libraries of the University of Michigan (Ann Arbor) and the University of California (Berkeley), and the Duke University library in Durham, NC. 

 

The ultimate source of my forecasts, of course, has been the thousands of publications published (andCall too oftenCnow out of print) by the hundreds of various national meteorological and climatological organizations in nations throughout the world.  Sometimes, you can order publications directly from these organizations, but it is ultimately an unsatisfying and patience-wearing exercise.  The Library of Congress gets copies of most of them in the long run, anyway. 

 

Useful intermediate sources include the many compilations made by scholars from those primary sources.  Among the most useful have been the Handbuch der Klimatologie by Köppen and Geiger; the two WWII-era compilations by the British Meteorological Office, Tables of Temperature, Relative Humidity, and Precipitation for the World, and Tables of Temperature, Relative Humidity, Precipitation and Sunshine for the World; the pricey but invaluable sixteen-volume World Survey of Climatology, edited by Landsberg; the U. S. Naval Weather Service’s World-wide Airfield Summaries (use with care, contains estimates as well as data of record); and the World Meteorological Organization=s two compilations, Climatological Normals for the Period 1931-1960, and Climatological Normals for the Period 1961-1990.  

 

Finally, there are the extremely useful compilations issued by the NCDC in various non-print formats.  These include the exhaustive Global Climatic Historical Network (GCHN); the International Station Meteorological Climate Summary (ISMCS); the Global Daily Summary (GDS) for 1977-1991; and the U. S. Navy=s Marine Climatic Atlas of the World.  I use these sources every day. 

 

This, then, is where I get the raw data for my forecasts.  The next section deals with what exactly I do with this data once I have it. 

 

 

METHODOLOGY

It would be nice if, for any given destination, there was a single set of climatic data that was the official set for that destination.  If would make my work so much simpler.  All I would have to do is to reprint that data-set (with the appropriate citation of credit, of course) and my work would be done.  Unfortunately, this is most definitely not the case. 

 

 

Multiple Weighted Extracts

Most destinations have dozens of sites scattered over their area for which climatic records of varying lengths have been compiled.  In most cases, they are all official, in that the records appear in the official publications of that nation’s meteorological organization.  I think that I once counted twenty-three for the Phoenix (AZ) metropolitan area, alone.  Some of these sites have long records, some have short ones.  Some are in areas that travelers are likely to visit, and some are not.  In addition, there are the numerous sites operated by airfields, military bases, universities, and agricultural experiment farms.  In many cases, these latter records are better and more detailed than the nation’s meteorological agency sites. 

 

Not only that, but not all of this data is equally useful.  Records may be fragmented, with no measurements being made at certain times of year or certain times of day.  This is very common in some areas of harsh weather.  Instruments may be badly located, or observers badly trained.  Quite often, estimates will be printed and passed off as actual readings.  Numerical errors creep in at all levels (you can safely disregard a temperature of -44° offered for the central Sahara in August).  A certain level of expertise is required to properly evaluate the overall validity and usefulness of many compilations. 

 

Because there are many possible sources of widely-varying quality, it is necessary to make decisions as to which particular sources will be utilized in the final numbers, and what weight to give to each.  Over the years, I have developed a computer program which enables me to do just this with some ease. 

 

Once I have decided which sources to use (the average is about eight), I give each source an array of weightsCone weight for each parameter that is being evaluated.  My program allows for fifty-six parameters, which means up to 56 weights for each source.  These weights are based on my professional judgment as to how useful that particular source is in regard to that particular parameter. 

 

The published value for a particular month and parameter (let’s say, January snowfall) is multiplied by the weight assigned to that combination of source and parameter.  The total weighted value for all sources used is then divided by the total of the weights to get the value that will appear in the table.  This sounds cumbersome, but in actual practice it goes fairly quickly.

 

More than half of the table values have been treated in this way.  If they appear to be the same as some published value, it is usually due to a rather unusual agreement among all the sources, or to simple rounding in print (the computer uses sixteen significant figures) of the table values. 

 

It is this process that makes the results a matter of professional judgment rather than a matter of fact.  It also avoids copyright infringement, since I make it a point never to use a single source for any parameter.  That and their presentation as forecasts rather than records, enables me justify my own copyright.  I decide how much weight to give to each individual source. 

 

 

Interpolation

 

Not every destination has records for all of the twenty or so climatic characteristics that I use in my Sunshine Guides.  Where no records exist, it is necessary to produce the table values by some means other than multiple weighted extracts.  One of these means is interpolation.  Care must be exercised in the use of interpolation, however, because not all climatic characteristics are suitable for interpolation.  I do not use it at all for precipitation values, humidity values, or fog.  It has only limited value for temperature characteristics.  It is most useful for sky data, since (except in mountainous areas) that sort of characteristic does not change rapidly with distance. 

 

If a destination has no sky data, I will use interpolation from surrounding sites to produce values for that destination.  This is an accepted climatological procedure, and is used by the meteorological services of many nations.  All maps using isopleths and isorithms use this technique.  The weights that I give to the various sources are in proportion to their similarity to and to their distance from the destination in question. 

 

 

Regional Ratios

I use interpolation much less than I used to, because I have replaced it with a technique that I call “regional ratios”.  This technique is based on two hypotheses: 1) certain climatological characteristics are related; and 2) these relationships are more similar in nearby areas than in distant ones. 

 

An example of the first hypothesis is that the percentage of sunshine is related (although not necessarily linearly) to the percentage of cloud coverCor more precisely its inverse, the percentage of clear sky.  An example of the second hypothesis is that the relationship between sunshine and clear skies at Grand Island, Nebraska, is more similar to the same relationship at Lincoln and North Platte than it would be to that relationship in San Diego or New York City. 

 

This relationship can be expressed in the formula:

 

SS(x):CS(x) = SS(y):CS(y)

 

Here, SS is the percentage of sunshine, CS is the percentage of clear sky (100% minus the percentage of cloud cover), x is the station for which no sunshine data is available, and y is a nearby site where it is available.  Both sites have cloud-cover data.  By solving for SS(x) from a number of nearby sites, and assigning the proper weights, a value for sunshine can be produced for use in the tables.  Since the ratios become imprecise at low numerical percentage values, the computer program switches over to incremental ratios in these situations.  This is a far from perfect technique, but it is demonstrably better than interpolation. 

 

 

Multiple Correlation

Many climatic characteristics are related to one another, although usually much less obviously than sunshine and clear skies.  Frequency of precipitation at level x is related both to total precipitation and to frequency at level y.  The relationship is non-linear.  My table characteristic, “Reasonably Dry Days” involves a level of roughly a tenth of an inch.  Since that measurement is rarely made outside of the U. S., I calculate the table values using values for the total precipitation and the values for “Dry Days”.  The percentage of dry days, in turn, is derived from the percentage of days with precipitation equal to or exceeding 0.1mm or 0.01".  The formulae involve multiple correlations.  The results are certainly accurate enough for long-term forecasts.  Snowfall is also estimated using this technique. 

 

 

Calculation

Many of the table values involve simple (or, not so simple) calculations.  The hours of daylight values are calculated using the latitude, longitude, and elevation values.  The techniques use simple spherical trigonometry.  The algorithms may be found in most elementary astronomy texts or manuals of navigation. 

 

Data in metric form are entered into the database directly.  The program converts data in other formats into metric units.  It will print in either metric format or American Standard. 

 

Values for relative humidity are often calculated from temperature values and values for vapor pressure or dew-point.  This is necessary because relative humidity is so strongly temperature dependent.  Readings taken at one station at 12:00 noon are not comparable with readings taken at another at 15:00, and there is no standardization of reading times.  My relative humidity values are intended for the hottest time of day.  In the absence of reliable records that approach this ideal, I calculate it.  Again, the technique has its drawbacks (calculated afternoon humidity values tend to be lower than recorded ones, with the inverse true for sunrise calculations), but it serves the purpose. 

 

          If you get the idea that all of these procedures result in values that are not very precise, you are absolutely right.  But this is in keeping with the realities of actual atmospheric variability.  Air temperatures at shoe-top level may be tens of degrees different from air temperatures at forehead level.  Showers fall at the airport, but not downtown—and vice versa.  Forest clearings get different weather than places under the forest canopy, and measurements taken in the clearings are not representative of those in the forest.  City streets differ from grassy parks, seashores from places inland, airports from downtown, and so on and so forth. 

 

          Don’t confuse precision with accuracy.  For pi, the value of twenty-two sevenths is far more accurate than a value of 3.15472861.  The latter is more precise, but less accurate.  My Guides are as accurate as I can get them—precision is not that important.  Just remember that weather conditions vary from morning to night, from day to day, from year to year, and from place to place at any instant in time—often over very short distances.  What you need is a professional judgment—and that is what I give you. 

 

 

Copyright 2007 by Patrick J. Tyson     www.climates.com

Last edited in January of 2010