Customization

WXSIM can be run for any location in the world. However, in order to enable all the features and produce really accurate forecasts, it must be customized for a particular location. The customization process consists of creating a unique file of site data, called cty.fdt. A cty.fdt file can actually have any number of sites; mine has all the sites I've ever done - over 1000 of them! For customers, though, I produce a file with the home site as the first entry (some users have a few such sites), followed by Atlanta - for use with the tutorials in the manual - and about 25 other, less carefully done sample sites.

Two other, much smaller files can also be regarded as 'customized'. The file custinit.txt contains information for program boot-up, including the default site along with many parameters that preserve various settings you choose when you run the program. The file retini.txt contains similar information for use with the data retrieval module.

Some description of the customization process is in order, partly so that if you choose to order, you know what kind of work you're paying for. The data types in cty.fdt file entries can be grouped roughly into the following categories:

• Geographical data, including latitude, longitude, elevation, and time zone. You could potentially enter this kind of information yourself, but it would be a bit tedious.
• Climatological data, consisting of parameters describing normal temperature, dew point, and 'water temperature' for the site. Also included are data on average wind speed relative to large scale flow, and diurnal variation of wind speed.
• Parameters describing surface type (ground cover), continental versus maritime aspects of the site, and degree of urbanization (which can be changed at run time). Some of this information is used to help estimate default soil moisture and the reflectivity of snow cover.
• Parameters describing the distances and directions to large bodies of water, along with allowance for island locations and data concerning sea breeze strength.
• Information about the most significant mountain ranges affecting the area's weather. This includes distance, direction, angular span, and effective height of the mountains. In some customizations it may also include mountain/valley breeze parameters.
• Parameters influencing diurnal temperature range and overall temperature bias. This is an especially critical area, for which I do a lot of careful testing and tweaking.
• Various other items, some of which are not used in all customizations. These include dependence of diurnal tempreature range on wind direction and on time of year (mainly for agricultural sites with a lot of transpiration from crops in summer), and tendency for short-term temperature fluctuations affecting reported maximum and minimum temperatures.
• Advection sites. I choose a large number of surrounding official reporting sites (METAR and/or synoptic, and in many cases buoys). This usually includes most or all stations within a couple hundred miles, and more scattered ones at distances of up to (sometimes over) 1200 miles (2000 km). For most North American and European sites, I select about 220-280 advection sites, though sparsely populated regions will naturally have less. The data includes distance and direction as well as elevation correction information.
• FOUS sites. For U.S. and some Canadian sites, I choose three NAM FOUS sites, among which WXSIM can interpolate to estimate values at the home site. This data is not available outside North America, though fortunately the GFS model, available from the READY site, has world-wide coverage.
• RAOB sites. I always choose three surrounding RAOB (upper air) sites from which WXSIM can interpolate upper air conditions at the home site.
• Climate narrative. This is simply a paragraph describing your site's climate, thouh you may well know more about it than I do!

As you might surmise, the customization process in quite involved. The various parameters interact in many subtle ways in the program, so considerable care is necessary. Over the years, I've developed a number of tools which help greatly. Some of these I use every time, and some only for certain types of sites. I have a program which helps me set up most of the data, but I always go back and make adjustments after the intital creation. I regularly use a program which maps out the parameters for previous sites I've done, so that I can build on that experience in an organized way. For many major U.S. sites, I have a vast amount of data from the National Climatic Data Center, and a program I wrote for powerful analysis of that data, including graphs of 'model days', consisting of dozens of days meeting selected criteria. I can then directly, graphically compare WXSIM output to this actual data and tweak cty.fdt to improve the fit. I have written other software which helps me do something similar with home weather station data (especially Davis stations). My climatological data sources include numerous internet sites and Tim Vasquez's Digital Atmosphere. Finally, I almost always run test forecasts using real-time data, and either wait a day or two to see how they verify, or compare to other forecasts for the area.

In the past, the above work took me anywhere from 2 or 3 hours in the case of very straightforward sites in locations where I have a lot of prior experience, up to 12 or more solid hours of work for especially challenging cases, such as a mountain near a coast, with years of home weather station data to analyze. On average, a customization took me perhaps 5 or 6 hours, spread out over two or three days.

I have found ways to slightly reduce the time spent customizing, but have also added new tasks, such as routinely analyzing users' home weather station data and repeatedly testing against climatology and also output from surrounding, previously done sites. This has kept customization time generally about 3-4 hours. However, there are now two options: basic customization, which includes the above bulleted points and takes about 2 hours or a bit more, and enhanced, in which I still (for a bit more pay) do the full-blown job, taking perhaps 4 hours. The accuracy difference should be small in most cases, so the basic option is fine for most customers. For some sites, though, the extra cost and effort may be worthwhile and yield a noticeable improvement in temperature forecasts. Click here for more details.

This work is the main thing you're paying for with customization. I could probably do it quicker, but I'm frankly rather obsessed about getting it right - or at least as good as possible - and that's good for you!

Customizing the companion program WXSIMATE takes much less time, once I've done all the work setting up WXSIM itself. I develop a registration code which enables WXSIMATE to use WXSIM's custom files, and - especially if you are ordering the WXSIM/WXSIMATE bundle - also leave various settings in a state appropriate to your region.