Extreme Temperature Data

[solarken]

Looks like the SolarABC / ucf website that used to have ASHRAE temperature data lookup was discontinued in December. Has anyone found another site that is easy to use to get extreme min and record high temperature data by location?

 

[ggunn]

Looks like the SolarABC / ucf website that used to have ASHRAE temperature data lookup was discontinued in December. Has anyone found another site that is easy to use to get extreme min and record high temperature data by location?

That site is now a subscription service where it used to be free. I have not found a free alternative.

 

[pv_n00b]

SolarABC had not been updated in over 10 years. If that's gone all the better since the ASHRAE data was two versions old. This website is not operated by ASHRAE but seems to have current data and a nice map interface to find the station closest to the site of interest.

ASHRAE climatic design conditions 2009/2013/2017/2021/2025

ASHRAE climatic design conditions station finder

ashrae-meteo.info

 

[ggunn]

SolarABC had not been updated in over 10 years. If that's gone all the better since the ASHRAE data was two versions old. This website is not operated by ASHRAE but seems to have current data and a nice map interface to find the station closest to the site of interest.

ASHRAE climatic design conditions 2009/2013/2017/2021/2025

ASHRAE climatic design conditions station finder

ashrae-meteo.info

That's great and the interface is cool, but it isn't straightforward (to me, anyway) which numbers out of all that data to use for min and max design temperatures.

 

[wwhitney]

which numbers out of all that data to use for min and max design temperatures.

Go to the "Extreme Annual Design Conditions" section. An example is excerpted below.

I take it you need the extreme high and low for calculating max Isc and max Voc. If you want a conservative choice, just look at the very right for the 50-year return period extreme temperatures. For a more detailed understanding:

In the example below, we are told that the extreme annual high is 33.2 C with a standard deviation of 3.4 (so it is being modeled as a normal distribution). Half the years it will be above 33.2, half it will be below (if the normal distribution model is accurate).

So what is your willingness for any year's annual extreme high to be actually higher than your chosen design value? If you say a probability of 20%, 10%, 5%, or 2%, the math has been done for you to the right. Those chances correspond to the 5-year, 10-year, 20-year, and 50-year return period values; as you can see the high temperatures increase with increasing return period (and decreasing chance of exceedance). So if you like 5%, you get an extreme high of 39.5C.

If you want a different chance of exceedance, you can use the mean and standard deviation to determine the corresponding value. Spreadsheets usually offer that as a function. See https://en.wikipedia.org/wiki/68–95–99.7_rule for more information.

For example, if you want a 2.5% chance of exceedance (40 year return period), that is approximately 2 standard deviations above the mean, so the temperature would be 33.2 + 2*(3.4) = 40C, which as expected lies between the 20 year and 50 year values provided.

Cheers, Wayne

 

[ggunn]

Extreme cold is a more important design consideration than extreme heat.

 

[wwhitney]

Extreme cold is a more important design consideration than extreme heat.

OK, but the information given is identical, and the procedure is basically identical, just subtract some number of standard deviations from the mean low, rather than add.

Cheers, Wayne

 

[ggunn]

OK, but the information given is identical, and the procedure is basically identical, just subtract some number of standard deviations from the mean low, rather than add.

Cheers, Wayne

In the past I have always used the ASRAE minimum for my low temp adjustment of Voc; what would be the equivalent?

 

[wwhitney]

In the past I have always used the ASRAE minimum for my low temp adjustment of Voc; what would be the equivalent?

As far as I can see, ASHRAE doesn't provide a "minimum". If the old website provided a "minimum" based on the ASHRAE data, then it was making a choice as to how often it was likely to actually get colder than that minimum. Hopefully disclosing that choice somewhere, but if not prominently, you may not have been aware of that.

The data excerpted provides an average yearly minimum, along with a standard deviation for that; the minimum temperature in a year will vary from year to year.

If you use the average yearly minimum for your design, then half the years it will be colder than your design temp at some point; half the years it will never quite reach your design temp. [This is assuming the model normal distribution fits the data well, and that the historic data is a good predictor of future results.]

If your project has, say, a 20 year expected life, then you could use the 20 year return period minimum as your design minimum. Then you'd expect that one year over the life of the project, the actual minimum temperature would be less than the design minimum; 19 years it would not.

If that happening one year over a 20 year project is more often than you like, you can use the 50 year return value number. Then you'd have roughly a 60% chance it never happens over a 20 year project lifetime.

So you have to choose your allowable chance of exceedance.

Cheers, Wayne

P.S. If you have one of the old reports, and it specifies the weather station and data set (time period) used, and the new website has the same weather station and option for the same data set, you can compare the old value with the new values and figure out what choice the old website was making.

 

[ggunn]

As far as I can see, ASHRAE doesn't provide a "minimum". If the old website provided a "minimum" based on the ASHRAE data, then it was making a choice as to how often it was likely to actually get colder than that minimum. Hopefully disclosing that choice somewhere, but if not prominently, you may not have been aware of that.

The data excerpted provides an average yearly minimum, along with a standard deviation for that; the minimum temperature in a year will vary from year to year.

If you use the average yearly minimum for your design, then half the years it will be colder than your design temp at some point; half the years it will never quite reach your design temp. [This is assuming the model normal distribution fits the data well, and that the historic data is a good predictor of future results.]

If your project has, say, a 20 year expected life, then you could use the 20 year return period minimum as your design minimum. Then you'd expect that one year over the life of the project, the actual minimum temperature would be less than the design minimum; 19 years it would not.

If that happening one year over a 20 year project is more often than you like, you can use the 50 year return value number. Then you'd have roughly a 60% chance it never happens over a 20 year project lifetime.

So you have to choose your allowable chance of exceedance.

Cheers, Wayne

P.S. If you have one of the old reports, and it specifies the weather station and data set (time period) used, and the new website has the same weather station and option for the same data set, you can compare the old value with the new values and figure out what choice the old website was making.

Well, of course the old website is gone and I never saved of the reports, but I have plenty of old designs where I show my calculations based on the numbers I got from them. All the designs I have done in the past few years have been in areas where I have worked before, so I have been using the same values. I would shy away from any string lengths that show a Voc of 999.9V on a 1000V maximum, anyway, so my interest is merely academic unless and until I have to build a design somewhere away from familiar territory.

 

[wwhitney]

P.S. My comments on the return period values and how to compute them are not correct. I dug up a second-hand reference to a formula from the ASHRAE document on how to compute them, which is to take the mean extreme annual low M, and subtract F times the standard deviation, where F is given by:

F = -sqrt(6)/pi * (0.5772 + ln(ln(1/(1-n)))

Which I find a bit mysterious (am still looking into where it comes from) but confirmed that it matches the excerpted data above at least for n=50 and n=10. Pretty sure it is based on an assumed distribution that is not a normal distribution (for the extreme annual lows).

Cheers, Wayne

 

[wwhitney]

P.P.S. The distribution the formula is based on is the Gumbel or double exponential distribution.

And return period does mean the reciprocal of the annual chance of exceedance. So if you're happy using a design extreme low temperature that has a 20% chance of exceedance each year, use the 5-year low; if you want only a 5% chance of exceedance each year, use the 20-year low; etc.

Cheers, Wayne