[Mesogeiakos]

Quote

If you are saying Malaga would be warmer than Crete annually then i know you would be wrong on that one.

Exactly,I think it will take more than 18 days of data for February(actually less days on record) with a hot spell in Iberia and 15 days of March data with a cold snap in Greece to actually suggest that Crete will be cooler annually.

[Ian Williams]

Thanks Pit,
The trouble is AEMET does not have any past data for Malaga Port, Estepona and Fuengirola, are they newish stations like Andujar? Now these three station may turn out warmer annually than Seville, because i have no past data, certainly it would seem they are warmer during the winter months, especially their minimums. However i can positively say that the likes of Karpathos, Ierapetra, Palaiohora and even Rhodes (Rodos) are all warmer annually than anywhere in Spain.

[Mesogeiakos]

Ferreiro,with the limitations listed below

1.Only 18 day of Feb data with an extreme heat spell in Spain
2.Only 15 day of March data with an extreme cold snap in Greece
3.With methodological errors in your calculations pointed out by Ian

Btw March is not over again and correct me if I am wrong but in January it was Karpathos the warmest area in Europe according to the official data in absence of the data from South Crete.

[Ian Williams]

Quote

Ferreiro - 17/3/2011 12:42 What matters are the official data. Within one year we will know. I began my research on February 10. In March is happening the same. I suspected in January was the same too. Official data from AEMET and HNMS from 10th February until 28th February.

OK Im happy for you to post your data, and i will post my data some of which will be adjusted if it needs to, I dont think our figures will be way out. But lets start this from March 1st for this thread OK? I will not post daily but say one or 2 times a week with my up to date data ok?

[Mesogeiakos]

I think from March onwards is a good idea Ian,the Feb data we have are really for 18 days (actually for even less days in various locations) plus Karpathos data are missing.

[ldavidcooke]

Hey Guys,

This is part of the battle in climate change as well, how do you handle a non-reporting day/month?

Doing as NASA did with gridding, or NOAA did with adjusted proximity substitution or even Ian's hourly +/- 0.5, when all is totaled even for the week will be different at a greater decimal precision value then the decimal precision of the data. (IE: 6 values say 1.5 totals to 9 or averages at 1.5, 7 values with 6 at 1.5 and one null or 0 and we get 1.2857142 ad nauseum. Since the precision was one decimal the output would be either 1.2 or rounded up 1.3.)

If we followed Ian's procedure (say the hourly was 1.2) we get an average weekly value of 1.5285714 or either 1.5 or 1.6 if you round up. So which is the more accurate?

Doing a little extra data table and calculation work/definition it is possible to only use those days for which you have a valid entry.
(Note: Two primary failures of spreadsheet and data base engine designs: 1. An empty/null field once you enter data in it and remove the data it is no longer null. 2. Formula value calculations will allow non-integer values and give them a value of zero.)

It is quite possible that everyone can be looking at the same data and get different results.

Cheers!
Dave

[Mesogeiakos]

Thanks Dave,

I see what you mean.It is a hassle all right!But even for days of valid entries there are limitations,for example say 14 days of data from one location compared to 28 days from another location,can we reach good results?I mean how do we address this?

[ldavidcooke]

Hey Pit,

It requires you put more sophistication into the effort. As to which is the most accurate a representation the best way is to report say both the average and the valid data count: Such as X-site reported an average temperature of 1.5C over 14 days while Z-site reported 1.3C over 28 days reporting.

What makes it more difficult for climate change is to be valid statistically the number of values being averaged should be a mininum of 30. So what do you do if between say 1949 and 1979 you are missing the average temperature for June the 1st 5 times. I would suggest using up to three values on either side of the central date in the calculation as it would be more representative then applying "fudge factors". Even if the margins contained outliers it at least are based on actual measures, hence allows for testing of the recording instrument and instrument interpretation.

As long as you document the process you used to arrive at your conclusion it is valid for the work you have done. If it differs from other works the two different processes/approaches should be tested either by the competing sources or an independent jury of peers.

Cheers!
Dave

[Ian Williams]

Interesting Dave, I think what i will do for the problematic spanish stations is run both Ferreiro`s method and my own. With missing days without the max and min temps im currently using the hourly data, what if i were not to subtract the 0.5 to the hourly minimum and not add the 0.5 to the maximum, would this be considered more scientific/accurate than just excluding it altogether?

Below is an example of Ferreiro`s and my own data for Malaga, there are a couple of discrepencies because Ferreiro uses the 24 hr min and maxes supplied by AEMET, i.e Max occured at 2350 hrs or a min temp occurring at 1420 hrs, when the Greek stations and likewise the UK uses mins between 2100 and 0900 hrs and maxes between 0900hrs and 2100hrs. So again if this happens i use the hourly minimum between the times i use to try and get like for like.

Minefield springs to mind!

[ldavidcooke]

Hey Ian,

Both are valid, it depends on the application. Measured values with time stamp will always be more valid in a data table.

(As to the analysis, there is nothing wrong with modulating the peaks with a standard value as long as you can demonstrate that the adjustment value when compared to known values, match within the significant values.

(IE: max. of 1.2C +0.3 at 1500 or -1.2C -.03 at 0600 versus a measure 1.5 min./max. would demonstrate a good "fudge factor". Even if it were +0.4 (1.6) for the max. and -0.3 (-1.5) the min. the difference to two significant values would match, not rolled-up.)

Hence, testing of the value at a given time as opposed to the maximum will provide a "fudge factor" for that site, if a random set of 30 values within a year can be defined. You would likely want to do this same test for each poor data record site. Then compare your sites and see if the "fudge factor" is valid for both or can you average the different values to get a "universal fudge factor" value. Like I said a lot of work up front; but, scientifically defensible.

As to which source to use, If the apllications are human activity oriented you would use the daytime 0900-1700 window. If the applications are water vapor limited solar energy flux or thermal dynamicaly oriented you would likely employ the 24hr. min./max.

(In the later case you should pair the prior days max. with current days min., if you want to demonstrate residual energy. You should use todays min./max. to represents the atmospheres impedance or thermal inertia. But, that is getting entirely to complex for your application and this Forum or thread.)

Cheers!
Dave

[Mesogeiakos]

While today 17.9C in Herakleion and 16.8C in Palaiohora and Karpathos as mean temps.From now on and especially from April onwards South Greece will see as always much higher minimums.Palaiohora would be SO interesting to watch in terms of its means from April onwards!!Especially in the summer the minimums of Palaiohora will be close to the maximums of coastal Iberia

http://www.hnms.gr/h...region=ObsCrete

[Costa]

Doe Portugal fix on this thread too? :D

[Ian Williams]

Quote

Costa - 18/3/2011 22:29 Doe Portugal fix on this thread too? :D

Yeah why not :-)

[Ian Williams]

AEMET has Fuengirola`s min temp as 15.9C but this was at 2359hr on the 18th, its minimum hourly temp until 0900hrs on the 18th was 16.8C so i deducted 0.5C. I dont get AEMET, Estepona just because it misses one hourly reading in a 24hr period it does not release its min and max temps. Estepona`s hourly max was 23.5C so have added 0.5C to make it 24.0C and likewise have deducted 0.5C from its minimum

[Ian Williams]

Malaga Port was the warmest on the 19th with a mean of 18.9C, Coldest from the group was Tatoi with a mean of 12.4C.

Andujar had the hottest max not just from this group but in Europe with a Max of 28.0C

[ldavidcooke]

Hey Ian,

In your posting from Sat. (two above this one), so by your adjustment you were 0.25 higher for one and 0.25 lower on the other, so what difference did it make to tbe record? For the site it artifically injected uncertainty into the data set. For the region the modification did not change a thing. If both min. and max. were missing the difference would be net zero. So at best you are net zero most of the time you "modulate" the known data and 1/2 the time you may understate the average by 0.25C (which should be either 0.2 or 0.3 as your temperature reading is only to tenths, not hundredths of a degree.)

This then begs the question of, "Is it really necessary to modulate the data? ". Also, if you have not marked all the ones modified how would you return to the raw data record? How would you recover the hour of max./min. observation, if you do not have a daily min./max. value?

Cheers!
Dave

[Ian Williams]

Hi Dave,
I am thinking of scrapping Estepona, as AEMET the Spanish meto, hasnt released any max or mins since the 13th, now my crude method was only going to be for the odd day here and there, thus not really altering the mean temps over a long period, however if i were not to use the method at all Esteponas mean with what max min data i do have (up to the 13th) would be significantly lower than it is with or without making the 0.5C adjustments. To be honest its not a major deal as both Malaga Port and Fuengirola are quite close to Estepona and similar in temperature anyway.

[ldavidcooke]

Hey Ian,

If air/ocean currents and altitude are truly similar it may be worth discontinuing. However, in my brief review, it appears different enough. As to cooler without an adjustment, I would simply add two columns "min. time , "max. time", keeps the data pure. I repeat all i/o info. is valuable, just it is more important to have raw rather then modulated data even if it seems cooler, at least it is a measured value.

Cheers!
Dave