CET currently at 16.36C by my calculations. The weekend looks quite cool mainly due to cold nights.

Next week temperatures should be slightly above average. This is mainly due to warmer nights returning though. Daytime maximums will only be close to average.

Currently estimating 16.14C by the 28th. So I suspect we will finish close to the July CET of 15.9C, possibly a little higher.

GW sent you a message on Aug 7th saying that had missed date for August due computer problems.No doubt penalty points but gave you my forecast of 16.4 for August. what penalties do I incur please?

Met Office Hadley            16.4c.      Anomaly        0.4c.   Provisional to  14th.

Metcheck                        16.75c.    Anomaly        0.53c.

Netweather                     17.24c.    Anomaly        1.05c.

Clevedon Weather           17.1c.      Anomaly        0.8c.

Mount Sorrel                   17.0c.      Anomaly        0.8c.

My Mean                         17.0c.      Anomaly       -0.5c.

CET currently at 16.36C by my calculations. The weekend looks quite cool mainly due to cold nights.

Next week temperatures should be slightly above average. This is mainly due to warmer nights returning though. Daytime maximums will only be close to average.

Currently estimating 16.14C by the 28th. So I suspect we will finish close to the July CET of 15.9C, possibly a little higher.

My guess shouldn't be looking too bad this month then!

So I suspect we will finish close to the July CET of 15.9C, possibly a little higher.

I've gone for 15.8 but hope that I turn out to be well over 1c too low, due to a so far unforecasted heatwave that kicks in from next weekend.

 

As an example here is my data for 2014. Generally the mean calculated by taking the average of the maximum and minimum gives a higher figure than the mean of all data points during the day. That is not always the case - especially in winter -e.g. if you have a sudden clearance of cloud just before dawn which pushes the minimum right down. Then the simple max / min average will be lower.

Often the differences are actually very small. Sometimes though the differences can be substantial as evidenced by March and September in 2014.

The size of the difference does not appear to be impacted by the absolute diurnal variation in temperature. It will be down generally to more subtle factors in most cases which are hard to pin down to specifics.

Thanks for the detailed response GW, funny that the two largest differences last year were in the equinox months.

My own short experiment, working with hourly data from 9th to 14th August 2015, has returned mean values as follows:

Back Yard: 16.63 from Max/Min, 16.38 from Hourly

Heathrow:  18.95 from Max/Min, 18.55 from Hourly

That gives a difference between the hourly and max/min means of -0.28*C for my location - which happens to be identical to your value for the whole of August 2014 - while for Heathrow the difference is a much larger -0.47*C, but this was strongly affected by yesterday, for which the max/min mean was a remarkable 1.64*C higher than the hourly data mean. Take that day out, and the difference for 9th-13th is -0.24*C which is actually lower than that for my location.

A good example as to why it's important to have large datasets to minimise the impact of unusual events. With data by the minute covering 10 years, I daresay you have that covered GW!  

 

 

GW sent you a message on Aug 7th saying that had missed date for August due computer problems.No doubt penalty points but gave you my forecast of 16.4 for August. what penalties do I incur please?

Unfortunately your entry was too late to be included in the list of predictions for this month. It will therefore count as a missed prediction for the purposes of the competition. The rules provide for a specific formula to be used to determine the prediction error for anyone who misses a prediction. This is based on the upper quartile figure of the range of prediction errors of everyone who made a valid prediction plus an amount of 0.4C. So I can't say exactly what your prediction error for this month will be until the final CET comes in at the end of the month. It looks like this month will have the lowest overall prediction error for the year so far. My best guess at the moment is that you will be allocated an error of 1.10C for this month.

 

 

Thanks for the detailed response GW, funny that the two largest differences last year were in the equinox months.

My own short experiment, working with hourly data from 9th to 14th August 2015, has returned mean values as follows:

Back Yard: 16.63 from Max/Min, 16.38 from Hourly

Heathrow:  18.95 from Max/Min, 18.55 from Hourly

 

That gives a difference between the hourly and max/min means of -0.28*C for my location - which happens to be identical to your value for the whole of August 2014 - while for Heathrow the difference is a much larger -0.47*C, but this was strongly affected by yesterday, for which the max/min mean was a remarkable 1.64*C higher than the hourly data mean. Take that day out, and the difference for 9th-13th is -0.24*C which is actually lower than that for my location.

A good example as to why it's important to have large datasets to minimise the impact of unusual events. With data by the minute covering 10 years, I daresay you have that covered GW!  

Here is a summary of my data since 2006 showing the difference between the mean calculated using one minute data points and the mean calculated just using the maximum and minimum. The positive numbers mean the one minute mean is higher and the negatives mean it is lower.

It is clear that from Nov-Jan generally the one minute mean is slightly higher whereas in all other months it is usually lower. The biggest differences occur in July to Sept.

Interesting to note that for the first half of August this year the difference is just 0.03C on my figures which is way down on the average of 0.40C for August. 7 of the 15 days had a one minute mean that was higher than the max/min mean which is an unusually high number. So this month is turning out to be a little unusual in that respect.

Met Office Hadley         16.2c.     Anomaly      0.2c.   Provisional to 15th.

Metcheck                    17.1c.      Anomaly      0.91c.

Netweather                 16.55c.    Anomaly      0.32c.

Clevedon  Weather       16.9c.     Anomaly      0.6c.

Mount Sorrel               16.8c.      Anomaly      0.6c.

My  Mean                    16.9c.      Anomaly     -0.6c.

Met Office Hadley     16.1c.       Anomaly     0.1c.      Provisional   to  16th.

Metcheck                 16.43c.     Anomaly     0.21c.

Netweather              16.95c.     Anomaly     0.76c.

Clevedon  Weather    16.8c.      Anomaly     0.5c.

Mount  Sorrel            16.6c.      Anomaly     0.4c.

My  Mean                  16.8c.      Anomaly    -0.7c.

 

Here is a summary of my data since 2006 showing the difference between the mean calculated using one minute data points and the mean calculated just using the maximum and minimum. The positive numbers mean the one minute mean is higher and the negatives mean it is lower.

It is clear that from Nov-Jan generally the one minute mean is slightly higher whereas in all other months it is usually lower. The biggest differences occur in July to Sept.

Interesting to note that for the first half of August this year the difference is just 0.03C on my figures which is way down on the average of 0.40C for August. 7 of the 15 days had a one minute mean that was higher than the max/min mean which is an unusually high number. So this month is turning out to be a little unusual in that respect.

Those differences in some of the months are higher than I would have guessed. A difference between the published monthly mean CET for (say) September 2006 of 16.80 would have been 16.13 on the minute-by-minute basis and may not have been the warmest September on record on that basis

Also, the pronounced seasonality is another surprise as I would have anticipated quite a random effect. I can't see a simple explanation for the seasonality

Thinking about it in the abstract, it seems that in order for there to be a big difference between the minute-by-minute  average and the current method of taking the mean of the high and low water-marks each day, there needs to be a sharp (or quite sharp) spike or trough in the continuous temperature profile. This could be (as was said earlier) a clear period during the afternoon of a generally cloudy day for summer months to give a spike in the max or conversely a clear period overnight during a generally cloudy night in winter to give a sharp drop in temperature. These would have to be quite short-lived so as to affect the high/low water marks much more than the 24 hour average minute-by-minute

Also, going back to my earlier comment, if you just take the full minute-by-minute average I think you can go back to double-counting. For (say) an unusually cold morning in a generally mild spell in winter it could be the lowest temperature of two consecutive 24 hour periods 9AM to 9AM if it were the lowest at both 8AM and 10AM that morning. I think GW explained earlier that this effect is deliberately suppressed in the 1/2(max+min) approach currently in use

NB I think it would be better to compare the median of the minute-by-minute figures each day rather than the mean

 

Those differences in some of the months are higher than I would have guessed. A difference between the published monthly mean CET for (say) September 2006 of 16.80 would have been 16.13 on the minute-by-minute basis and may not have been the warmest September on record on that basis

Also, the pronounced seasonality is another surprise as I would have anticipated quite a random effect. I can't see a simple explanation for the seasonality

Thinking about it in the abstract, it seems that in order for there to be a big difference between the minute-by-minute  average and the current method of taking the mean of the high and low water-marks each day, there needs to be a sharp (or quite sharp) spike or trough in the continuous temperature profile. This could be (as was said earlier) a clear period during the afternoon of a generally cloudy day for summer months to give a spike in the max or conversely a clear period overnight during a generally cloudy night in winter to give a sharp drop in temperature. These would have to be quite short-lived so as to affect the high/low water marks much more than the 24 hour average minute-by-minute

Also, going back to my earlier comment, if you just take the full minute-by-minute average I think you can go back to double-counting. For (say) an unusually cold morning in a generally mild spell in winter it could be the lowest temperature of two consecutive 24 hour periods 9AM to 9AM if it were the lowest at both 8AM and 10AM that morning. I think GW explained earlier that this effect is deliberately suppressed in the 1/2(max+min) approach currently in use

NB I think it would be better to compare the median of the minute-by-minute figures each day rather than the mean

 

While this is a useful academic exercise, isn't the purpose of measuring the CET series in the way it is, to provide consistency with the existing 350 year dataset?

 

While this is a useful academic exercise, isn't the purpose of measuring the CET series in the way it is, to provide consistency with the existing 350 year dataset?

Correct. I think most people agree that a simple average of maximum and minimum is not a particularly accurate way of assessing the mean temperature for a day although it is a very good proxy. So if we were starting from scratch we would use a different system. But we have to maintain the current approach in order to enable comparisons to the historic CET series. Of course we could run both systems in parallel to start building up a record of the minute by minute or hour by hour data.

 

While this is a useful academic exercise, isn't the purpose of measuring the CET series in the way it is, to provide consistency with the existing 350 year dataset?

I think that's the point though. If you look at GW's data you can see that the consistency is questionable

btw I'm not seriously suggesting it would be practical to switch to minute-by-minute recording for every location but it does show up some anomalies

 

Unfortunately your entry was too late to be included in the list of predictions for this month. It will therefore count as a missed prediction for the purposes of the competition. The rules provide for a specific formula to be used to determine the prediction error for anyone who misses a prediction. This is based on the upper quartile figure of the range of prediction errors of everyone who made a valid prediction plus an amount of 0.4C. So I can't say exactly what your prediction error for this month will be until the final CET comes in at the end of the month. It looks like this month will have the lowest overall prediction error for the year so far. My best guess at the moment is that you will be allocated an error of 1.10C for this month.

Thanks GW for the full explanation.

Met Office Hadley    16.1c.     Anomaly       0.1c.   Provisional to 17th.

Metcheck                16.39c.   Anomaly       0.16c.

Netweather            16.91c.    Anomaly       0.72c.

Clevedon Weather   16.9c.     Anomaly       0.6c.

Mount   Sorrel         16.6c.    Anomaly       0.4c.

My    Mean              16.8c.    Anomaly     -0.2c.     

I have 16.2*C up to and including today's mean.

Uncannily close to the raw CET value, but the real CET is probably a little lower as usual (by a tenth or two).

Eyeing up a possible warm spell later next week - different origins to the one that affects the SE and E during the coming days, currently looking to affect a wider swathe of the UK but then so was this upcoming spell based on runs initiated prior to the weekend! 

I wouldn't mind being a little low (up to 0.5*C, say) with my 15.7*C 'gut instinct prediction' - the error would still be lower than that for the majority of my model-based predictions 

Met Office Hadley         16.0c.       Anomaly         0.0c.     Provisional   to  18th.

Metcheck                     16.35c.      Anomaly          0.12c.

Netweather                  16.82c.      Anomaly          0.63c.

Clevedon  Weather        16.9c.       Anomaly          0.6c.

Mount Sorrel                 16.6c.       Anomaly          0.4c.

My Mean                       16.8c.       Anomaly          0.7c.  

Met Office    Hadley            16.0c.      Anomaly     0.0c.    Provisional  to  19th.

Metcheck                           16.33c.    Anomaly     0.11c.

Netweather                        16.78c.    Anomaly     0.59c.

Clevedon Weather               16.8c.     Anomaly     0.5c.

Mount   Sorrel                     16.6c.     Anomaly     0.4c.

My    Mean                          16.7c.     Anomaly   -0.6c.

Met Office Hadley        16.1c.     Anomaly       0.1c.   Provisional to  20th.

Metcheck                    16.47c.   Anomaly       0.24c. 

Netweather                 16.85c.   Anomaly       0.66c.

Clevedon Weather        16.9c.    Anomaly       0.6c. 

Mount   Sorrel              16.7c.    Anomaly       0.5c.

My   Mean                    16.9c.    Anomaly      -0.7c.