Re: [EnergyPlus_Support] Re: Long Term Weather Data Predictions

Regarding this thread, it might be of your interest the research done by Dr. Halimi Sulaiman "Probabilistic assessment of discomfort risk in buildings" which I will be presenting on her behalph at BS2013.

Because of luck of reliable weather data in certain regions, Dr. Sulaiman proposes a methodology of assessing discomfort risk by combining whole building analysis and statistical synthetic weather data (eg. 1000 years, generated from 5 measured). Runs the model for 1000 years and ends up with output that considers statistical variations in weather, including generated extremes. The output now is the whole probability distribution spectrum instead of a single deterministic results. In this way you have more information to asses risk involved in predictions of how the building is going to behave, not only internal comfort; energy, operation cost, etc.

Best,

Santiago.

Papers Abrstract Follows,

2013/7/30 Leandro Biondo <leandromet@xxxxxxxxx>

Jim and Joe,

You can run a weather model for as long as you want, and it may help the analysis for weather purposes, that would be to have an idea of what to expect for the next few weeks or couple of months. Normally the atmosphere models respond very well for up to 8 days ahead, then it degrades or deviates more but still can be used as a good tendency information for 10 to 15 days more, then it tends to deviate more and starts to be a rough tendency that should be looked carefully up to 45-60 days.

The problem is again that the errors you will find on the weather model may be large, depending on the variable you are interested in, for example if you want to have an idea of the dry bulb temperature for the next month or so, it can be closer to the model, if you need to know if the rain will supply your building water systems for the same month it can be much worse.

All depends on the model you are looking at and what you want from it. There are some approaches that are better to get the big picture with lower resolution (like hurricane prediction and cold/heat waves) and others that are better to get detailed information (like rain ad temperature on your neighborhood). Here in Brasil, for example, the center for weather prediction (cptec) does a discussion every 3 months on what to expect for the next season based on lots of runs from lots of models, and the prediction is given on percentages compared to the climatology means (like 20|55|25, would be 20% chance to be 2 units lower than the mean, 55% equal and 25% higher). Other countries do similar things and you can use this on a E+ model, but you need to create your data considering this prediction, there won´t be consistent data on an hour-to-hour resolution for the next 90 days.

About the "change" from older meteorologists to younger, this sounds funny and probably is wrong (I am sure it is wrong for Brasil, US, Australia and most the EU) to think the newer computer models may influence too much on getting the long term predictions worse. First because most "old" meteorologists already use computer models for a long time and they are the ones that helped creating and improving the new ones and second because today we have newer models and better knowledge of the atmosphere.

There could be a loss of quality on the discussions and conclusions but it is more likely that the longer-term predictions can get worse thanks to the changes on the environment, like heat island effect, forests changes and global warming. Also we might expect today a better quality and precision prediction than people used to expect on past decades, this brings more criticism that helps improving the area.

Anyway, for any E+ run "in the future" you should expect better results with shorter periods and may create biased weather data for scenarios comparison on longer periods.

sheers,

Leandro Biondo
Environment Analyst / Meteorologist
Brazilian Forest Services - Environment Ministry