Simulation or historic weather files for Northern Pakistan

5 posts / 0 new
Last post

Hi everyone,

I'm working on a project to design affordable housing and temporary
shelters for refugees and disaster recovery in Pakistan. One of our target
areas is the remote northern areas of the country, called Hunza Valley.
Some larger cities here include Gilgit and Skardu. The nearest weather
files available from EnergyPlus are 200-300 miles away. The Hunza Valley
region is mountainous, so given the influence of altitude and microclimate
effects I'd like to at least find historical weather data to validate
simulation weather file selection.

Let me know if you have any thoughts on where I can get weather data. It's
a student project, so we don't have budget to purchase weather files from a
commercial service.

Best,
Arfa

*Arfa Aijazi*
PhD Student, Building Science and Sustainability
Center for the Built Environment | Architecture
| UC Berkeley

(m) 703-508-6535 | (e) arfa at berkeley.edu

Arfa Aijazi's picture
Offline
Joined: 2018-07-02
Reputation: 0

Arfa,

We looked into data availability and there are few hours of data (and no
nighttime data) for those locations. Please contact me by email -- there
are other options (no cost).

Dru

On Fri, Mar 1, 2019 at 2:36 PM Arfa Aijazi via Bldg-sim <

Drury B Crawley's picture
Offline
Joined: 2011-10-02
Reputation: 0

As you probably realize, this is an? area of the world with almost no observed weather
data.? The only remotely usable data I have from the ISD (Integrated Surface Hourly
Database)? are nine months of data (April - October) 2011 in Skardu. However, there are
satellite-derived solar radiation and computer-modeled weather conditions from the MERRA-1
Climate Forecast Model for anywhere in South Asia from 2000 through 2014 available as part
of NREL's National Solar Radiation Database on line at https://nsrdb.nrel.gov/nsrdb-viewer
.? Unless you want to spend several weeks doing so, I would not suggest that you delve
into either data sources, nor would I recommend using the modeled weather data from
MERRA-1, which I have found can be wildly off, especially in South Asia locations.

Since I've already got the data for Skardu for both data sets, I can see if I can use the
partial 2011 data for Skardu to bias correct the MERRA-1 data,?? which would then get you
a weather file for 2011.? I'm not promising anything though, as I've got a lot of others
things to do as well.

Joe

Joe Huang
White Box Technologies, Inc.
346 Rheem Blvd., Suite 205A
Moraga CA 94556
yjhuang at whiteboxtechnologies.com
http://weather.whiteboxtechnologies.com for simulation-ready weather data
(o) (925)388-0265
(c) (510)928-2683
"building energy simulations at your fingertips"

Joe Huang's picture
Offline
Joined: 2011-09-30
Reputation: 406

After a number of hours I've created a "typical year" weather file for Skardu using the
data source that I describe below, along with some tweaking, or bias correction if I want
to sound technical, using climatic information from Wikipedia.

The file is available for free at this link on my company server:

http://weather.whiteboxtechnologies.com/custom-np-link-download?F1=7806138414*PAK_SKARDU_415170_35.25-75.65mod_TYP.zip

For those interested, I like to explain the technique I used that I had? thought about for
more than a year, but never actually tried out until now, with results that I'm very
pleased to see, although those who are more critical might say that there's no hard
evidence like actual measurements, but then that's a circular argument because this
technique is being used only in cases when there's no measured data!

The two sources I'm using are (1) satellite-derived solar radiation, and (2)
meteorological reanalysis data.? The first is pretty self-explanatory, but the second
needs a bit of explanation. Reanalysis is using massive climate forecasting models in a
retrospective mode after the time is past and all the observational data have been
incorporated.? There are several of these reanalysis efforts around the world, whose
outputs cover the entire globe at a half-degree grid and span a time period going back a
couple of decades. Look up reanalysis on the Web if you want more information.

While the information is public, both of these types of data can be difficult to access,
especially reanalysis, but in this case the location in Pakistan is covered by NREL's
National Solar Radiation Database (NSRDB) courtesy of a India - US solar project that
ended 5 years ago.? Since I use the NSRDB quite extensively, I've already downloaded? both
the satellite solar and MERRA reanalysis data (NREL puts them together on the same file)
for Skardu for 2000-2014. There are no issues with the satellite-derived solar, but the
MERRA reanalysis data can be quite different from measured data, and particularly off for
locations in India and Pakistan.? Therefore, I've been hesitant in using it verbatim, but
found it helpful to fill data gaps once the data has been correlated with some type of
"ground truth".

In this case, the only measured data, a few months in Skardu in 2011, was also found to be
very poor and unreliable.? Turning to the Wikipedia, I found the following summary
climatic table for Skardu:

I then calculated the same monthly statistics from 15 years of the MERRA reanalysis data,
and found that monthly temperatures were about 12 C too cold, although the daily range
(difference between the daily max and min) seemed quite similar.

The average monthly relative humidity at 5 pm were also quite different except during July
and August.

What I then did was to adjust the dry bulb temperatures by calculating a difference in the
mean daily average per month, and a percent different in the mean daily range by month,
and then used these to adjust or bias correct the MERRA data,? producing this plot for all
15 years of data:

Adjustment for the humidity must be done after the dry-bulb temperatures, since relative
humidities are dependent on the air temperature.? The other complication is that since the
raw weather file contains only dewpoint temperatures,? a trial and error process with the
dewpoint temperature had to be done.? After three iterations, the humidities also match
quite closely.

SInce the NREL files were given for India time, another small routine was used to shift up
the data by two hours, which also brought the solar radiation in line with the calculated
sun angles.

The last step was to create a typical year weather file from these 15 years of weather
data. That was quite quick because I have all the software in house.

A final look at the "typical year" file showed that it all looked reasonable.? There's no
real way to say how good is this file, but the solar should be quite reliable, and the
monthly statistics are almost exactly the same as those from the Pakistan Met Office.

Joe

Joe Huang
White Box Technologies, Inc.
346 Rheem Blvd., Suite 205A
Moraga CA 94556
yjhuang at whiteboxtechnologies.com
http://weather.whiteboxtechnologies.com for simulation-ready weather data
(o) (925)388-0265
(c) (510)928-2683
"building energy simulations at your fingertips"

Joe Huang's picture
Offline
Joined: 2011-09-30
Reputation: 406

Joe, thank you very much this effort and the detailed explanation!

*Arfa Aijazi*
PhD Student, Building Science and Sustainability
Center for the Built Environment | Architecture
| UC Berkeley

(m) 703-508-6535 | (e) arfa at berkeley.edu

On Mon, Mar 4, 2019 at 6:15 AM Joe Huang via Bldg-sim <

Arfa Aijazi's picture
Offline
Joined: 2018-07-02
Reputation: 0