|
|
 |
Vital Signs
Project: Siegel House Case Study
Weather Data
As part of the data collection at the Siegel House, we set up a weather station on the
Siegel's back patio so we would have some accurate site data to use in our analysis. We
had some problems with the data from that station, but luckily, we were able to obtain
comparable data from a nearby weather station on the University of California, Davis
campus. This page describes the problems we had with our site weather station data, and
why we believe using the UC Davis data is an acceptable substitute.
(Note: You may click on the charts on this page for more detailed views.)
We Missed Some Data
When we set up the weather station, we did not set the data collection interval correctly.
We intended to assign a collection interval of 30 minutes which would, over a 3 week
period, produce about 1000 measurements. These measurements would have fit in the memory
of the instrument because its limit is 1280 measurements.
Unfortunately, we forgot to set this interval when installing the weather station on site,
so the station used the previous setting of 10 minute intervals instead. Sampling 3 times
as often produces 3 times as many measurements, so all the data would no longer fit in
memory. Now, only a little more than 1 weeks worth of data would fit in memory. We ended
up with the last week of data because the station keeps recording even when the memory is
full. It overwrites earlier data as it continues recording.
We compounded this problem by forgetting to turn off the station when we picked it up. We
didn't turn the unit off until three and a half days later. All the while, the station was
recording useless data on top of our site data. We ended up with about 6 days worth of
data from the site.
We compensated for our mistake by obtaining a complete set of weather data from a nearby
weather station on the University of California, Davis, campus. The data from Davis,
though, has some drawbacks. Namely, the data is not exactly from our site. The Davis
weather station is about 5 miles away from Village Homes. This is not a significant
distance, but, as described later in this section, some local effects are evident when
comparing the two data sets. Also, the data from Davis is not nearly as detailed as our
site data would have been. The Davis data consists of average temperatures and insolation
values calculated hourly. Our intended sampling rate of every 30 minutes would have given
us twice as many data points, and our actual sampling rate ended up giving us six times as
many data points.
Below is a chart of the temperature data we recorded on site versus the temperature data
from the Davis weather station for the whole 3 week period. That we only ended up with the
last 6 days worth of data from the site weather station should be evident.
A graph of outdoor temperatures showing site and UC Davis data (18 k
gif)
Radiation Affected our Outdoor Ambient Temperature Data
The Davis data and the site data do not match well, as can be seen in the chart
immediately above. We were hoping they would match closely so that we could feel confident
about using Davis data in place of our site data. As it looks on the chart, there were
much larger ranges of temperatures experienced at the site.
After examining the data and discussing it with Professor Arens, we think that the much
larger swings on site are probably due to radiation striking the thermistor attached to
the weather station. We did put a radiation shield over the thermistor to prevent this
from happening, but the shield must have been positioned incorrectly to prevent direct
sunshine from striking the thermistor.
Given this conclusion, we decided that the Davis data does a much better job of
representing outdoor air temperature at our site than our site data, even though it was
from about five miles away. We will use the Davis temperature data in the rest of our
analyses.
We had to Scale the Data from the Pyranometer
As can be seen in the chart below, the insolation data from the weather station at the
site did not match the data from Davis either.
A graph of insolation showing site and UC Davis data (10 k gif)
We decided, after another discussion with Professor Arens, that the difference was due to
a calibration error from the pyranometer. To adjust the lower values from the site to
match the Davis values, we needed to scale our values by a calibration factor. But, we did
not calibrate the instrument to a known strength radiation prior to our field study. To
work around this, we chose a data point from the Davis data set and figured out the
scaling factor we would need to multiply the site data by to match the Davis data at that
same time. We then scaled all the site data by this same value.
In the chart below, the data point we chose to scale by is marked with a red circle and
labeled "Calibration point." The chart shows that once the data is scaled, the
site and Davis data match well.
Once scaled, the site data matches the UC Davis data (14 k gif)
Scaling the data reveals different qualities in the data due to their different reporting
techniques. The Davis data consists of hourly averages, while the site data is actual
values recorded every 10 minutes. The difference in data quality is most strongly
displayed on Monday, February 19, the second day shown on the chart. The Davis values show
smoother, lower values while the site data shows many peaks and a much greater range due
to sun striking the pyranometer intermittently through patchy clouds.
We use the Davis Data in place of our Site Data
Once the insolation data from our site collection was scaled, it matched the Davis well
enough for us to feel confident using the Davis data in the rest of our analyses. The
chart below summarizes the outdoor temperatures and insolation values for the 3 week
period of our study.
A composite graph of outdoor temperature, insolation, and weather
patterns for Davis, CA, Feburary 1996 (27 k gif)
One last transformation we performed on the data is displayed at the bottom of the chart.
We simplified the insolation curves by representing cloudy days with a cloud icon, partly
cloudy days with a partly cloudy icon, and sunny days with a sunny icon. We will use this
simplified weather representation in our analyses of overheating in the sunspace and
radiant temperatures in the sunspace. |
