The forecasts presented on these pages are derived from three major weather forecast products supplied by the national weather services. (For more details, see: Weather Forecasting For Radio Astronomy: Part I The Mechanisms and Physics). The downloaded forecasts are either ground-level "Point" forecasts (added on April 2, 2009) or the results of forecasts of vertical weather profiles.
"Point" forecasts use sophisticated interpolation to provide details for any location within the US at a resolution of 5 km, extend for 6 days with a time resolution of 1 hr, and are updated on the average of nine times per day. These are supposedly the highest resolution forecasts and, as such, should be the best predictors of the weather in Green Bank. The 5 km cell I have chosen is centered at latitude=38.429, longitude = -79.840, just south of the GBT. "Point" forecasts provide only "ground-level" values, the quantities that humans are mostly interested in (precipitation, temperature, etc.), but do not include any information from which opacities can be determined. The forecasts are downloaded and archived as often as the forecasts are updated.
The majority of the results presented here are from 3.5 and 7.5 day forecasts of vertical weather conditions. The forecasts are based on computer models that use balloon soundings and GOES satellite soundings as input. These very detailed forecasts are only available for a limited number of locations. The three nearest to Green Bank are Elkins, WV, Lewisburg, WV, and Hot Springs, VA, all about 45-60 miles from the observatory. By 'averaging' the data from the three sites, it is hoped that the results represent the conditions over the observatory.
The 3.5 forecasts are based on the "North American Mesoscale" (NAM, formally known as ETA) model. ".... The model is run four times a day (00, 06, 12, 18 UTC) out to 84 hours. It is currently run with 12 km horizontal resolution and with 1 hour temporal resolution, providing finer detail than other operational forecast models." ( https://en.wikipedia.org/wiki/North_American_Mesoscale_Model).
The 7.5-day forecasts are based on the the first half of the 16-day GFS (Global Forecast System, previously AVN) atmospheric models. The half of the model I use has a 35 km horizontal resolution, a 3 hour temporal resolution, and is updated twice a day ( https://en.wikipedia.org/wiki/Global_Forecast_System). The courser resolution of the GFS model suggests that one should only use the GFS results only beyond the 3.5 day cutoff of the NAM forecasts.
Each vertical forecast consists of a time series of ground weather conditions (pressure, temperature, wind speed and direction, dew point...), gross weather conditions (cloud cover, total precipitable water, ...) and, most importantly for the work presented here, pressure, humidity, dew point, cloud fraction, ... as a function of time and height above the above mentioned towns. The models divide the atmosphere into ~64 layers that extend to over 20,000 m. Since these forecasts and models are what the weather services use to predict weather for our area, including the input to the "Point" forecasts, it's one of the best data sets for local forecasting. The data are downloaded from ftp://ftp.meteo.psu.edu four times a day and archived for future use.
The calculations performed on these data, unlike most other attempts at calculating weather conditions for cm- and mm-wave observations, are not based on a model atmosphere with assumed pressure heights or temperature lapse rates. Rather, calculations are performed for each layer of the atmosphere using the forecasted conditions for that layer, thereby eliminating any assumptions concerning the atmospheric profile above the observatory. Calculations are performed every two hours in TCL by a CLEO application (http://www.gb.nrao.edu/~rmaddale/GBT/CLEOManual/index.html) which also automatically generates these web pages.
The number of results that can be produced by the CLEO application is rather extensive and only a subset of what is possible is presented here. There's usually about thirty graphs that are given here that I hope help our observers plan their observations and help with their data calibration. The contents of these pages will change as we learn more about what our observers and staff require.
As I've added various weather products to this work, I have also
maintained an archive of the NWS forecasts. In June 2005 I
released a CLEO graphical interface that allows users
the ability to generate graphs for just their desired weather
parameters from wither the current or archived data.
One can use the archive of forecasts either to determine weather
parameters for past observations or to accumulate weather
statistics. Because the forecasts are generated within a couple
of hours before the time for which they apply, the archive can be
considered a reasonably good representation of the actual weather
conditions. Furthermore, starting in Sept 2009, I have been
archiving "Rapid Update Cycle" (RUC) vertical forecasts.
According to
Wikipedia:
"The RUC runs at the highest frequency of any forecast model at the
National Centers
for Environmental Prediction
(NCEP), assimilating recent observations to provide very high frequency
updates of current conditions and short-range forecasts. This update
frequency is still only once an hour (the standard interval for
ASOS
observation reporting), and with current computational limitations and
the time required to assimilate all of the data, there is approximately
an hour delay in producing the forecasts. Because of this, it is common
practice to use a one-hour forecast from the RUC as a current analysis,
as the one-hour forecast comes out only a few minutes before the time
it is forecasting for. There is also little possibility for error in a
one-hour forecast, meaning that the RUC's one-hour forecast will not
usually vary greatly from the actual state of the atmosphere at that
particular point in time."
The following table shows the extent of the archives for the
various types of forecasts I have mentioned:
| Forecast |
NAM | GFS3 |
RUC |
Point Map |
|---|---|---|---|---|
| Archive
start date |
26 Apr 2004 |
17 Sept 2007 |
15 Sept 2009 |
13 Mar 2009 |
The following sections give a general guide as to what is usually available as links off of this web page.
The displayed graphs consist of wind speeds (average ground level winds, average at 75-m, and gust) and temperatures as a function of UT, which should help observers avoid times when the weather conditions either preclude observing (too cold or windy for the telescope to operate) or wind conditions that are unsuitable to high frequency observing. I include estimates of precipitable water vapor, which can be used to predict roughly whether the opacity will be good for a particular observing frequency. A better measure of opacity is provided off of the "Opacity" or "Relative Effective System Temperatures" links above. also provide plots of dew point, humidity, and pressure. Some quantities are provided by only the "Point" forecasts, some quantities are provided only by the vertical profile forecasts, and some quantities are provided by both. All "Point" forecast plots will have "Green Bank" included in the legend to the plot.
Added on Feb. 25, 2009. The displayed graphs consist of various values as a function of UT. There are plots of fractional cloud cover in the low, middle, and upper atmospheres, and a plot of the total cloud cover. Usually lower-level and often middle-level cloud cover will influence continuum mapping projects. Plots also show the precipitation rates per hour in units of kg/m2. This is essentially the same as rainfall in mm/hr. Since snow is less dense than water, you can expect snow fall rates to be on the order of 10 mm for each kg/m2. Finally, I provide graphs that indicate the expected precipitation type, precipitation probabilities, the likelihood of thunder, ceiling height, and visibility. All "Point" forecast plots will have "Green Bank" included in the legend to the plot.
Opacities are derived via the MWP model of Liebe 1985), with some modifications by Danese and Partridge (1989). Opacities are calculated based on the contributions from 40 O2 resonance lines, three H2O resonance lines, H2O continuum, and the dry air. The model should be accurate for most purposes up to 120 GHz.
As of Sept, 2005, I have added in the contributions to opacity from hydrosols (fog, cloud water droplets, etc.). I'm using the cloud model described by Schwab and Hogg (1989), combined with the Liebe hydrosol continuum model. It's a compromise technique and assumes a cloud is present in any layer of the atmosphere where the humidity is 95% or greater. The thickness of the cloud layer determines the density of water droplets -- 0.2 g/m3 for clouds thinner than 120 m, 0.4 g/m3 for clouds thicker than 500 m, with linearly-interpolated densities for clouds of intermediate thickness.
The opacity plots typically provided are:
Once the opacity is calculated for each layer in the atmosphere, one can then use radiative transfer to derive the contribution to the system temperature that is due to the atmosphere. This quantity is helpful since Tsys directly influences the noise in an observation. The derived graphs show only that part of Tsys that is contributed by the atmosphere. They do not include the contributions from the receiver, spillover, and the 3K microwave background.
Plots consist of:
Neither system temperatures or opacities alone determine the affects of the atmosphere on observations. Firstly, the atmosphere attenuates the atmospheric signal and, secondly, the atmosphere emission can be a significant contributor to the system temperature. Both the atmospheric attenuation and emission are important factors in determining the amount of observing time needed to achieve a certain signal to noise.
I define the Effective System Temperature (EST) as Tsys*exp(Tau*AirMass). EST is proportional to the square root of the integration time needed to achieve a desired signal to noise. Tsys is the sum of the contributions from the atmosphere at the observing elevation (Tatm*(1-exp(tau*AirMass))), spillover (assumed to be 3 K for the GBT), the cosmic microwave background (3 K), and the receiver. Thus, EST is receiver, frequency, telescope, elevation, and weather dependent. (Note that I am not including in EST the contribution of any strong continuum source or background galactic emission.)
I next define the Relative Effective System Temperature (REST) as EST / EST0 where EST0 is the value of EST under the best possible weather conditions for Green Bank for the same elevation at which EST is determined. REST is exactly equal to sqrt(t/t0), where t is the integration time needed to perform an observation under the current weather conditions and t0 is the integration time needed under the best possible weather conditions. I used the weather conditions between 1 Oct 2004 and 1 May 2005 to determine values for EST0.
Plots on the forecast page consist of:
It might be useful to consider the following guiding principles in using REST plots.
Classically, observers use a measure of Tsys as a function of elevation (a 'tipping') to determine atmospheric opacity. The usual problem is that one has to assume a representative temperature (Tatm) for the atmosphere in this analysis. The degree to which the assumed Tatm is wrong is directly reflected in the inaccuracy of the derived opacity. These assumptions are no longer needed since one can determine Tatm from vertical weather data.
Plots consist of:
Probably only staff will find the estimates of refraction interesting. Like the other calculations presented here, one can use vertical profiles and in-situ measures of the index of refraction to derive the amount by which the telescope's pointing needs to be adjusted for the difference between the refracted and true elevation of a source.
I provide a comparison between the refraction derived from vertical profiles to two other methods that are based on ground-level weather data. The first is that produced by the SlaLib refraction package which uses a standard lapse rate and pressure heights, with ground level weather parameters (temperature, pressure, and dew point). The other uses an empirical fit to the model presented in GBT Memo 112 (Maddalena 1994) and the same ground weather values. The latter model, except for the empirical fit, is that which is currently in use by the GBT.
Additionally, the GBT has an interesting optics problem due to refraction. Richard Simon (1994, private communication) first pointed out that, since the top and bottom edge of the GBT are at two very different elevation, the atmospheric paths for rays that hit the top of the dish will pass through a different atmosphere than the rays hitting the lower edge. This differential refraction will alter the shape of objects that are observed close to the horizon, essentially elongating sources in the elevation direction. The telescope will have a virtual astigmatism that is due solely to the atmosphere. One of the plots I present illustrates the magnitude of this differential refraction on source size for each meter of aperture. The amount of virtual astigmatism is very weather dependent and poses a challenge to those wanting to observe close to the horizon. One way to correct for this astigmatism is to properly deform the telescope's shape out of a parabola in the elevation direction.
The refraction plots typically provided are:
D. S. Balser,
DSS memo 6.2, Forecasting
the Wind Speed in Green Bank,
2010
Bufkit (http://training.weather.gov/wdtd/tools/BUFKIT/)
(and associated web pages.)
B. Butler, "Precipitable Water Vapor at the VLA -- 1990 - 1998",
1998, NRAO MMA Memo #237 (and references therein).
J.J. Condon and D. S. Balser, DSS memo 5.3, Dynamic Scheduling Algorithms, Metrics, and Simulations, 2010.
L. Danese and R.B. Partridge, "Atmospheric Emission Models:
Confrontation between Observational Data and Predictions in the 2.5-300
GHz Frequency Range", 1989, AP.J. 342, 604.
K.D. Froome and L. Essen, "The Velocity of Light and Radio Waves",
1969, (New York: Academic Press).
H.J. Lehto, "High Sensitivity Searches for Short Time Scale
Variability in Extragalactic Objects", 1989, Ph.D. Thesis, University
of Virginia, Department of Astronomy, pp. 145-177.
H.J. Liebe, "An Updated model for millimeter wave propagation in
moist air", 1985, Radio Science, 20, 1069
R.J. Maddalena "Accurate Weather Forecasting for Radio Astronomy", Bulletin of the American Astronomical Society, Vol. 42, p.406.
R.J. Maddalena "Refraction, Weather Station Components, and Other
Details for Pointing the GBT", 1994, NRAO GBT Memo 112 (and
references therein).
R.J. Maddalena Weather Forecasting For Radio Astronomy: Part I The Mechanisms and Physics, Green Bank Technical Seminar, August 1, 2008.
R.J. Maddalena The Effects of the Atmosphere and Weather on Radio Astronomy Observations-- Lecture given during the July, 2011 Single Dish Summer School in Green Bank, WV (PowerPoint, 2.4 MBytes). Also in PDF (2.0 MBytes).
J. Meeus, "Astronomical Algorithms", 1990 (Richmond:
Willman-Bell).
K. Rohlfs and T.L. Wilson, "Tools of Radio Astronomy, 2nd
edition", 1996, pp. 165-168.
J.M. Rueger, "Electronic Distance Measurements", 1990 (New
York: Springer Verlag).
F.R. Schwab and D.E Hogg, "Analysis of Radiosonde Data for the MMA
Site Survey and Comparison with Tipping Radiometer Data" (1989), from
the IAU Symposium on "Radio Astronomical Seeing", pp 116-121.
Slalib (http://star-www.rl.ac.uk/star/docs/sun67.htx/sun67.html/)
W.S. Smart, "Textbook on Spherical Astronomy", 1977, (New
York: Cambridge Univ. Press).
Weather Watcher (http://www.singerscreations.com)