The Weather: Everyone talks about it – how well can we really predict it?

(forecast ¾”, actual up to 6”), the issue of how good forecasts can become local and/or national news (this year it is snowfall and icing in the Northeast and in the Carolinas).  In this exercise we will take look at how good predictions are.  To simplify the problem we will restrict ourselves to weather, and specifically temperature.  To do this, you are asked to look up data (preferably at the same time of day each day) on one of two major national forecast services, Weather Underground (www.wunderground.com)* and the National Weather Service (www.noaa.gov)**, for seven to ten days to see how good our predictions for temperature are.  The class will be split into four groups, half of the class will take each of these two services, and half of each of those will work with forecast high temperatures and half with forecast low temperatures.  We are in a highly variable season, so the results may be quite different.

Procedure:  

10. Select a time of day that you know you will be able to access the internet fairly regularly over a continuous seven to ten day period.  This is important – they change the forecast several times a day, so remember to record the time and try to look up the forecasts at about the same time each day.  This gives us consistency.  Select two days (at least 7-10 days in the future, exclusive of the day you start; that means between the 11^th and the 18^th, inclusive) for which you will check the forecast temperature.  Start anytime between February 5 and February 11 (but at least 7 days before the dates you picked, exclusive).  So, for example, if you selected February 15 and 16, you would need to start on Feb 9 and 10.

            What you are doing is picking two dates and tracking what the forecasts are for those two days from (more or less) a week out and seeing how they turn out.

1. At that selected time each day, go to the appropriate site (you probably should bookmark it once you have the right site) and look up the forecast temperature forecast Lansing Airport (use the links*^,** below) for the dates you have selected.  The reason we are using Lansing airport is that the official temperature site that is archived for Lansing by the National Weather Service is the airport location.  Use degrees Fahrenheit (°F).  Try following the directions below, if you have trouble, contact us immediately; do not put this one off.

1. Do this for the next 7-10 (Weather Underground) 6-7 (NOAA), recording the predicted high or low (whichever you are assigned) temperatures for each the two dates you selected and enter them into the table on the attached page.

1. At the end of the six-seven (ten if you are ambitious and the data are available) days, compare the predictions with the actual temperatures, which I will post on the class site (or you can look them up on a day by day basis on the NOAA web site**, but that has complications, see note below) and fill in the far right hand column.  Then analyze your results (see below).  While the data collection is a bit tedious, hopefully we will find out something about the reliability of temperature forecasts and (maybe) whose is the best.  I have minimized the time interval and number of days you need to track temperatures – a longer data set would be better, but impose more on you.

            If you are unsure, this will be briefly explained in class on February 5 or e-mail the instructor or talk to the classroom assistant.

*Weather Underground (https://www.wunderground.com/us/mi/capital–city.html) provides forecast data for ten days into the future.  The above link should come up “Capital City, MI.” Click on the tab labelled “10-DAY.”  This will give you a graph with forecast temperatures by day at the top.  Enter the forecast temperature (highs are in red and lows are in blue, as you have been assigned) for the dates you have selected on to the table on the attached page for those days.  Because of the connection now between The Weather Channel and Weather Underground, the forecasts the two use are almost identical, so we will stick with just Weather Underground.  WARNING:  Use the link above.  Entering Lansing for the location, usually gives you a private location.  You want KLAN (airport) or Capital City Station as the location.   I have also had issues with it defaulting to Capital, Costa Rica, so if you get tropical temperatures, something is wrong.

**NOAA. The National Weather Service of the National Oceanic and Atmospheric

Administration (http://forecast.weather.gov/MapClick.php?lon=–

84.61340620117186&lat=42.78305890174329) is the federal agency that monitors weather conditions and issues watches and warnings.  The NOAA (NWS) site has forecast temperatures going out for seven days, including the day you look it up, so it only gives you 6 days into the future.  At the top of the page it should start with “Current conditions at Capital Region International Airport (KLAN).”  Immediately below is a block titled “Extended Forecast for Lansing, Capital City Airport MI.” A set of icons with temperatures below them give you the next 3½ to 4 days of high and low temperatures (unfortunately, they are given by day of week, not date, so you will need to convert).  However, if you go further down to the  “Detailed Forecast” text below the icons, you can get all six to seven days.  Enter the forecast temperature (high or low as you have been assigned) from the text forecasts on to the table on the attached page for the dates selected.  WARNING:  Searching for the location from the NOAA home page will usually give you something in downtown Lansing.  Use the link above.  If that does not work, you can clock on the map next to the detailed forecast at the location of the airport.

Actual Temperatures.  These will be posted (with up to a day delay) on D2L.  WARNING:  There can be one major point of confusion if you look up actual temperatures yourself.  The low temperature will usually occur during the early morning.  Usually a forecast gives the high for the day, and the low for the following night (but that is usually going to be after midnight so it will officially recorded on the following day).  Depending your source and the time of day at which it occurs you may need to figure out what day the actual low temperature is for.  The high is seldom ambiguous.  I will post daytime high and nighttime low to avoid confusion; use my posts to avoid confusion.

1. Then answer the following questions (type or write neatly on a separate sheet(s) of paper or the back of the data sheet [with your name on every page!] – remember this is an analysis of the data and is somewhat subjective).  This requires you to EXAMINE YOUR DATA and use them to JUSTIFY YOUR CONCLUSIONS:

1. How good are the temperature forecasts? Compare the predictions with the actual.  Can you rely on them?  And how far in advance?  How can you quantify how good the forecasts are (for your data set) so the can be compared with those doing other sources?

2. Do the forecasts improve with time (that is, are they more accurate the closer you are to the day in question)?  If so, by how much?   How much variability is there?  Is there a trend?  Sometimes the conclusion might be ‘no.’  Suggest some reasons why.

3. If you want to plan an activity that requires that we know the temperature to within 2 degrees, based on this data, how far in advance can you safely schedule the activity (or can you even do so)?