Model Spectrum ***Experimental***
One plot combining the NWS forecast with the output from multiple weather models.

Elko Click Here
Graph Y-axis
Date, UTC
XML Graphic
Explanation of Graph

This plot shows a box and whisker timeseries of statistical data summarizing the output from many numerical weather models.
Often, the NWS forecast will fall within the range of the numerical models; however, sometimes models are incorrect. The NWS forecast or the observed value (what actually happens) may occur outside of the minimum and maximum model bounds depicted by the graph.

  • The green color text shows the inner box which contains 50% of the data for that period. This region is bounded by what are commonly referred to as the upper quartile or the 75th percentile and the lower quartile or the 25th percentile. Simply stated, 25% of the remaining data fall above the inner box, and 25% below.
  • The pink colors show where the highest and lowest model forecast values occur.
  • The dark blue color shows the NWS forecast. If you hover your mouse over the NWS forecast a window will pop-up with statistics summarizing the model data for that time period. This window also shows the time range for the point in Local Time.
  • The median (middle value of the data) is represented by a small gray horizontal line located in the inner box.
  • The degree of uncertainty (the color of the bars) is determined from the standard deviation of the models used.
  • Observed values (brown dots in the graph) may not agree with those published in the climate records. This occurs as a result of verifying the model points, which are grid cells, over an area. Some fields cannot be verified over an area and thus observations are not shown for these.


  • When the "Show Climatology" button is checked, climate data will display for those sites which have quality controlled climate records (This option only works for temperature). The overlay displays the record high (Red shade) and record low (Dark Blue shade) and normal high (Top of green shade) and normal low (Bottom of green shade) temperatures. These overlays are useful in determining if the upcoming temperatures are above or below normal and also highlight the potential for record breaking days.

    More info such as the year the record was broken and the observed values can be found by hovering over the dark blue NWS forecast dot.

    When the "Show Trends" button is checked, information from past model (grey bars) and NWS forecasts (light blue dots) are shown. The gray bars are just the box portion of the previous forecasts, i.e. the whiskers are omitted. All of the gray bars and light blue dots are displayed to the right of the time period in which they are valid for.

    Why is the NWS forecast sometimes outside of the Model Range?

    This can occur due to many reasons:

    1. Forecasters may tune into something the models can not resolve.
    2. The forecast has not been updated to reflect the latest model trends.
    3. The forecaster believes the current model runs are outliers compared to the prior runs.

    How are the Bar Colors/Uncertainty Determined?

    The colors used for the box and whisker plots are determined by the standard deviation of the current set of models. Basically, the colors define the current model spread which often is a good indicator of uncertainty. For example, if a field shows low uncertainty on the day 5 forecast, this means the current set of models converge on a value. In reality, there is the potential for the value to fall outside of this range. So be vigilant in assessing other factors when making decisions based on this display. Call your local forecast office if you need more help. The threshold values that define the uncertainty as low, moderate, or high were subjectively chosen for each field.

    What is Bias Correction?

    Applying a 30 day bias correction attempts to correct for systematic numerical model output biases. This often leads to a better model forecast when the upcoming weather pattern is similar to the previous 30 day pattern. Applying a bias correction involves applying a linear regression of past numerical model forecast errors to the current model forecast in hope of minimizing the expected forecast error. To calculate the bias from the linear regression method, a training period of 30 days is used. Because forecast errors occur every day, the training period needs to be sufficiently long to reflect the consistent biases inherent in a particular model. A 30 day training period is used because a period less than 30 days could be too responsive (and flip-flop between random errors), while a period much longer than 30 days would be too slow to respond to changes.

    What is a Numerical Weather Model?

    Numerical Weather Model - a mathematical simulation of change in the atmosphere. These models attempt to simulate the physical world by solving complex equations over time. The equations used in these models are not exact and therefore require assumptions in order to be solved. Assumptions impart a degree of uncertainty to the solutions. This webpage shows the distribution of the various solutions. To learn more about these models visit this link: Numerical Model Courses

    The values shown in these graphs represent point values (technically one 2.5 km x 2.5 km grid cell -- a very small area) and may not be representative of your exact location.
    Product Description Document
    Comments End: 2011-11-01
    Last Modified: 2012-07-30

    Webmaster
    US Dept of Commerce
    National Oceanic and Atmospheric Administration
    National Weather Service




    Disclaimer
    Information Quality
    Credits
    Glossary
    Organization
    Privacy Policy
    Freedom of Information Act
    About Us
    Career Opportunities