This plot shows a box and whisker timeseries
of statistical data summarizing the output from many numerical weather
models. 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. 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:
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 flipflop 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.



