About Forecasts

This tool uses 48 experimental forecasts from the University of California Merced. These forecasts come from the official CFS Operational Forecasts from NOAA's National Center for Environmental Prediction(NCEP) but they have been interpolated to 1/24-deg (4-km) grid cells and bias corrected utilizing the mean climatology from the gridMET climate dataset.

Interpretation of Daily Forecasts

When you select an analysis of 'Daily Forecasts', you get a view of the daily time series of all 48 individual forecasts. In the example shown below for minimum daily temperature, the 48 pink lines represent these 48 forecasts. These forecasts all start on May 3rd and extend for the next 4 weeks, where the week ends are indicated in the dotted lines.

Initially all the forecasts start at roughly the same value on May 3rd but then start to diverge after a couple of days. One would expect this from forecasts that are all initialized with similar values.

Overlaid on the graph is a red line which represents the middle of the forecast which is computed from the median of the 48 forecasts but is in itself not a forecast and is just shown to give some idea as to the central tendency of the 48 forecasts.

Also overlaid on the graph is a black line, which is the median daily values as computed from historical observations (1980-2018). Additional gray shadings on the graph show historical percentiles from observations. These extras lines and regions allow you to put the forecasts in perspective. For example, we can see that most of the forecasts in the first couple of days after May 3rd are for 'above normal' minimum daily temperatures as most of the pink lines are above the black line. For the middle of the first week of May 34d, most of the forecast of minimum daily temperatures are very 'above normal' since most of the pink lines fall within the 70-90th percentile gray shaded region. This means that most of the forecasts rank in the higher end of what has been seen historically forecasting.

Interpretation of Daily Categorical Forecasts

When you select an analysis of 'Daily Categorical Forecasts', you get a view of the daily time series in forecast categories. In the example shown below for minimum daily temperature, historical percentiles from observations are used to create 6 categories:

Below 10 Percentile

Between 10 and 30 Percentile

Between 30 and 50 Percentile

Between 50 and 70 Percentile

Between 70 and 90 Percentile

Above 90 Percentile

On the y-axis is the percent of the 48 forecasts that fall into each category, where the total adds up to 100%. In the example, you can see that on May 3rd most of the forecasts are in the lighter to darker pink (or the 50-70 and 70-90 percentiles) and that only less than 6 % of the forecasts are for the below normal categories in shades of blue/green.

As you glance at these forecasts for the next 4 weeks, the amount of shades of red/pink versus blue/green give a quick indication as to if there are more forecasts for above or below normal minimum temperatures. In the example, in the first week there are more above normal forecasts and then again in the middle of the third week with more below normal forecasts at the end of the first week. Further, if you concentrate on the amount of dark pink and dark green-blue forecasts you see that 20% of the forecasts in the first week are for minimum daily temperatures in the highest 90 percentile of past observations. Finally, over weeks 2,3 and 4 there seem to be equal chances of the different categories as there are equal number of forecasts in each category, i.e about 10-20% in each category.

Interpretation of Weekly Forecasts

When you select an analysis of 'Weekly Forecasts and Skill', you get a view of the weekly time series of forecasts. In the example shown below, the graph shows the weekly average of minimum daily temperatures for week 1,2,3 and 4 from the initial forecast of May 2. Here you can see all 48 forecasts as the dots on the boxplot. The boxplot gives an indication of the middle (median, 50 percentile) of the forecast with the blue horizontal line. Also in the boxplot are the 25th and 75th percentile values from the 48 forecasts in the extents of the rectangle about the median. Finally in the boxplot are the lowest and highest value from the forecasts.

The horizontal pink line is the historical mean based on past obsevations from 1979-2019 to give perspective on how the weekly forecasts are above or below normal. Here we see that in week 1, the entire box is above the historical mean with only 25% of the forecasts below the mean. We can also see from the upward trend in the horizontal blue lines from the boxplot that the weekly forecasts generally are for increasing minimum daily temperatures. Finally, the spread of the dots increases from week 1 to week 4 showing how the uncertainty increases the longer the forecast.

Interpretation of Weekly Forecast Skill

When you select an analysis of 'Weekly Forecasts and Skill', there is a table of the skill of the weekly forecasts below the graph. In the example shown below, the table shows the skill for the week 1-4 forecasts for minimum daily temperature with forecasts starting April 1st.

Here the table indicates that there no skill in the Week 3 and 4 forecasts for the month of April at this location. However, there is good skill in the Week 1 and 2 forecasts for the month of April.

Below the characterization of the skill is the precise correlation value that was computed for these weekly forecasts for the given month from a comprehensive analysis of hindcasts from 1980-2010 compared to actual observations. The results shown here are for the 12-ensemble mean of the forecasts. It has been seen that a correlation r less than 0.2 indicates no statistically significant skill, which we label with a skill of 'None'. Correlations higher than 0.2 have varying levels of skill. See the documentation tab on Skill for more information about the hindcast analysis or our assignment of the strength of the skill.