<p>
The Solar Forecast Arbiter's preprocessing algorithms applied the following
operations to the data:
</p>
<ol>
  <li>Drop or fill missing forecast and reference forecast data points according to the report's forecast fill method selection.</li>
  <li>Apply the <a href="#data-validation-resampling">data validation tests</a> specified with <em>discard before resample</em> == True to the observation data and discard
    the matched data. This procedure is typically used to exclude erroneous observation data.
    </li>
  <li>Resample the observation data to the forecast interval length using the mean.</li>
  <li>Apply the <a href="#data-validation-resampling">data validation tests</a> specified with <em>discard before resample</em> == False to the observation data and discard
    the resampled intervals where the percentage of matched points exceeds <em>resample threshold</em>.
    This procedure is typically used to exclude intervals that contain valid but undesirable data, such as the hours of sunrise/sunset or hours that are mostly clear.
    This step also discards intervals where the percentage of points discarded before resampling exceeds the <em>resample threshold</em>.
    </li>
  <li>Align the time series of the resampled, filtered observations, the forecast, and, if selected, the reference forecast.
    Discard intervals that do not exist in all time series.</li>
</ol>
<p>
The <a href="#data-preprocessing-results-table">table of data preprocessing results</a>
summarizes the number of points matched by each of these operations.
</p>
