Duration

Definition

Fall pulse flow duration is in most cases calculated as the number of days from the beginning of the flow event until the flow event peak. See step 6 in the calculation steps for the exception to this rule. This metric is measured in number of days.

Steps

1. The start date calculated for the fall pulse flow timing and the Gaussian filtered flow data are used as inputs for the duration algorithm.

   def calc_fall_flush_durations_2(filter_data, date):

2. Set a requirement for slope and magnitude on either side of the event peak (left-side rising limb and right-side falling limb). The slope threshold shown in the code snippet below (der_percent_threshold). The end of the peak on each side is set when slope decreases below the defined thresholds. The derivative threshold is higher for the left side (requires a steeper rising limb) and lower for the right side (requires a less steep falling limb). This is based on the typical shape of a storm-caused peak in flow.

   The flow percent threshold (flow_percent_threshold) is a magnitude percentile requirement for the left or right side of the fall pulse flow peak. For the left side rising limb, the bottom of the rising limb must be below a relative magnitude threshold (default 80th percentile) based on all flow values on the left side of the peak. The same requirement is used for the right side of the peak.

   """Left side sharp"""
      der_percent_threshold_left = 50
      flow_percent_threhold_left = 80
   
      """Right side mellow"""
      der_percent_threshold_right = 30
      flow_percent_threhold_right = 80

3. Find the valleys on either side of the fall pulse flow peak, using the minimum values from the peak detection algorithm. If a valley is within 10 days of the fall pulse flow peak on either side, then automatically set that valley as the end of the fall pulse flow. This could occur on one side of the fall pulse flow peak and not the other side. If there are no identified valleys on either the left or right side of the fall pulse flow peak, then continue to the next step to identify the valleys (the right and left durations are set as zero by default, then later reassigned when the actual duration is found).

    left_maxarray, left_minarray = peakdet(filter_data[:date], 0.01)
    right_maxarray, right_minarray = peakdet(filter_data[date:], 0.01)
   
    if not list(left_minarray):
        left = 0
    else:
        left = int(left_minarray[-1][0])
   
    if not list(right_minarray):
        right = 0
    else:
        right = int(date - 2 + right_minarray[0][0])

4. If the left side of the fall pulse flow has not already been detected as a valley within 10 days of the fall pulse flow peak, find the left side through the following requirements: 1. The slope of the rising (left-side) limb falls below the derivative percent threshold. 2. The flow magnitude of the far left side falls below the percentile threshold.

     if date - left > 10:
                 """create spline, and find derivative"""
                 x_axis_left = list(range(len(filter_data[left:date])))
                 spl_left = ip.UnivariateSpline(x_axis_left, filter_data[left:date], k=3, s=3)
                 spl_first_left = spl_left.derivative(1)
   
                 """check if derivative value falls below certain threshold"""
                 spl_first_left_median = np.nanpercentile(spl_first_left(x_axis_left), der_percent_threshold_left)
   
                 """check if actual value falls below threshold, avoiding the rounded peak"""
                 median_left = np.nanpercentile(list(set(filter_data[left:date])), flow_percent_threshold_left)
   
                 for index_left, der in enumerate(reversed(spl_first_left(x_axis_left))):
                     # print(der < spl_first_left_median, filter_data[date - index_left] < median_left)
                     if der < spl_first_left_median and filter_data[date - index_left] < median_left:
                         left = date - index_left
                         break

5. If the right side of the fall pulse flow has not already been detected as a valley within 10 days of the fall pulse flow peak, then find the right side through the following requirements: 1. The slope of the falling (right-side) limb falls below the derivative percent threshold. 2. The flow magnitude of the far right side falls below the percentile threshold.

     if right - date > 10:
                 x_axis_right = list(range(len(filter_data[date:right])))
                 spl_right = ip.UnivariateSpline(x_axis_right, filter_data[date:right], k=3, s=3)
                 spl_first_right = spl_right.derivative(1)
   
                 spl_first_right_median = abs(np.nanpercentile(spl_first_right(x_axis_right), der_percent_threshold_right))
                 median_right = np.nanpercentile(list(set(filter_data[date:right])), flow_percent_threshold_right)
   
                 for index_right, der in enumerate(spl_first_right(x_axis_right)):
                     # print(date+index_right, der < spl_first_right_median, filter_data[date + index_right] < median_right)
                     if abs(der) < spl_first_right_median and filter_data[date + index_right] < median_right:
                         right = date + index_right
                         break

6. Once the left and right ends of the fall pulse flow have been detected:

   • Set the duration as the number of days between the left side of the fall pulse flow peak and the fall pulse flow peak:

     if left:
     duration = int(date - left)

   • If there is no left side identified, set the duration as the number of days between the fall pulse flow peak and the end of the falling limb of the fall pulse flow peak:

     elif not left and right:
     duration = int(right - date)