
Activity Recorded by Data Logger  by Ralph Taylor 



Introduction 
Old Faithful Geyser has been monitored electronically using data loggers at least since 2000 (the first year for which I have found data files). There was an infrared system that recorded temperatures on a paper chart before that, but the data is not available in electronic form. The data logger for Old Faithful measures the runoff water temperature at a point about 20 meters (a rough estimate at best) from the vent toward the west. The sensor picks up preplay and the eruption start, with a delay of between one and ten minutes from the visual times recorded in the OFVC logbook. The mean delay for a period of several months in 2001 was found to be 3m46s with a standard deviation of 1m05s. The position of the sensor is far enough from the geyser that the eruption duration cannot be determined from the temperature trace.
It has been known for many years that Old Faithful's eruptions can be predicted by timing the duration an eruption and using a regression formula to find the expected interval to the next eruption. Since the data logger information does not contain duration information, it is not possible to find the regression data from the data set or to classify eruptions by duration.
In recent years Old Faithful has had increasing numbers of "long" eruptions (more than about 3m30s duration) and very few "short" eruptions. There are few eruptions of intermediate duration. Since this behavior pattern became established, predictions have been made using one prediction interval for short duration eruptions and another for long duration eruptions (with some special handling of the few intermediate duration eruptions).
For the purpose of the analysis presented here, I have only interval data, so I have classified eruptions as long or short based only on the interval.



Activity in 2011 
The overall statistics for 2011 are shown at Old Faithful 2011 Statistics. This summary gives the statistics for all intervals, and separately for long and short intervals for the current yeartodate, the past month, and the past week. A pdf of this summary is at Old Faithful Recent Activity Summary.




The interval graph shows all of the intervals for 2010. The orange triangles mark instances of short intervals. The graph is updated about every six weeks.

Click for a larger image



The next graph shows the intervals for the three months preceding the last data download at an expanded time scale. Again, the orange triangles denote short intervals, which often occur in clusters separated by days of exclusively long intervals.
In recent months Old Faithful is showing signs of returning to an old pattern of alternating long and short eruptions (and therefore intervals).

Click for a larger image



This graph shows the activity for the month preceding the last download. The graphs all include a blue line showing all intervals, a green or yellow line showing the moving median interval. The orange triangles at the 43 minute levels denote the short intervals. In the past few years short intervals have tended to occur in clusters separated by periods of days or even a week with no intervals less than 1h15m. In the years that I have been analyzing Old Faithful, there have been no recorded instances of successive short intervals.

Click for a larger image



The interval distribution shows that the greatest number of intervals were between 1h30m and 1h35m, which contains about a third of the long intervals. The category between 1h0m and 1h05m contains most of the short intervals.
Note that in the histograms displayed here the labels shown on the Xaxis represent the upper boundary of the class, not the midpoint. Geyser times are traditionally truncated. The graph at the right has class widths of 5 minutes. The bar appearing above the label "01:35," for example, contains intervals from 1h30m through 1h34m (90 to 94 minutes).

Click for a larger image



The second histogram, which shows the distribution of intervals with a resolution of one minute six seconds shows the main cluster of intervals around the 90 minute mark.
Again, recall that the label represents the high limit of the class, for example, the label "1:33" is the upper boundary of a class that contains intervals ranging from 1h32m1s to 1h32m59s, the class labeled "1:33"
contains intervals of 1h32m (92 minutes).
I shifted this histogram to 1m6s to match the logging interval I adopted in 2010 to allow for fullwinter runs of the data logger.

Click for a larger image



The final chart in this section is a representation of where the actual intervals occur in the prediction window. The chart includes results for the past week and month.

Click for a larger image


Activity since 2000 
Old Faithful has been consistent in its behavior for the years covered by the data logger information. A simple plot of interval vs time (shown at right) does not reveal any noticeable change. There are two large gaps in coverage, one in early 2002 and one in early 2005, both caused by equipment problems. A few shorter gaps resulted when downloads were delayed due to winter conditions and the logger memory filled.

Click for a larger image



More detail about the longterm variation in a geyser's behavior is often shown by a moving median graph, which eliminates the very long and very short intervals and gives a good picture of the overall behavior. For Old Faithful since 2000 this plot (at right), shows a slow increase in intervals throughout 2001, amounting to four minutes increase in that year. Since that time, the median interval has remained fairly steady with some fluctuation.

Click for a larger image



Another way to look at the long term behavior is to plot a graph of the interval distribution for the different years. The graph at the right shows this plot. The shape of the distribution has remained generally the same since 2000, confirming the lack of change in Old Faithful in that time. The curve for 2009 shows a slight shift to shorter intervals.

Click for a larger image



Finally, the chart at right shows how the maximum, minimum, mean, and median intervals for each month have changed. Since April of 2002 there was essentially no change in Old Faithful's behavior until 2007 when the mean interval decreased slightly. Note also that there is no visible seasonal change in Old Faithful's intervals.
Although past earthquakes (notably the 1959 Hebgen Lake quake and the 1983 Mt. Borah, ID quakes) were followed by noticeable increases in Old Faithful's intervals, the Denali AK quake in November of 2002 did not result in a measurable change in Old Faithful, even though it did result in changes in several other Yellowstone geysers, such as Daisy, Castle, and Lone Pine geysers.

Click for a larger image


Activity in 2010

Activity in 2009

Activity in 2008

Activity in 2007

Activity in 2006

Activity in 2005
