Lightning Data - General Discussion

    The lightning data used on this page comes from the National Lightning Detection Network (NLDN), operated by Global Atmospherics, Inc [ ].  The NLDN detects cloud to ground lightning strikes across the US with a network of over 100 sensors. More technical information on how this is done can be found in the following reference [1]:

[K.L.Cummins, E.A.Bardo, W.L.Hiscox, R.B.Pyle, A.E.Pifer NLDN '95: A Combined TOA/MDF Technology Upgrade of the U.S. National Lightning Detection Network from the Proceedings of the 12th Conference of Interactive Information and Processing Systems (IIPS) For Meteorology, Oceanography, and Hydrology]

     The NLDN has been in operation since 1989.  Initially, strike locations were estimated to be only accurate to a rather coarse 10 km.  Upgrades occured in the early 90's and by 1995 there was a significant increase in both detection efficiency and location accuracy. The detection efficiency (the percentage of the actual strikes that are detected by the network) is reported to be 80%-90%.  The reported locations of strikes are accurate on average to about 0.5 km.  This resolution raised, for the first time, the possibility of correlating lightning with terrain features on a fairly fine scale; for instance, the scale of a mountain ridgeline, a canyon edge, an isolated desert mountain peak or even an isolated tall tower. The decision was made to not use the less accurate NLDN data from before 1995 in the analysis here.  Currently the data used in this work runs from Jan 1, 1995 to Dec 31, 1999.  The times of day reported in the original NLDN data are UCT, but for all of the work done here the times were converted to (MST) time.
     A small sample fragment of NLDN data is shown below:

09/30/95 23:56:57 34.834 -113.248 0.0 -20.4 kA 2
09/30/95 23:56:58 32.063 -111.071 0.0 -10.2 kA 1
09/30/95 23:57:58 34.329 -110.970 0.0 38.0 kA 1
09/30/95 23:59:59 34.879 -113.354 0.0 -35.9 kA 12

The first four data fields present are Date, Time, Latitude, and Longitude.  The last two give the Peak Current (including polarity) and the Multiplicity of strokes per flash.  To correlate lightning with terrain, an additional data field that would be of interest would be the elevation at which the strike occured.  Computer code was written to use a strike's latitude and longitude to lookup and assign an elevation to each strike.  The elevations came from the U.S. Geological Survey datasets called Digital Elevation Models (DEM's) [see:  Elevation Data - General Discussion ].  The NLDN data reports lightning strike latitude and longitude to the nearest 0.001 degree. This corresponds to 3.6 arcsec, which is close to the 3 arcsec resolution of the elevation data found in the USGS DEM's. A resolution of 3.6 or 3.0 arcsec would pinpoint a location to within about 100 m, but the error bars on the lightning location are actually larger. The average accuracy of the strike location is more like 500 m.
    The image below illustrates the situation. The fine grid represents 3 arcsec latitude and longitude bins. A strike is reported at the red 'x', but there is a 50 % chance that the actual location is somewhere within the small red ellipse and a 90% chance it is within the larger ellipse (see Ref. [1] above).

    There are over a dozen of the small bins that all have nearly an equal chance of containing the actual strike. The darker grid lines represent a 12 arcsec resolution. The decision was made to use this coarser grid to assign elevations. The 12 arcsec resolution better matches the accuracy of the lightning data. There is nothing to be gained by an elevation assignment at a resolution which is significantly higher than the lightning resolution.  12 arcsec elevation data was obtained by a straight forward averaging of 4 x 4 arrays of 3 arcsec values. The 12 arcsec data still contains a wealth of topographic information as can be seen in the terrain images elsewhere on these pages.

Region and Period of Study:
     This study covers the 72 quadrangles from latitude 31 N to 43 N and longitude 114 W to 108 W.  This includes virtually all of Arizona (except a slender strip bordering Calif.), all of Utah, New Mexico and Colorado and pieces of Idaho, Wyoming and Nebraska.  The 72 quad region is quartered into sections of 6 quads north to south by 3 quads east to west.  This division correpsonds approximately to the four states of Arizona, Utah, Colorado and New Mexico.  Those  four regions will be referred to by those titles, but it should be kept in mind that the 'Arizona' region contains a slice of far western New Mexico, the 'Colorado' region contains parts of Wyoming and Nebraska and so on.
    The study is limited to the summer monsoon months of July and August.  To see how this season fits into the larger picture, the following graphic shows a plot of the monthly total lightning for each of the four state regions for the whole year.  The number of strikes is a total for that month for the five year period 1995-1999.  1 = Jan., 2 = Feb. and so on.

    The bulk of the lightning occurs in the summer monsoon period of July and August, when 67% of the annual lightning occurs.  Again, this is the period of interest in the study.   Of the rest of the year, by far the next biggest contribution is from September or June.  However, in September storm systems from the Pacific often begin to play a role and in June the high plains are more active so the signal begins to contain more than just the summer monsoon pattern.
    In July and August, Arizona and New Mexico dominate the lightning count.  The numbers are not adjusted to reflect density, which provides a small bias, but this only amounts to about 5% of the enhancement of the southern tier of states.  (All four regions cover the same size rectangle in terms of latitude and longitude, but the area on the ground is less for the more northern regions due to the curvature of the earth.)   New Mexico and Colorado both cover a large area of high plains, so they dominate in May and June.  It is interesting to note that Arizona and New Mexico are both strong in July and August, but are roughly mirror image in terms of a slight August peak for Arizona and a slight July peak for New Mexico.  Lightning in Colorado doesn't strongly favor July or August and Utah has fairly constant counts for July, August and September.

Display Options:
    With the lightning data chosen, the question arises as to how to display the data.  The lightning strike locations are given to  the nearest 0.001 degree latitude and longitude.  This is close to 100 m distance, however the typical error in these locations are more like 500 m.  The following image shows the southwest corner of the Santa Catalina mountains NE of Tucson:

Each red dot is a separate lightning strike.  All strikes for July and August of 1995 -1999 are present.  This image is just large enough to be sensitive to the 0.001 of a degree.  Only strikes that happen to occur at the same latitude and longitude to the nearest 0.001 of a degree are plotted on top of each other, but the strikes appear to be sufficiently spread out so that there are probably not a huge number of  dots plotted on top of other dots.
   Variations in strike density can clearly be seen, but the patterns are probably not boldly obvious to the viewer.  A quantitative comparison of densities in different regions is in order.  Also there is the issue of the location accuracy not really being this fine anyway.  As a result, the decision was made to count up strikes in a grid with a lower resolution.  Since the location accuracy is roughly 500 m, something closer to this range would be reasonable.
   The following three images show the results of some possible choices.  Each image show a 0.5 deg latitude by 0.5 deg longitude region that includes all of the Santa Catalina mountains NE of Tucson.
   These three images represent bin sizes of 0.005 degree, 0.01 degree and 0.02 degree.  The middle resolution (0.01 degree) is the one used in the lightning maps posted at this site.  The other two plots show the results of a resolution of half as much or twice as much.  The first image shows lightning counted in bins of 0.005 degree latitude by 0.005 degree longitude.  Naturally, the smaller the grid rectangles, the smaller the count, and the 0.005 degree grid counts will average one fourth as much as the 0.01 degree grid.  For the 0.005 grid, there are a lot more bins where only one or two (or no) strikes occured in the entire five year period.  Even in fairly active areas, there will be more 'holes' due to random fluctuations of the smaller numbers.  In the 0.01 degree resolution, regions of higher strike density seem to stand out more solidly.  The 0.01 degree grid rectangles give slightly finer than 1 km resolution [about 930 m north to south and about 800 m wide east to west].  This is just slightly beyond the typical error in the strike locations, so one wouldn't expect to be able to go much finer and still see any new signal in the data.  The 0.02 degree plots looks smoother yet, but the approximate 2 km x 2 km bin size is getting to be enough bigger than the lightning location accuracy, that details of the signals are being lost.  A 4 sq. km. bin size makes for a pretty coarse grid to lay over a small desert mountain range.
   One subtle issue involves the use of fixed latitude and longitude grids to count lightning.  This is an easy method, but it means that the area of the grid rectangle will not stay exactly fixed as one moves from north to south.  As a result the numbers, and subsequently the colors, on these images are not strickly proportional to strike density.  There would be a tendency for the colors to get more intense at the southern end of a map.  Fortunately this tendency is very small over a single 1 degree by 2 degree quad.  For example, for a quad such as Tucson, the area of the grid rectangles increases by cos(32)/cos(33) which amounts to 1.1%.  So a 20 strike bin at the top of the Tucson quad would need to be the same color that the value 20.22 has at the bottom of the quad, in order for the colors to be truely proportional to strike density.  20 and 20.22 on the color scale are indistingishable, so a correction for this effect was not made.  It is a little more of an issue on maps that cover a much larger area, such as the whole state of Arizona, but even then it amounts to a few percent.

Total Lightning Strikes - Discussion of some Results

   This section discusses one particular type of map.  These maps show total lightning strikes added up for July and August during the five year period of 1995-1999.  The strikes are totalled in a grid of rectangles with a resolution of 0.01 degree latitude by 0.01 degree longitude.  These grid rectangles are a simple way to divide up a quadrangle into bins, but they do have the disadvantage of changing area as one goes from north to south.  The rectangles range from as big as 1.11 km high and 0.95 km wide (for an area of 1.05 km2) along the southern edge and as small as 1.11 km high and 0.81 km wide (for an area of 0.90 km2) at the northern edge.  So the area of the grid rectangles decreases about 15% from the bottom to the top of entire region.  To display an unbiased lightning density, would require correcting for the area change.

Full Region Plots.

    Figure 1. below shows lightning over the entire region, using data that is not corrected to show the proper density.  It is a Mercator projection, so a strip across the north end is just as wide as a strip along the south end, while in reality there is less actual real estate for lightning to hit.  This lightning is spread out into the same number of bins so there will be a tendency for the lightning plot to appear weaker as one moves from north to south.  This turns out to be a very small effect, however, compared to the fact that there really is much less lightning activity over the north.

   The next graphic (Figure 2.) shows the exact same data, but an actual density calculation has been made.  The same total number of strikes that was used to generate Figure 1. , were then divided by the area in square kilometers of that rectangular bin.  The numbers don't change much though, since the area per bin is close to 1.0 km2 for the whole plot.  In the center, especially, the area is very close to 1.0 km2.  The area is slightly more than one along the bottom edge (1.05 km2) so these values are reduced about 5% and the area is slightly less than one along the top edge (0.90 km2) so these values are increased by about 10%.  However, a 10% change will turn a 20 into a 22 or a 6 into a 6.6 and a look at the color bar shows that the difference will be hard to notice.  Figure 1. and Figure 2. are hard to tell apart.  For this reason no correction for latitude was incorpated in the results that follow.  In a plot for a single quad the correction amounts to only about 1% from the top to the bottom which is even more trivial.  All other Total Lightning Strike maps here will use totals in 0.01 degree by 0.01 degree resolution bins.  The numbers are close to a strikes per km2 value and that value can always be calculated more accurately, given the latitude, if desired.
     The above figures show the total lightning strikes for the full region of study.  Maps are also available at three levels of finer detail (state regions, half state regions and quadrangle).  The detailed maps show many features that are lost on the large plot and several examples will be seen below.  The more detailed maps are especially useful for the weaker areas, such as Utah, because the color scale can be darkened there to enhance the weaker features.  The plot of the full region needs a color scale that spans the full range of activity.  This is also a strength of the full plot though, because it shows how the different quads compare with each other.  The separate quad plots have color scales that use 0-20, 0-30, 0-40 or 0-50.  This should be carefully noted when viewing multiple quads.
   A number of interesting features can be seen in the plot of the full region.  There is a definite trend toward less and less lightning as one goes from south to north.  Even using the density values to correct for shrinkage of the bin size.  The highest value in the whole plot is a 58, which is found along the lower left (in Sonora, Mexico just south of the Nogales, AZ area).  Along the northern edge values of about 15-20 are the highest.  There are several mountain peaks that hit values of about 50 in southern Arizona and New Mexico.  These also correspond to about 50 strikes per km2.  This means about 10 strikes per km2 on an annual basis.  It should be noted that this is very high and it only includes the months of July and August.  On nationwide lightning maps, one typically only sees values as high as 10 strikes per year per km2 in parts of the southeast.  The southeast also tends to have lightning spread out more through the course of the year, so values of 5 strikes per month per km2 in parts of this study region (such as the Santa Catalina Mtns north of Tucson, AZ or the Sacremento Mtns of New Mexico) may rank as the highest in the nation.
   Figure 3. shows a plot of total lightning versus latitude.  The whole region consists of 1200 by 1200 rectangular bins and for this plot all the bins in the same latitude row were just totalled together.  The bottom strip, for instance, has a total area of about (1200)x(1.05 km2) = 1260 km2.  The total number of strikes in that strip was about 12,500.  So this strip averaged about 10 strikes per km2 for the five year period or about 2 strikes per year per km2.  The activity drops to about half this value at the middle of the region (the AZ/UT and NM/CO borders) and to about a third of that value at the north edge of the region.  These values here are corrected for the decrease in bin size with latitude.  The numbers are basically a strike density with the bottom width used as a reference so they are all strikes per 1260 km2.
   The plot of the full region also does a nice job of revealing the lightning hot spots.  There are perhaps five or so important areas of thunderstorm development on this largest scale.  Figure 4. highlights and labels these regions.  The general movement of storms within these areas can be seen on the diurnal animations (seen and discussed elsewhere on this site) and the arrows here are included to reflect the overall movement trends.  The five zones are discussed breifly below:
1.)  The most intense activity seems to originate in the mountainous areas along and south of the Ariz./Sonora border from Nogales to Douglas.  This activity tends to build to the NW.  This area extends farther into Mexico and the even larger mountains of the Sierra Madre, that are too far south to be seen on these maps, probably play a role as a genesis region.

2.)  Another large genesis area extends along the Mogollon Rim of Arizona, through the Arizona White Mountains and ends at the Mogollon mountains in New Mexico.  Storms tend to fire first along the high elevation axis that runs WNW to ESE and then they migrate downhill to the SW.

3.)  There is a strong genesis region in the front range mountians around Pikes Peak which blossoms out onto the Palmer's Divide which is a wedge of higher elevation that sticks out into the high plains.
4.)  The front range mountains of northern New Mexico are also a strong genesis region.  Again the presence of a peninsula of higher terrain heading east into the plains (the Raton Divide in this case) seems to yield an enhanced corridor for storm development.

5.)  The Sacremento Mountains of New Mexico.  The diurnal animations reveal this area as a very early genesis region.  Unlike the above mountain areas, however, storms do not seem to develop off of these mountians in one particular direction with the strong signal seen in those cases.  In particular the extension out into the plains to the east is not pronouced as in the other two front range areas of 3.) and 4.).

Detailed Plots from Selected Quads.
   This section deals with the data as viewed on a much more detailed scale.  All of the maps cover only a single quad or a portion of a quad.  A number of interesting features can be seen.

Example 1.)   The first examples show a strong correlation with elevation and lightning intensity.  Figure 5. is a close-up of the Santa Catalina Mtns just north of Tucson, AZ.  These mountains are a nice example of a large, but isolated island of high mountians (just over 9000 ft) surrounded on all sides by lower desert (2000ft - 3000ft).  The map covers an area roughly 30 miles on each side.  The lower elevations have typically 5-15 strikes per bin for the five year period, whereas the number is 3 to 5 times that at the highest zones.  The lower left part of the map would be in the city of Tucson.  Note how the lightning strikes follow the two parallel ridges that run north from the center of the mountains, while the canyon between them is much suppressed.  Sabino Canyon, which cuts into the south side of the mountains, can also be picked out from the lightning signal as well.  The lightning maxima associated with the Santa Catalinas appears to be centered quite well (on the average) on the peaks with no clear trend for storms to effect the NW side or NE side and so on.

Example 2.)    The next example (Figure 6.) shows the Sierra Estrella mountains southwest of central Phoenix, AZ.  Downtown Phoenix is in the upper right hand portion of the map.  The area covered is similar to the previous map.  Recall that each rectangular bin is close to a kilometer square and there are 10 bins to the 0.1 degree of latitude or longitude.  The Sierra Estrellas are the mountain chain that lie along a NNW to SSE line in the lower middle of the map.  This is another isolated mountain island surrounded by flat desert, but different in many ways from the Santa Catalinas.  The Sierra Estrellas are not nearly so high (peaking at about 4500 ft in a 1000 ft desert plain).  They are a long narrow range, only 2 or 3 miles wide in places, while the Santa Catalinas are much more massive, being 20-25 miles wide in any direction.  Even so the Sierra Estrellas appear to have a distinctly enhanced lightning signal.  The signal is not always very sharp each individual year (which it is in the case of the Santa Catalinas) but has definitely emerged after the five year period.  Even the individual hot spots along the range correspond well with the actual peaks in the mountains.  There is a lot of random scatter too, of course.  Note the two hot bins (in black) in the low desert to the west of the mountains.
    There is a smaller mountain range that runs more east to west on the middle of the right edge.  This is the South Mountain Park area and is in the Phoenix city limits.  These low mountains only reach about 2500 ft.  These appear to be too small to produce a clear lightning enhancement although there are some clusters of 'hotter' bins around.  Perhaps a weak signal would emerge in a longer time period.  The highest part of these mountains is also home to a forrest of TV and radio towers.

Example 3.)     Clearly, a great many of the maximums and miniumums in the lightning strike totals correspond to mountains and valleys, but there are exceptions.  Inspection of several maps will turn up a variety of blotches, streaks or even arcs of lighting strikes that at first glance might seem to suggest an underlying structure, but on closer examination look to be in the middle of nowhere, topographically speaking.  Random noise is to be expected, of course, but the multi-year summation is supposed to help smooth that out.  However, five years (ten different months) is not enough to completely smear out the contributions from individual storms.  Particularly in areas with less total lightning, that still get an occasional intense storm (e.g. the low deserts of SW Ariz., the Colorado Plateau area of the four corners and the high plains of New Mexico), one sees blotches and streaks that are the signature of just one or two events.  The higher lightning activity areas do not tend to have as much fluctuation that look like individual storms.
   One of the best examples of these single storm signals can be seen in the map below from the Ajo quadrangle (Figure 7.).  There is a pronounced peak of lightning in the NW part of the map at about 32.8 N and 113.5 W.  This area is in a broad flat valley of the low desert, near the Gila River and yet the intensity is comparible to some of the active mountain areas in the SE part of the map.  It turns out though, that this peak is primarily due to a single tremendous storm on Aug 27, 1997.  The two plots in the image differ only in that one does and the other does not include Aug. 27, 1997 in the five year lightning strike totals.  The large bullseye signal dissappears when that one day is left out.  It is hard to see much difference anywhere else on the two maps, besides that storm.  The Aug. 27, 1997 storm was an unusual event.  For a solid hour (5:30 MST-6:30 MST) a very intense storm sat over the same area with 5 to 10 flashes per minute for that entire time.  Less frequent lightning occured a half hour to an hour on either side of this period.  Nothing else approaching that intensity was seen elsewhere in the quad that day.

    In an effort to uncover these one day anomalies another set of plots was generated.  Subtracting out hand selected individual days is not practical for all the quads, however the same information can be obtained by counting just the number of different days with lightning instead of the total number of strikes.  That way a single storm that produces multiple strikes in a bin will still only count for one event.  These plots are labeled 'Total Storm Days' instead of 'Total Lightning Strikes'.  The Total Storm Days plot for the Ajo area is shown below in Figure 8.  The suppression of the strong single storm maxima is just as effective here as it was in the special plot above.
 The Total Storm Days plot overall tends to be a little more smoothed out, which one would expect.  Still there are adjacent bins where the number of days with lightning jump from say 2 to 12.  This random scatter is partly due to the fairly small size of the bins.  It is possible for a storm to roll through your location with heavy rain and lightning, but if there didn't happen to be a detected cloud to ground strike in that 1 km2 bin, it would not be counted as a storm day.

Example 4.)   Figures 9. and 10. show the Lukeville quad.  The first plot is a topographic map and the second shows lightning strikes.  This region is interesting because of the steep lightning gradient between the Gulf of California and the mountains to the east.  The Nogales quad (to the east of Lukeville) has the highest strike densities of anywhere in the study area, while lightning is very rare over the northern gulf.  It can be seen along the southern edge of the Lukeville lightning plot that within a span of about forty miles (say from 112.7 W to 112.1 W) there is five times as much lightning without a major change in the terrain.  As one travels from west to east along the south edge of the quad there is of course a general rise in elevation and an increase in mountains but nothing dramatic and it would seem clear that the lightning pattern reflects something beyond just local terrain.  There are mountains on both sides of the lightning boundary and there is a large isolated volanic peak on the quiet western side of the map.  Even though this peak rivals the highest elevations to the east, it shows only the most minimal of a lightning signatures.  This lightning gradient becomes more diffuse to the north in Arizona.  It would be interesting to see how this feature continues farther into Mexico.

Example 5.)  The clearest signals in the data are mountain and valley signals, but there are other possible features to be found.  Figure 11. shows a more expanded view of the Phoenix quad.  This includes the Sierra Estrellas that were discussed above.  Central Phoenix is in the NE quarter of the plot.  There are two mountain ranges to the southwest and west of Phoenix; the Sierra Estrellas with a fairly strong maxima and the White Tank mountains with a minimal enhancement.  The Salt River flows between these two ranges in a broad flat valley.  There appears to be an area of enhanced lightning (highlighted below) in this valley.  Storm activity in the Phoenix valley often builds down from the north, northeast or east.  As discussed earlier, the Mogollon Rim is an important genesis area and these storms often build southwest towards the desert.  Phoenix is often at the tail end of this migration.  Storms in the valley are often preceeded by or associated with cooling outflows from the northeast quadrant.  The space between the White Tanks and Sierra Estrellas would represent a constriction for any shallow outflows that are generally out of the northeast.  Perhaps the highlighted valley signal is due to this convergence.  The plot of Total Storm Days is not shown here, but it looks very similar, so the enhancement is not due to one or two anomalous storms.
Example 6.)   Figures 12. and 13. show a topogrpahic map and a lightning map for the main portion of Grand Canyon National Park, Arizona.  The south rim of the canyon represents a sharp drop-off for lightning frequency.  The north slopes inside of the canyon presents a larger complex of side canyons and spires and the lightning reflects this.  Lightning maxima can seen with some of the pinacles inside the canyon itself, on the north side of the river.  A tendency can be seen for more lightning in the low elevations off to the west rather than the low deserts off to the east.  This trend shows up nicely in the larger full quad maps and the diurnal animations.  The high country above the north rim turns out to be a very early genesis area for storms, but activity here has already peaked by 1:00 PM to 2:00 PM MST, and storms tend to develop further west and south later.  Overall, the north rim has a suprisingly weak total signal, with more lightning found in the lower, but very rugged, mountainous areas to the west such as the Mt. Trumbull area.
Example 7.)  The next three examples illustrate a general trend that occurs from south to north in the full region.  In the southern most peaks, such as the Sacremento Mtns in New Mexico and the Santa Catalinas Mtns in Arizona, the lightning density is highest at or near the very highest peaks.  Farther north however, there appears to be more lightning on the side slopes at some intermediate elevation, with a lower density at the highest elevation.  Figure 14. shows the Pikes Peak, Colorado area.  Pikes Peak is just left of the center of the map.  The highest lightning activity occurs in a rough ring around Pikes Peak.  The ring forms a backwards 'C' as it wraps around the north, east and south sides of the highest peaks.  This lightning maxima occurs at an elevation that is distinctly lower than the peak in the center.
Example 8.)   Figure 15. and 16. show the Unitas Mtns in Utah.  These mountains consist of a broad, wide area of high elevation cut into with numerous long canyons, especially on the south side.  The lightning plot shows the most frequent strikes not at the highest altitudes, but in a rough ring around the sides.  The long peninsular ridges to the south seem to have the strongest signal.  Within this middle altitude band there are lightning minima tracing out some of the bigger canyons on the south side.
 Example 9.)   Figure 17. and 18. are from the far northern part of the study region.  They include the southern end of the Wind River Mountain Range in Wyoming.  This is a large prominent range but it doesn't produce much of a signal in the lightning pattern.  There is perhaps a suggestion of lightning forming a band around the mountains at the middle elevations as was seen in the prior examples.  The western slopes in particular seem to be enhanced.  It may take more years of data for a signal to develop at this lower lightning activity latitude.
Example 10.)   These last graphics show the Mesa quad., but instead of just the five year total, plots are also shown for each year separately.  This gives a nice picture of the year to year variation.  One can also see how the signal builds from year to year.  Each year there are signs of strong individual storms in a variety of places, but the places that end up having a strong signal will have some contibution from each year.  The Pinal Mtns are located just below and to the right of center (about 110.7 W & 32.3 N).  There is a very strong peak in 1996, but there is some degree of enhancement all of the other years as well, thus adding up to a definite signal.