Supplies:

Data were collected on two separate visits to Enchanted Rock where two routes were photographed each visit. The supplies for the experiment included a digital camera (in this case an iPhone), a tape measure, and a PVC quadrat. A partner was also needed to capture photographs safely. The PVC quadrat was divided into 25 5 x 5cm squares measuring a total area of 25 x 25cm. It was enclosed in PVC pipe where the squares were constructed with thick nylon string.

Safety considerations:

It is important to address the safety concerns that accompany collecting data in the field. Packing enough water for the day was necessary as some approaches can take approximately one hour. The same goes for food as it is important to bring snacks along while ensuring to leave no trace and pick up what others may have left behind. The trails can easily be hiked in sandals but is a good idea to have some closed-toe shoes with good tread as it is more conducive for the granite slabs. Species native to the area such as the Western diamondback rattlesnake also makes shoes a sensible choice as well as self-awareness of one’s surroundings. All data was collected from the ground on the four established routes of interest with the help of my partner. This way neither of us was introducing anything into the experiment that could have dangerous implications such as technical climbing.

Route selection:

The four established routes were all of classic status for the area with a difficulty ranging from easy to moderate. This classic status practically ensures the routes are frequently climbed when climbers visit the area. Data were collected for an easy traditionally protected crack (Sweat) and a moderate sport route (Pro Sweat) both located on News Wall of the Main Dome area of Enchanted Rock. The other two routes existed in Echo Canyon; a separate area technically considered to still be a part of the Main Dome complex. In Echo Canyon, a moderate traditionally protected crack (Orange Peel), as well as a moderate sport route (Orient Express), were the subject of interest. Below is a map courtesy of Google Maps of the area with the two areas marked along with the trail entrance that starts at the main parking lot.

Data collection:

The tape measure was used to measure the length from the ground to the highest point of interest. I decided to take photos of each route with the PVC quadrat measuring 7ft and 9ft off the ground as well as 3ft to the right of each route at these same heights. These distances were chosen to standardize data collection and provide a distance high enough off the ground without leaving the ground. For the sport routes, this meant two photos directly below the bolt line and two photos 3ft to the right of the line. As for the case of the cracks, this meant two photos immediately left of the crack (7 and 9ft) as well as the four photos 3ft to the right of the crack. There were a total of four routes analyzed and four photos were taken at each site for a total of 16 photos to be used in the analysis.

After photos were taken it was important to take notes about what part of the route the various photos pertained to to conduct the analysis. The PVC quadrat served nicely to separate the surface area of the rock in the 25 x 25cm grid so that the growth of lichen could be quantified. In the photo above one can see the different colors of lichen covering the granite dome, right of the crack Sweat. I used the tape measure to find the spot that was 3ft to the right to then collect the off-route data with the quadrat. Below shows one of the many photos taken with the PVC quadrat held against the rock, in this case it is being held immediately to the left of Sweat to collect on-route data. I attached one more on route photo of the sport line Orient Express. One can take notice of the differences in lichen cover between these two photos.

Data processing and uncertainty:

For the analysis I thought it was best to focus on my two questions of interest being; how does lichen cover differ between off-route and on-route conditions when climbing is concerned and how does lichen cover differ on-route between traditional and sport climbs. I ended up with four observations for each of the four different routes. Lichens of different colors were measured and sorted into the following groups: black, orange, light grey, dark grey, dark green/brown, and neon green. Light grey and dark grey were then consolidated into the “grey” group and neon green was consolidated with dark green/brown to obtain the “dark green brown” group. Neon green colored crustose lichen was seen in very low proportions compared to the other colors, which is also why I thought it was appropriate to combine the data. The light grey colored lichen was the only one of a foliose variety, but it was consolidated with the grey colored lichen to simply look at the four main colors seen on the surface of the granite.

Counting the amounts of lichen proved to be trickier than I expected. It also took longer than anticipated once I realized the shortcomings with counting particular lichens on the first route Orient Express. This route was visited again at the end for a recount using the new methods I found for obtaining counts for each color. The free open-source photo editing software Gimp was used to edit the photos and count lichens. The first step was separating the area of interest from the PVC quadrat as well as the region beyond the quadrat. By using the ‘scissors selector’ tool I was able to cut out the 25 x 25cm area. This gave me the total amount of pixels in the frame, which I recorded for each photo for the data analysis. Then, I thought it would be simplest to use the ‘select by color’ tool that lets one select a color as it finds all related colors in the photo. This proved to have some troubles. The uncertainty and sources of potential error go as follows. First, light grey colored lichen was identified as in the same color range of the cord separating the quadrat, so it was apparent light grey lichen needed to be counted manually. This was done using the ‘lasso’ tool to encircle the light grey lichen in each 5 x 5cm square to record the values for each patch of light grey lichen. These numbers were then summed to obtain the total pixel count for the light grey lichen for the observation. This was also done for the color orange since the shades of orange lichen varied too drastically to obtain an accurate representation from the ‘select by color’ tool. The ‘select by color’ tool was used for the rest of the lichen colors by selecting five different places on the photo, recording the pixel count, and taking the average of those numbers. This was done for black, dark grey, lime green, and dark green/brown. The black-colored lichen showed the max value among the total proportion of lichen cover in each photo as it was frequently seen covering over 10% of the photo area used for analysis. The issue with the black lichen was that it could also be mistaken for minerals in the granite such as hornblende or biotite. However, upon analysis of the photo, it was apparent which photos contained black lichen. If the photo contained an amount resembling a lawn of black lichen, then five points were selected and averaged to obtain the black lichen pixel count. If it was patchy and less abundant the lasso tool was used on each area individually.

Visualization and statistical analysis:

To better visualize the data, I needed to separate it into two groups: data gathered directly on-route and data gathered off-route. The ‘erockpixels’ data was filtered by ‘location’ to make the following four plots. I ended up making two plots for each group to plot lichen color proportions as well as the mean pixel counts for each color of the lichen. I was able to successfully add error bars on the second plot concerning the mean pixel counts, which is why the second plot was constructed for each group.

erockpixels <- read.csv("/Users/bitoFLO/Desktop/Website/content/I4pixelcounts.csv")
erockpixels <- erockpixels %>% filter(location == "on")
erockpixels <- pivot_longer(erockpixels, c("black", "grey", "orange", 
    "darkgreen.brown", "no.growth"), names_to = "color", values_to = "values")
erockpixels <- select(erockpixels, -c(total.pixels))
erockpixelssum <- erockpixels %>% group_by(color) %>% dplyr::summarize(mean = mean(values), 
    sd = sd(values), count = n(), se = sd/sqrt(count)) %>% mutate(meanfrac = mean/sum(mean))
erockpixelssum <- erockpixelssum[-c(4), ]  #removes 'no.growth' for plot clarity
# On-route: Proportions plot
ggplot(data = erockpixelssum, aes(x = reorder(color, mean), y = meanfrac, 
    fill = color)) + geom_bar(stat = "summary") + scale_y_continuous(limits = c(0, 
    0.1), breaks = seq(0, 0.1, by = 0.025)) + geom_text(aes(label = round(meanfrac, 
    4)), position = position_dodge(width = 0.9), vjust = -0.25, 
    color = "black") + ggtitle("On-Route: Lichen Proportions by Color") + 
    xlab("Color of Lichen") + ylab("Mean Proportion") + scale_fill_manual(values = c(orange = "#fcae1e", 
    darkgreen.brown = "#5c4827", grey = "#c0b9ac", black = "black"))

# Off-route: Mean pixel counts plot
ggplot(data = erockpixelssum, aes(x = reorder(color, mean), y = mean, 
    fill = color)) + geom_bar(stat = "summary") + geom_errorbar(aes(y = mean, 
    ymin = mean - se, ymax = mean + se), width = 0.3, color = "green") + 
    scale_y_continuous(limits = c(0, 6e+05), breaks = seq(0, 
        6e+05, by = 1e+05)) + geom_text(aes(label = round(mean, 
    2)), position = position_dodge(width = 0.9), vjust = -0.25, 
    color = "black") + ggtitle("On-Route: Lichen Pixel Counts with Error Bars by Color") + 
    xlab("Color of Lichen") + ylab("Mean Pixel Count") + scale_fill_manual(values = c(orange = "#fcae1e", 
    darkgreen.brown = "#5c4827", grey = "#c0b9ac", black = "black"))

erockpixels <- read.csv("/Users/bitoFLO/Desktop/Website/content/I4pixelcounts.csv")
erockpixels <- erockpixels %>% filter(location == "off")
erockpixels <- pivot_longer(erockpixels, c("black", "grey", "orange", 
    "darkgreen.brown", "no.growth"), names_to = "color", values_to = "values")
erockpixels <- select(erockpixels, -c(total.pixels))
erockpixelssum <- erockpixels %>% group_by(color) %>% dplyr::summarize(mean = mean(values), 
    sd = sd(values), count = n(), se = sd/sqrt(count)) %>% mutate(meanfrac = mean/sum(mean))
erockpixelssum <- erockpixelssum[-c(4), ]  #removes no.growth for plot clarity
# Off-route: Proportions plot
ggplot(data = erockpixelssum, aes(x = reorder(color, mean), y = meanfrac, 
    fill = color)) + geom_bar(stat = "summary") + scale_y_continuous(limits = c(0, 
    0.1), breaks = seq(0, 0.1, by = 0.025)) + geom_text(aes(label = round(meanfrac, 
    4)), position = position_dodge(width = 0.9), vjust = -0.25, 
    color = "black") + ggtitle("Off-Route: Lichen Proportions by Color") + 
    xlab("Color of Lichen") + ylab("Mean Proportion") + scale_fill_manual(values = c(orange = "#fcae1e", 
    darkgreen.brown = "#5c4827", grey = "#c0b9ac", black = "black"))

# Off-route: Mean pixel counts plot
ggplot(data = erockpixelssum, aes(x = reorder(color, mean), y = mean, 
    fill = color)) + geom_bar(stat = "summary") + geom_errorbar(aes(y = mean, 
    ymin = mean - se, ymax = mean + se), width = 0.3, color = "green") + 
    scale_y_continuous(limits = c(0, 6e+05), breaks = seq(0, 
        6e+05, by = 1e+05)) + geom_text(aes(label = round(mean, 
    2)), position = position_dodge(width = 0.9), vjust = -0.25, 
    color = "black") + ggtitle("Off-Route: Lichen Pixel Counts with Error Bars by Color") + 
    xlab("Color of Lichen") + ylab("Mean Pixel Count") + scale_fill_manual(values = c(orange = "#fcae1e", 
    darkgreen.brown = "#5c4827", grey = "#c0b9ac", black = "black"))

Just from eyeballing the proportions plots and taking notice of the number atop each bar that indicates the proportion of the area covered with a particular color of lichen shows there to be more growth on-route than off-route. This is what I was expecting to see as it seems to intuitively make sense, especially if the routes are frequently traveled compared to others in the park. Both black and grey lichen proportions are greater off-route from the data, and this was exciting to see despite the uncertainty introduced when counting the pixels for these two colors. However, it is important to conduct a proper statistical test before making any inferences. The mean pixel count plots show error bars that are fairly large for each color. Collecting more data across routes in Enchanted Rock would help bring these error bars down.

I used R plyr functions to rename the column headers for the different lichen colors so I could then convert the wide data into a long format. Since I wanted to compare the means between the on-route group and the off-route group and normality was violated after looking at the plots created for each group, I decided on using the Mann-Whitney U test. Since the total pixel count differed for each photo, the proportions needed to be used when conducting statistics. Essentially fractions of lichen cover are being compared in this analysis. Using ‘filter’ I removed the observations for no growth to obtain 64 observations. The null and alternative hypotheses are stated below.

Ho: The distribution of mean proportions for lichen cover between on-route quadrat placements and off-route quadrat placements are the same.

Ha: The distribution of mean proportions for lichen cover between on-route quadrat placements and off-route quadrat placements differ significantly.

erockprops <- plyr::rename(erockprops, c(prop.black = "black", 
    prop.grey = "grey", prop.orange = "orange", prop.darkgreen.brown = "darkgreen brown", 
    prop.no.growth = "no growth"))
erockprops <- pivot_longer(erockprops, c("black", "grey", "orange", 
    "darkgreen brown", "no growth"), names_to = "color", values_to = "props")
erockprops <- erockprops %>% filter(str_detect(color, "growth", 
    negate = TRUE))
wilcox.test(erockprops$props ~ erockprops$location)

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  erockprops$props by erockprops$location
## W = 566, p-value = 0.4697
## alternative hypothesis: true location shift is not equal to 0

The results ended up not coming back significant. The null hypothesis that the mean proportions for lichen cover were equal between the two groups could be not be rejected (W = 566, n = (32,32), p = 0.4697). One cannot conclude from the data on the four routes and statistical test conducted that the lichen cover differs on-route as opposed to off-route.

The next question concerned how lichen cover differs on-route between traditional and sport climbs. In this case ‘filter’ was used again to remove the off-route data to obtain 32 observations of on-route data. Then, I made another plot in a similar format as the previous two. It made more sense in this instance to plot the total proportions for lichen cover per color for each route in this case. This additional chart once again showed orange-colored lichens are not found in high proportions and that most of the lichen present are black and grey. Orient Express (sport) shows to be the one route that shows considerably lower proportions of lichen when eyeballing although at the same time has the greatest proportion of orange-colored lichens. It is important to take into consideration factors such as sun and surface runoff. Orient Express is very exposed and stays in the sun most of the day whereas routes like Sweat (trad) and Pro Sweat (sport) do not receive nearly as much sun while also receiving much more runoff from the Main Dome. Orange Peel (trad) is about as tucked away as it can be only receiving morning sun briefly. It also seems that higher proportions of black colored lichens appear on these three routes when compared to Orient Express.

erockprops2 <- erockprops %>% filter(location == "on")
erockprops2 <- erockprops2 %>% mutate(route_clean = str_sub(route, 
    1, -4))
erockprops2sum <- erockprops2 %>% group_by(route_clean, color) %>% 
    summarize(meanprop = mean(props))
ggplot(erockprops2sum, aes(x = route_clean, y = meanprop, fill = color)) + 
    geom_bar(position = "stack", stat = "identity") + xlab("Route") + 
    ylab("Mean Proportion per color") + ggtitle("Lichen Distribution Between Routes (On-Route)") + 
    scale_fill_manual(values = c(orange = "#fcae1e", `darkgreen brown` = "#5c4827", 
        grey = "#c0b9ac", black = "black", `no growth` = "#a4877e"))

wilcox.test(erockprops2$props ~ erockprops2$style)

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  erockprops2$props by erockprops2$style
## W = 118.5, p-value = 0.7325
## alternative hypothesis: true location shift is not equal to 0

When looking at this plot Pro Sweat is seen with the most lichen cover across the two spots where the quadrat was placed. It is interesting to see how this route is a sport line and shows more lichen cover than the two traditional lines. When looking at Pro Sweat compared to Orient Express (the other sport line) they are complete opposites; a min and a max for total lichen cover among the four routes. I conducted the Mann-Whitney U test once again.

Ho: The distribution of mean proportions for lichen cover between traditional and sport routes are equal.

Ha: The distribution of mean proportions for lichen cover between traditional and sport routes differ significantly.

This time the p-value comes back even higher as the results are once again not statistically significant. The null hypothesis that the mean proportions for lichen cover were equal between the two groups could be not be rejected (W = 118.5, n = (16,16), p = 0.7325).