Wildfires Threaten Saguaro Cacti in Pima County, AZ
Global warming is increasing the frequency of wildfires, posing a threat to the Saguaro Cactus, a species endemic to the Sonoran desert
Introduction
In the summer of 2020, the Bighorn Fire burned nearly 120,000 acres of desert near my home in Tucson, Arizona. Not only do wildfires threaten the lives of residents, they are also a growing threat to our native flora and fauna.
The Saguaro Cactus is the largest species of cactus and endemic to the United States South West and Baja California. It can grow up to 78 feet tall and provides food and shelter for other desert dwelling species.
Global Warming is increasing the intensity and frequency of wildfires all across the world. Mitigating the climate crisis requires a two part solution: countries must make drastic changes to cut their carbon emissions, while local governments prepare solutions unique to their location.
The purpose of my project is to locate areas for concentrated fire prevention based on their threat to Saguaro cacti.
The Sonoran Desert is the only place in the world where the Saguaro Cactus is found. Warm, hot air from the Gulf of Mexico condenses into yearly monsoon storms, providing the moisture and temperature necessary for the Saguaro to survive.
Pima County is my study area for this project. It spans 9,189 square miles and is home to thousands of Saguaros.
Based on aerial imagery of known Saguaro locations, around 54% of land in Pima County is viable for Saguaro cacti to live.
I performed this calculation by dissolving the cactus point locations into a single shape and calculated that its area is 1.38E11 square meters, while the total area of Pima County is 2.56E11 square meters.
Left: Points of known Saguaro locations Right: Estimated area viable for Saguaro cacti.
This map displays the extents of wildfires in Pima County from 1972 to 2021. You can view each fire's name, year, and acres it burned by clicking on the map.
Wildfires in Pima County
Click the play button on the map to view how fire extents have changed over the years. The time lapse intervals span five years from 1972 to 2021.
The graph to the right displays the number of acres burned in per year. Since 1972 there has been an increase of approximately 853.68 acres per year (below).
Linear Regression model
Factors of Wildfire Risk
To identify where Saguaros are at risk from fire, I decided to create a probability model of areas where wildfires are likely to occur based on environmental factors.
Land Cover
The following map displays a 30 meter land cover NLCD classification .
Wildfires commonly occur where there is dry, flammable vegetation. In Pima County, many fires burn in the Ponderosa Pine forests of the Santa Catalina Mountains.
Recently, the spread of invasive buffelgrass has become an urgent problem for wildfire prevention. Buffelgrass, a noxious weed native to the Middle East, poses a direct threat to Saguaro cacti due to its fast rate of growth and flammability.
Buffelgrass causes the Mercer Fire to spread in 2019
Topographic Relief
Relief is a measure of the total amount of variation in elevation of a region, also known as steepness. The rate at which fires spread uphill is faster on steeper slopes because as the heat rises, it "preheats" the vegetation above the fire, making it more susceptible to burning.
I constructed a raster image of total relief from a 10 meter Digital Elevation Model (USGS). Total relief was calculated as the range of elevation values from the Digital Elevation Model in a four cell radius. White represents higher values in each image.
Left: Digital Elevation Model of Pima County Right: Total Relief
Precipitation
The amount of precipitation affects moisture in the air and in the surrounding vegetation. Areas that experience low precipitation often contain dry materials that can fuel wildfires.
The raster image on the right displays the average yearly precipitation (mm of rain) from 01/01/1979 to 03/08/2024. The data was accessed through the Google Earth Climate Engine application.
Legend (mm)
Temperature
Hotter temperatures increase the risk of wildfire by evaporating moisture from the soil and vegetation.
The map to the right displays the average maximum temperature per day (degrees Celsius) from 01/01/1979 to 03/08/2024. I accessed this data from the Google Earth Climate Engine application.
Legend (degrees Celsius)
Probability Modeling
To model of areas of fire risk, I ran a multivariable logistic regression to produce a binary model. A multivariable regression determines the amount of influence various factors have on a certain outcome.
Sampling Data Points and Fire Extents
Sampling
The first step of my process was to compare the locations of known wildfires to a control group. I generated random points throughout Pima county and extracted the values of land cover, relief, precipitation and temperature at each point into a table. Then, I performed the same operation to the wildfire events, averaging their variable values for each fire.
Regression Data Table
On the right is a portion of the regression data table. The value 1 in the column "Case Field" identifies the object as a event group (wildfire) location while 0 represents a point from the control group. The regression will compare the fire locations to the control group to isolate each variable's impact on the occurence of a fire.
Running the Regression
First regression output (click to expand)
I ran the logistic regression in R, a programming language used for data manipulation.
The rightmost column of the output displays the significance, or p value, of each variable in predicting the locations of wildfires. The statistical standard for p-value is < 0.05. Based on this rule, total relief is not statistically significant because its p-value = 0.556.
Second regression output (click to expand)
After removing total relief, I ran the regression again and the p-values were all < 0.05. The first column of the regression labelled "estimate" can now be used to construct the probability model.
Creating the Probability Model
Before creating a binary logistic model, I first had to make a multivariable linear equation using the output from R.
I used the following expression to create the probability model on the right:
4.3164 + ( "precipitation" * 0.09379) + ( "land cover" * 0.08811) + ( "temperature" * -0.49005)
Precipitation, land cover, and temperature are raster datasets (see Factors of Wildfire Risk ) scaled by their respective coefficient values from the regression output.
The constant 4.3164 is the corrected intercept from the output.
The equation a' = a + ln(n 2 /n 1 ) was used to create the corrected intercept because the control group size was different than the study group size.
a' = corrected intercept a = intercept from R output n 2 = larger sample size n 1 = smaller sample size
After creating the first probability model, I performed a logistic transformation on the raster image using the following formula:
1/(1+Exp(-probability model))
This scales my model between 0 and 1, where 1 represent areas likely for a wildfire to occur and 0 represents unlikely areas.
Logistic Transformation Visual
Finally, I created the binary model by assigning all values below 0.5 a value of 0 and all values above 0.5 a value of 1.
1 represent areas of potential wildfire risk and 0 represents areas at low risk.
Results and Conclusions
Limitations
A limitation in my project was the resolution of the temperature and precipitation data I used. This caused my resulting probability model to have a coarse resolution, which reduces its accuracy in predicting wildfire events.
Identifying Saguaro Risk
I compared Saguaro point locations to the areas of wildfire risk predicted by the binary model. The result displays only the Saguaros at risk from wildfires.
Saguaros in Pima County vs Saguaros at Risk
Then, I created a heat map based on the greatest concentration of at-risk Saguaros. The heat map is divided into 20 categories, where higher values represent a greater number of Saguaros. I selected the locations with values of 18 or greater, which yielded the four locations indicated on the map below.
Saguaro Risk Heat Map
Based on my project, the four locations displayed above should be the subject of concentrated fire prevention efforts in order to protect Saguaro cacti. This would require cooperation with tribal authorities and the National Parks Service.
The number of wildfires are increasing across Pima County, and targeted solutions are the best way to mitigate the worst effects of the climate crisis.
