Visualizing the impact of hurricanes on mangroves over time
Author
Isabel Liston
Published
December 13, 2024
Introduction:
Mangrove trees are an important aspect of many coastal ecosystems. Often, these areas are also impacted by hurricanes, therefore it is beneficial to understand how hurricanes damage mangrove trees and how long it takes trees to recover from said damage. Hurricanes impact mangroves through “…three primary mechanisms: wind damage, storm surges, and sediment deposition” (Smith III et al., 2009, p. 24). The wind can damage the trunk, branches and leaves, the storm surges can uproot or knock over the mangrove tree, finally, “…deposited materials interfere with root and soil gas exchange leading to eventual death of the trees (Smith III et al., 2009, p. 24). With knowledge of how hurricanes impact mangroves, we can look towards quantifying the response of mangrove forests to these impacts. To do this, mangroves must be studied over time and at different areas within a hurricane’s impact area. This leads us to the data set used in this project, following Hurricane Andrew in 1992 permanent study plots were established in the Everglades region of southern Florida. More were added later to compare modern observations to historic measurements, in total there are twenty-three plots with data spanning 1992-2011. The goal of this project is to visualize various aspects of how individual mangroves, different mangrove species and the plots overall respond to hurricanes over time.
Materials and Methods:
Data Sources
United States Geological Survey’s Wetland and Aquatic Research Center
To begin this project, I loaded all the necessary libraries, listed below, and imported the data sets. I then set up the study area using a bounding box, cropped the us_states data using this bounding box to get a polygon of the study area and filtered the us_states data to just the state of Florida for a scope of reference. For the Everglades Mangrove Data, I removed the “NA” values for DBH, grouped by Plot ID, reformatted the date to be compatible with R, arranged in date order then transformed the dataframe into a spatial object for mapping purposes. I also used the distinct() function to make a list of the 23 plots with their unique IDs, their sitename and their geometry. Finally, I mapped the tracks of the hurricanes that passed through the study area between 1992 and 2011 using the hurricane exposure package. I extracted the point information of the hurricane tracks and converted it to a spatial object. I then cropped this down to Florida and further cropped it to just the study area.
# Downloading packages -------------------------------------------------------
- Downloading kableExtra from CRAN ... OK [2 Mb in 0.57s]
- Downloading svglite from CRAN ... OK [197.9 Kb in 0.34s]
Successfully downloaded 2 packages in 2.4 seconds.
The following package(s) will be installed:
- kableExtra [1.4.0]
- svglite [2.1.3]
These packages will be installed into "~/work/final-project-IListon/final-project-IListon/renv/library/linux-ubuntu-jammy/R-4.4/x86_64-pc-linux-gnu".
# Installing packages --------------------------------------------------------
- Installing svglite ... OK [installed binary and cached in 0.34s]
- Installing kableExtra ... OK [installed binary and cached in 0.53s]
Successfully installed 2 packages in 0.96 seconds.
#create the study area bounding boxbbox_coords <-c(-81.2133, 25.5831,-80.9134, 25.1589)#create a polygon from the bounding boxbbox_polygon <-st_as_sf(st_sfc(st_polygon(list(matrix(c( bbox_coords[1], bbox_coords[2], bbox_coords[3], bbox_coords[2], bbox_coords[3], bbox_coords[4], bbox_coords[1], bbox_coords[4], bbox_coords[1], bbox_coords[2] ), ncol =2, byrow =TRUE))), crs =4269)) #used EPSG 4269 to match us_states #set the study coordinate reference systemstudy_crs <-st_crs(bbox_polygon)#create the study area polygon study_area <-st_crop(us_states, bbox_polygon)#filter the "us_states" data to just FloridaFL_state <-filter(us_states, NAME =="Florida")#tidy everglades mangrove dataEVG_tidy <- EVG_raw_data %>%subset(dbh_cm !="NA") %>%group_by(plot_ID) %>%mutate(across(contains("date"), ~mdy(.))) %>%arrange(., date)#convert tidies mangrove data into an st object EVG_tidy_spat <- EVG_tidy %>%st_as_sf(coords =c("longitude", "latitude"), crs = study_crs)#gather the distinct values of plot_ID, site name and geometryEVG_plots <- EVG_tidy_spat %>%distinct(plot_ID, site, geometry)#map_tracks for the hurricanes passing through the study area hurr_paths <-map_tracks(storms =c("Andrew-1992","Irene-1999","Katrina-2005"))#extract the specific point data from the hurr_paths object,#convert it into a st object, and crop it to fit the state of Floridahurr_paths_spat <- hurr_paths[["layers"]][[2]][["data"]] %>%st_as_sf(.,coords =c("longitude", "latitude"), crs = study_crs) %>%st_crop(.,FL_state)#crop hurr_paths to just the study areastudy_area_hurr_paths <-st_crop(hurr_paths_spat, study_area)
Results:
Map the spatial distribution of study plots
Code
if (knitr::is_html_output()) {plot_dist <-leaflet(EVG_tidy_spat)%>%addTiles()%>%fitBounds(., lng1 =-81.2133, lat1 =25.5831,lng2 =-80.9134, lat2 =25.1589)%>%addMarkers(data = EVG_plots, popup = EVG_plots$site)plot_dist} else { plot_dist_static <-ggplot()+geom_sf(data = study_area)+geom_sf(data = EVG_plots, aes(color = plot_ID))+theme(axis.text.x =element_text(angle =270, vjust =0.5, hjust=1))+labs(title ="Study Area with Study Plots", color ="Plot ID")+theme(plot.title =element_text(hjust =0.5)) plot_dist_static}
Plot the path of hurricanes in the context of the study area and plots
Code
#make a map showing Florida, the study area and the hurricane pathsstate_view <-ggplot() +geom_sf(data = FL_state) +geom_sf(data = study_area, fill ="green")+geom_sf(data = hurr_paths_spat, aes(color = storm_id))+labs(title ="Florida with the Study Area and Hurricane Paths", color ="Storm ID")+theme(plot.title =element_text(hjust =0.5))state_view
Code
#make a map showing the study area, cropped hurricane paths and study plots study_area_view <-ggplot()+geom_sf(data = study_area)+geom_sf(data = study_area_hurr_paths, aes(color = storm_id))+geom_sf(data = EVG_plots, shape =15)+theme(axis.text.x =element_text(angle =270, vjust =0.5, hjust=1))+labs(title ="Study Area with Hurricane Paths and Study Plots", color ="Storm ID")+theme(plot.title =element_text(hjust =0.5))study_area_view
Plot species make up of each plot over time
Code
#make a cluster and register it for parallel processingcl <-makeCluster(4)registerDoParallel(cl)#factor plots for use in foreach loopsites <-factor(EVG_tidy$plot_ID)#create foreach loop that calculates plot make-up, #then makes and saves a bar chart #representing the quantity fo each species overtimeplot_makeup_list <-foreach(i =levels(sites), .packages =c("dplyr", "ggplot2")) %dopar% { plot_makeup <- EVG_tidy %>%filter(plot_ID == i) %>%group_by(plot_ID, date, scientific_name) %>%summarize(individual_count =n(), .groups ="drop") bar_chart <-ggplot(data = plot_makeup, aes(fill = scientific_name, y = individual_count, x = date)) +geom_bar(position ="fill", stat ="identity")+labs(x ="Date", y ="Proportion of Species", fill ="Scientific Name") +ggtitle(paste("Plot", i))+theme(plot.title =element_text(hjust =0.5)) file_name <-paste("plot_", i, ".png", sep ="")ggsave(filename = file_name, plot = bar_chart, width =10, height =8, units ="in")return(file_name)}stopCluster(cl)#convert saved .png files to graphic objects (grobs) for use with grid.arrangegrob_list <-lapply(plot_makeup_list, function(file_name) { img <-readPNG(file_name) raster_grob <-rasterGrob(img) return(raster_grob)})#arrange grobs in a grid and save image all_plots <-grid.arrange(grobs = grob_list, ncol =4, nrow =6)
Find and visualize the directly impacted plot/plots
Find the plot/ plots that fall in the direct line of a hurricane
Code
# Transform to a projected coordinate system (NAD83 / UTM Zone 17N)EVG_plots_trans <-st_transform(EVG_plots, crs =32617) study_area_hurr_paths_trans<-st_transform(study_area_hurr_paths, crs =32617)study_area_trans <-st_transform(study_area, crs =32617) # Create buffer for EVG_plots and hurr_paths (1 mile = 1609.34 meters)EVG_plots_buffer <-st_buffer(EVG_plots_trans, dist = (1609.34*0.5))study_area_hurr_paths_buffer <-st_buffer(study_area_hurr_paths_trans, dist = (1609.34))# Find the intersection of the buffersoverlap <-st_intersection(EVG_plots_buffer, study_area_hurr_paths_buffer)overlap_table<-overlap %>% st_drop_geometry%>%kbl() %>%kable_minimal()overlap_table
plot_ID
site
storm_id
date
BRU
Broad River
Irene-1999
1999-10-15 21:00:00
BRM
Broad River
Katrina-2005
2005-08-26 04:45:00
Code
#Visualize the intersectionggplot()+geom_sf(data = study_area_trans)+geom_sf(data = study_area_hurr_paths_buffer, aes(color = storm_id))+geom_sf(data = study_area_hurr_paths_trans, aes(color = storm_id))+geom_sf(data = EVG_plots_buffer, aes(fill = site))+geom_sf(data = overlap, aes(fill = site), color="yellow")+labs(title ="Intersection of Study Plots and Hurricane Paths",fill ="Site Name", color ="Storm ID",caption ="Study Plots have a 1/2 mile Buffer, Hurricane Paths have a 1 mile Buffer")+theme(plot.title =element_text(hjust =0.5),plot.caption =element_text(hjust =0.5),axis.text.x =element_text(angle =270, vjust =0.5, hjust=1))
Plot the average DBH of each species in the impacted plot/plots over time
Code
#filter the mangrove data based on the overlap results, #the first overlap plot is BRUplot_BRU <- EVG_tidy_spat %>%filter(plot_ID =="BRU") %>%subset(status !="dead") %>%group_by(date, scientific_name) %>%mutate("avg_species_dbh_cm"=mean(dbh_cm)) #averaged to minimize #visual noise#plot the change in average Diameter at Breast Height (DBH) for each species over timeggplot(plot_BRU) +geom_point(aes(x = date, y = avg_species_dbh_cm, color = scientific_name)) +geom_line(aes(x = date, y = avg_species_dbh_cm, color = scientific_name, group = scientific_name)) +geom_vline(xintercept =as.Date("1999-10-15"), color ="gray", linetype ="dashed", linewidth =1) +annotate("text",x =as.Date("1999-10-15"), y =0, label ="Hurricane Irene", color ="black")+labs(title ="Change in Average DBH per Species in Plot BRU", x ="Date", y ="Average DBH in cm", color ="Scientific Name")+theme(plot.title =element_text(hjust =0.5))+theme_minimal()
Code
#filter the mangrove data based on the overlap results again,#the second overlap plot is BRMplot_BRM <- EVG_tidy_spat %>%filter(plot_ID =="BRM") %>%subset(status !="dead") %>%group_by(date, scientific_name) %>%mutate("avg_species_dbh_cm"=mean(dbh_cm))#plot the change in average DBH for each species over time ggplot(plot_BRM) +geom_point(aes(x = date, y = avg_species_dbh_cm, color = scientific_name)) +geom_line(aes(x = date, y = avg_species_dbh_cm, color = scientific_name, group = scientific_name)) +geom_vline(xintercept =as.Date("1999-10-15"), color ="gray", linetype ="dashed", linewidth =1) +annotate("text",x =as.Date("1999-10-15"), y =0, label ="Hurricane Irene", color ="black")+geom_vline(xintercept =as.Date("2005-08-26"), color ="gray", linetype ="dashed", linewidth =1) +annotate("text",x =as.Date("2005-08-26"), y =0, label ="Hurricane Katrina", color ="black")+xlim(min(plot_BRM$date), max(plot_BRM$date) +365) +labs(title ="Change in Average DBH per Species in Plot BRM", x ="Date", y ="Average DBH in cm", color ="Scientific Name")+theme(plot.title =element_text(hjust =0.5))+theme_minimal()
Results (Text):
Through this project I was able to plot the location of the twenty-three permanent plots in the context of the state of Florida in an interactive map. Using a static map, I visualized the study area within the state of Florida and the various hurricanes that have passed through the area of interest during the period of study. I also made a map of the plots within the study area and the hurricane paths at that scale. Furthermore, I used bar charts that had the proportion of each species observed on each study date to visualize the species make-up of each plot over time. These plots are useful for understanding which species thrive at different hurricane exposure levels and for comparing how the species make-up changes through time. Using a buffer I was able to determine which sites were most directly impacted by the various hurricanes that passed through the study area. The result of this buffer and intersection was the identification of two plots that were directly in a hurricane’s path. Using the average Diameter at Breast Height (DBH) for each species I visualized the change in these trees over time.
Conclusions:
For both directly impacted plots the species Laguncularia racemosa and Rhizophora mangle had increasing average DBH over time indicating that the trees were still thriving despite being in the direct line of a hurricane. Looking at the species make-up bar charts for both of these plots supports this idea because both species are observed to have a general increase in the number of individuals. On the other hand, the species Avicennia germinans, in Plot BRM, had a sharp decline then continued to have a slow decline in average DBH indicating the loss of some larger individuals early in the study period and then a continued struggle to grow and the loss of more individuals as the study period progressed. This is reflected in the bar chart of species make-up for this plot, the bar for the first year has a larger portion of Avicennia germinans than any of the following years’ bars. These limited findings indicate that there is merit in using past changes in species make-up and changes in traits such as DBH to understand which species of mangroves fare better in hurricane zones. This research can be further expanded by looking at more traits such as stem breakage of the trees identified here to see if these conclusions are consistent across multiple assessment criteria.
References:
Anderson, B. (2024). Geanders/hurricaneexposuredata [R]. https://github.com/geanders/hurricaneexposuredata (Original work published 2016)
Bivand, R., Nowosad, J., Lovelace, R., Mimis, A., dataset), M. M. (author of the state vbm, & dataset), G. S. (author of the state vbm. (2024). spData: Datasets for Spatial Analysis (Version 2.3.3) [Computer software]. https://cran.r-project.org/web/packages/spData/index.html
Doyle, T. W., Smith, T. J., & Robblee, M. B. (1995). Wind Damage Effects of Hurricane Andrew on Mangrove Communities Along the Southwest Coast of Florida, USA. Journal of Coastal Research, 159–168.
geanders/hurricaneexposure: Functions to Create County-level Time Series of Hurricane Exposure. (n.d.). Retrieved December 12, 2024, from https://github.com/geanders/hurricaneexposure
Laura C Feher, Thomas J. Smith III, Gordon H Anderson, Ginger Tiling-Range, Karen M. Balentine, Greg A Ward, Kevin R. T. Whelan, Christa L. Walker, Andre Daniels, Michael J Osland, & Fara S. Ilami. (n.d.). Everglades mangrove vegetation data from 23 long-term plots (1992-2011) [Dataset]. U.S. Geological Survey. https://doi.org/10.5066/P13DONMU
Smith III, T. J., Anderson, G., Balentine, K., Tiling, G., Ward, G., & Whelan, K. (2009). Cumulative Impacts of Hurricanes on Florida Mangrove Ecosystems: Sediment Deposition, Storm Surges and Vegetation. Wetlands, 29, 24–34. https://doi.org/10.1672/08-40.1
