FOR 128 Final Project

Authors

Nicole Bommarito

Abigail Matson

Avrie Hoell

Kaily Hurt

Voka Schiller

Published

December 9, 2024

knitr::include_graphics("NETN_Parks.png")

Introduction

In our project we chose to work with the Northeast Temperate Inventory & Monitoring Network (NETN),(2006-2004) data package from The National Park Data Base. Using this data, we aim to analyze the relationship between tree growth and the environment as well as the way trees can influence their environment. Specifically we investigate factors like the conditions of the forest floor and their relation to tree height classification, for both how fores

#| code-fold: TRUE
#| code-summary: "Code"

suppressWarnings(suppressMessages(library(tidyverse)))

suppressWarnings(suppressMessages(library(dplyr)))
suppressWarnings(suppressMessages(library(gt)))

Methods

Park Data

The NETN data set includes the 13 parks within the network which fall along the eastern coast of the united states, stretching from Maine to New Jersey. It monitors each plot once per year, stretching 18 years so far.

This CSV in the data set examines the Tree Foliage of plots within these parks

tree_foliage <- read_csv( "TreesFoliageCond_NETN.csv", show_col_types =  FALSE)

dim(tree_foliage)
[1] 35511    30
head(tree_foliage) %>% select(1:10) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode IsAbandoned PanelCode SampleDate IsQAQC
SAGA-017 NETN SAGA SAGA VS 17 FALSE 3 2012-06-14 FALSE
ROVA-004 NETN ROVA ROVA_HOFR_West VS 4 FALSE 2 2011-06-02 FALSE
ACAD-057 NETN ACAD ACAD_MDI VS 57 FALSE 2 2019-06-19 FALSE
SARA-021 NETN SARA SARA VS 21 FALSE 3 2008-06-02 FALSE
MABI-023 NETN MABI MABI VS 23 FALSE 3 2016-06-08 FALSE
ACAD-034 NETN ACAD ACAD_MDI VS 34 FALSE 1 2010-07-14 FALSE

This next CSV examines 5 different forest floor conditions within the parks as well as their health and height

forest_floor <-read_csv("StandForestFloor_NETN.csv", show_col_types = FALSE)
dim(forest_floor)
[1] 10158    21
head(forest_floor) %>% select(1:6) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode
ACAD-001 NETN ACAD ACAD_Schoodic VS 1
ACAD-001 NETN ACAD ACAD_Schoodic VS 1
ACAD-001 NETN ACAD ACAD_Schoodic VS 1
ACAD-001 NETN ACAD ACAD_Schoodic VS 1
ACAD-001 NETN ACAD ACAD_Schoodic VS 1
ACAD-001 NETN ACAD ACAD_Schoodic VS 1

This CSV examines different individual trees, plots, and their height over the years

head(forest_floor) %>% select(1:10) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode IsAbandoned PanelCode SampleDate IsQAQC
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 6/6/2006 FALSE 
tree_heights <- read_csv("StandTreeHeights_NETN (1).csv", show_col_types = FALSE)
dim(tree_heights)
[1] 6876   21
head(tree_heights) %>% select(1:10) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode IsAbandoned PanelCode SampleDate IsQAQC
ACAD-009 NETN ACAD ACAD_MDI VS 9 FALSE 1 2006-06-14 FALSE
ACAD-009 NETN ACAD ACAD_MDI VS 9 FALSE 1 2010-07-13 FALSE
ACAD-139 NETN ACAD ACAD_MDI VS 139 FALSE 4 2009-07-08 FALSE
ROVA-024 NETN ROVA ROVA_VAMA VS 24 FALSE 4 2009-06-04 FALSE
MABI-006 NETN MABI MABI VS 6 FALSE 1 2006-07-26 FALSE
MABI-007 NETN MABI MABI VS 7 FALSE 1 2010-06-10 FALSE

The 4th CSV measures individual trees and what conditions they classify under

tree_conditions <- read_csv("TreesConditions_NETN.csv", show_col_types = FALSE)
dim(tree_conditions)
[1] 37464    55
head(tree_conditions) %>% select(1:10) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode IsAbandoned PanelCode SampleDate IsQAQC
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE

Finally, the last data set examines the plant cover within the plots ( notably it can not individual trees)

plant_cover <- read_csv("StandPlantCoverStrata_NETN.csv", show_col_types = FALSE)
head(plant_cover) %>% select(1:10) %>% gt()
Plot_Name Network ParkUnit ParkSubUnit PlotTypeCode PlotCode IsAbandoned PanelCode SampleDate IsQAQC
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2006-06-06 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2010-07-27 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2010-07-27 FALSE
ACAD-001 NETN ACAD ACAD_Schoodic VS 1 FALSE 1 2010-07-27 FALSE

Filtering the Data

This data set has a lot of columns, to make it easier to work with we are going to remove columns not relevant to what we are trying to investigate.

tree_conditions <- tree_conditions %>%
select(-ExportDate, -contains(c("DPL")), -ProtectedStatusCode, -IsAbandoned, -IsQAQC, -SampleDate, -PlotTypeCode, -PlotCode)

We are also going to be pivoting our forest floor data set to tidy it up

forest_floor <- forest_floor %>%
select(-c("PlotTypeCode", "PlotCode", "IsAbandoned", "SampleDate", "IsQAQC", "ForestFloorCode", "CoverClassLabel", "DPLCode", "DPLUserID", "DPLDate", "EventID", "ExportDate"))
plant_cover <-plant_cover %>%
select(-ExportDate, -contains(c("DPL")), -IsQAQC, -IsAbandoned, -SampleDate, -PlotTypeCode, -PlotCode, -StrataCode, -CoverClassLabel)
tree_foliage <- tree_foliage %>%
select(-ExportDate, -contains(c("DPL")), -IsQAQC, -IsAbandoned, -SampleDate, -EventID, -PlotID, -Network, -ParkUnit, -ParkSubUnit, -PanelCode, -PlotCode, -PercentLeavesLabel)

Joining the Data

Now that our data is cleaner, we need to combine the CSV datasets into one dataset we can work from

forest_floor and plant_cover join

forest_floor <- forest_floor %>%
group_by(ForestFloorLabel) %>%
mutate(row = row_number()) %>%
tidyr::pivot_wider(names_from = ForestFloorLabel, values_from = CoverClassCode)

plant_cover <- plant_cover %>%
group_by(StrataLabel) %>%
mutate(row = row_number()) %>%
tidyr::pivot_wider(names_from = StrataLabel, values_from = CoverClassCode)

combined_forest_plant <- left_join(forest_floor, plant_cover, by = c("row", "Network", "ParkUnit", "ParkSubUnit", "PanelCode","PlotID")) %>%
select(-Plot_Name.y, -SampleYear.y)

Results

combined_forest_plant_long <- combined_forest_plant %>%
  pivot_longer(cols = c("Bare Soil", "Rock", "Water", "Trampled", "Non-Vascular", "Lichen"),
               names_to = "Condition_Type", values_to = "Condition_Value") 
combined_forest_plant_long <- combined_forest_plant_long %>%
  filter(!Condition_Value %in% c("NC", "PM", "0"))

P1<- ggplot(combined_forest_plant_long, aes(x = Condition_Type, fill = Condition_Value)) +
  geom_bar(position = "dodge") +
  labs(title = "Distribution of Condition Types Across Forest Floor Coverage",
       x = "Condition Type", y = "Count") +
  facet_wrap(~ParkUnit, scales = "free") +
  theme_bw()
ggsave("condition_types.png", plot = P1, width = 20, height = 15)
knitr::include_graphics("condition_types.png")

For this graph we filtered out conditions like 0, which indicated this condition type covered 0% of the forest floor and conditions like “permanently missing” or “not collected”. As we can see, the condition’s vary greatly across the parks. Many do not ever experience the “Lichen” condition which is a combination of fungus an algae. For most parks in this data set, conditions like “Non-vascular” or Rock tend to often cover at least a small percentage of the surface area of the plots within the park. These observations may in part be due to park size and number of trees ( as you can see the counts are much higher for Acadia (ACAD) than many of the other parks in the network). This insight into the forest floor compositions of the parks can help us identify which forest floor conditions are positively correlated with average tree height within the park over time.

combined_long_coverage <- combined_forest_plant %>%
  pivot_longer(cols = c("Ground", "Mid-understory", "High-understory"),
               names_to = "Condition_Type", 
               values_to = "Vascular_Coverage_Rating")

P2 <- ggplot(combined_long_coverage, aes(x = Condition_Type, fill = as.factor(Vascular_Coverage_Rating))) +
  geom_bar(position = "stack") +
  facet_wrap(~ParkUnit,) +
  labs(title = "Count of Vascular Plant Coverage Ratings by Condition Type",
       x = "Condition Type", y = "Count of Coverage Rating") +
  scale_fill_brewer(palette = "Set3", name = "Vascular Coverage Rating (1-6)") +
  theme_minimal() 
ggsave("vascular_coverage_plot.png", plot = P2, width = 20, height= 15)

knitr::include_graphics("vascular_coverage_plot.png")

The relationship between trees and the forest floor conditions can be bidirectional as well.

In this example the “Ground” condition represents trees that are less than 0.5 meters above ground, “Mid-understory” represents trees that have heights between 0.5 and 2 meters above ground, and “high understory” represents trees that are over 2 meters above ground. in height.

From what we can see in the graph, trees classified as”Ground” are the only trees with 95-100% vascular plant coverage on the forest floor, this phenomena is observed across parks,this could indicate that high vascular plant coverage is negatively correlated with tree growth but it is more likely this observation is due to the fact that these trees lack large amounts of canopy cover due to their short stature, allowing for needed sunlight to reach the vascular plants on the forest floor. Also likely explaining why the taller trees tend to have higher counts of lower vascular plant coverage. Though interestingly, plots rated with 0% vascular plant coverage could be found across a variety of conditions. For example trees in Acadia National Park (ACAD) that were classified as “Ground” show the highest rates of 0% with high-understory showing the lowest. This may tie back to Acadia’s larger variety in conditions as well as high percentage of non-vascular plant forest floor coverage compared to other parks.

Tree_heights_clean <- tree_heights %>%
  filter(!is.na(TagCode))

Tree_heights_clean <- Tree_heights_clean %>%
  select(-PanelCode, 
         -IsAbandoned, 
         -SampleDate, 
         -IsQAQC, 
         -contains("DPL"),
         -ExportDate)
avg_tree_heights <- Tree_heights_clean %>%
  group_by(ParkUnit, SampleYear) %>%
  summarise(Average_Height = mean(Height, na.rm = TRUE))
`summarise()` has grouped output by 'ParkUnit'. You can override using the
`.groups` argument.
ggplot(avg_tree_heights, aes(x = SampleYear, y = Average_Height, color = ParkUnit, group = ParkUnit)) +
  geom_line() +                                 
  geom_point() +                                
  labs(title = "Average Tree Height by Park Over Time",
       x = "Year",
       y = "Average Tree Height (meters)") +
  theme_bw()

print(avg_tree_heights) %>% gt()
# A tibble: 58 × 3
# Groups:   ParkUnit [8]
   ParkUnit SampleYear Average_Height
   <chr>         <dbl>          <dbl>
 1 ACAD           2011           14.4
 2 ACAD           2012           15.2
 3 ACAD           2013           13.4
 4 ACAD           2014           12.8
 5 ACAD           2015           13.1
 6 ACAD           2016           14.2
 7 ACAD           2017           13.7
 8 ACAD           2018           13.0
 9 ACAD           2019           13.5
10 ACAD           2021           14.0
# ℹ 48 more rows
SampleYear Average_Height
ACAD
2011 14.44887
2012 15.18538
2013 13.36230
2014 12.78370
2015 13.10598
2016 14.15951
2017 13.72857
2018 12.98045
2019 13.51667
2021 13.99456
2022 13.37591
2023 13.45625
2024 14.31102
MABI
2012 29.03333
2014 25.30149
2016 25.83922
2018 25.58438
2022 25.61698
2023 25.14545
MIMA
2012 20.65938
2014 22.32632
2016 17.41277
2018 22.12000
2022 19.53750
2023 22.83929
MORR
2011 30.55250
2013 28.77273
2015 25.51268
2017 28.09286
2019 27.67941
2022 27.47959
2024 27.74531
ROVA
2011 24.81692
2013 23.14842
2015 22.50488
2017 23.78108
2019 23.10168
2022 22.11458
2024 24.11346
SAGA
2012 26.67576
2014 26.19153
2016 24.36533
2018 25.44545
2022 23.27812
2023 24.63030
SARA
2012 23.80976
2014 18.64789
2016 21.14571
2018 19.65571
2022 22.68406
2023 20.27639
WEFA
2011 24.25333
2013 22.26923
2015 22.28667
2017 21.32143
2019 22.55172
2022 23.18571
2024 24.27632

A lot of interesting conclusions can be drawn from this graphic and table

For example as you can see the “WEFA” (Weir Farm National Historical Park) park has significantly shorter trees than the other parks in the NETN network, Though this significant difference cannot be explained by tree height alone, when you look back to it’s forest floor data you notice high proportions of the forest floor conditions “Rock” and “Trampled” as well as “Bare Soil”, WEFA also has well as higher percentages of high vascular plant coverage. Similar trends can be seen for other low ranking parks like MIMA which has a forest floor predominantly consisting of rock.

Another interesting thing you can observe through this data is the dip in tree heights experienced across all parks in the network in the early 2010s. This points to some sort of event in the area that likely killed off some of the taller trees in the parks. It also indicates forest floor conditions as a relatively unimportant factor in accounting for a park’s average tree heights in the face of environmental events.

Fortunately, you can also observe consistent gains on the average tree heights across the parks indicating these trees are living longer and growing taller. However some parks like SARA (Saratoga National Historical Park) continue to suffer in terms of average tree height. Indicating once again, more park specific environmental factors outside of forest floor condition to be the root of their average tree height decline.

#Conclusions

Overall, in our analysis of the relation between parks, forest floor conditions, and tree height. We found forest floor conditions to exhibit a mild correlation with tree height. In our graphical observations parks with higher proportion of rock and bare soil were shown to have fewer trees as well as shorter trees. This lower average tree height indicates a large proportion of shorter trees and thus more vascular plant coverage. This was observed in many parks who tended to rank lower on average height across years, these parks had large proportions of high vascular plan coverage within their “Ground” trees category. However, when observing the trends over time it is clear that forest floor conditions are more influenced by tree heights than tree heights are influenced by the forest floor. Significant changes in tree heights were observed during specific time periods indicating a regional/ park specific event that likely killed older, taller trees. Showing that even in forests with great conditions for tree growth, catastrophic environmental events can set forest growth back more than anything.

knitr::include_graphics("Acadia.jpg")