## Class "Spatial" [package "sp"]
## 
## Slots:
##                               
## Name:         bbox proj4string
## Class:      matrix         CRS
## 
## Known Subclasses: 
## Class "SpatialPoints", directly
## Class "SpatialMultiPoints", directly
## Class "SpatialGrid", directly
## Class "SpatialLines", directly
## Class "SpatialPolygons", directly
## Class "SpatialPointsDataFrame", by class "SpatialPoints", distance 2
## Class "SpatialPixels", by class "SpatialPoints", distance 2
## Class "SpatialMultiPointsDataFrame", by class "SpatialMultiPoints", distance 2
## Class "SpatialGridDataFrame", by class "SpatialGrid", distance 2
## Class "SpatialLinesDataFrame", by class "SpatialLines", distance 2
## Class "SpatialPixelsDataFrame", by class "SpatialPoints", distance 3
## Class "SpatialPolygonsDataFrame", by class "SpatialPolygons", distance 2

The purpose of this study is to analyze Wildfire Trends from 1984-2021. As our global climate is connected to the natural environment around us, it has become increasingly important to track trends in natural hazards so that we may adjust our economic/human resources to mitigate negative impacts caused.

Data for Analysis

1. Utah Surface Management Agency (SMA)

a. This dataset was collected by the BLM and retrieved from the Department of Interior’s website. The layers in this dataset are polygon features depicting the surface area of Utah and which management agency oversees them.

b. https://data.doi.gov/dataset/blm-ut-surface-management-agency-polygon-3e9c5

2. United States Wildfires

a. This dataset was collected by the US Forest Service and retrieved from the Department of Agriculture’s website. The layers within this vector point shapefile depict points where wildfires have occurred since 1984.

b. https://data.fs.usda.gov/geodata/edw/datasets.php?xmlKeyword=monitoring+trends+in+burn+severity

c. A boundary shapefile map is available for this dataset; however, the file size is very large so the vector point file will be used as this study is only interest in general location and counts of wildfires. For those interested in exporting this data in the form of a KML/KMZ output, the file size and complexity is too great.

d. This dataset inventories wildfires greater than 500 acres in the eastern U.S. and 1,000 acres in the western U.S. While the study does show some limitation in that wildfires are confirmed using satellite imagery and smaller fires are omitted, the dataset is quite extensive and accurate showing recent historical wildfires. On a greater scale of time and history, this dataset does not provide a greater long-term depiction of wildfire behavior.

3. Land Cover

a. This dataset was collected by the SWReGAP project which is a multi-institutional group. The layers within this raster file depict landcover over a five state region to assess biodiversity.

b. https://swregap.org/data/browse/?prefix=public/data/landcover/UT/

c. The generalizability of this study is outstanding in expressing numerous land cover types validated and assessed across multiple agency groups. It is interesting to note the study is limited to 5 states in the intermountain west and southern U.S. as inclusion of northern states in the intermountain region would reduce bias to land cover types dominate in southern climates.

Load and Clean Data:

1. Surface Management (SMA) Shapefile

## OGR data source with driver: ESRI Shapefile 
## Source: "C:\Users\joshc\Documents\USU\2023\GEOG 4870\FInal Project\GEOG4870 Final\Final Data\SMA\utah_sma.shp", layer: "utah_sma"
## with 12009 features
## It has 9 fields

Data Cleaned

Drop column 4 as it was all NA

Drop missing values within rows

Remove duplicate rows

2. Wildfire Shapefile

## OGR data source with driver: ESRI Shapefile 
## Source: "C:\Users\joshc\Documents\USU\2023\GEOG 4870\FInal Project\GEOG4870 Final\Final Data\US Wildfires\S_USA.MTBS_FIRE_OCCURRENCE_PT.shp", layer: "S_USA.MTBS_FIRE_OCCURRENCE_PT"
## with 29926 features
## It has 22 fields
## Integer64 fields read as strings:  MAP_ID DNBR_OFFST DNBR_STDDV

Data Cleaned

Drop columns with no data: wildfires <- wildfires[, -c(14,17,20)]

Drop missing values within rows: wildfires <- wildfires %>% na.omit()

Remove duplicate rows: wildfires <- (wildfires[!duplicated(wildfires$FIRE_ID), ])

## class       : SpatialPointsDataFrame 
## features    : 29926 
## extent      : -166.0911, -65.35063, 17.95656, 70.13926  (xmin, xmax, ymin, ymax)
## variables   : 19
## # A tibble: 29,926 × 19
##    FIRE_ID       FIRE_NAME ASMNT_TYPE FIRE_TYPE NODATA_THR GREENNESS_ LOW_THRESH
##    <chr>         <chr>     <chr>      <chr>          <dbl>      <dbl>      <dbl>
##  1 AZ3359910999… MUD SPRI… Initial    Wildfire        -970       -150        100
##  2 TX3129409771… OWL CREEK Initial    Wildfire        -970       -150        110
##  3 FL2686508190… UNNAMED   Initial (… Unknown         9999       9999        400
##  4 CO4001210858… SPRING C… Extended   Wildfire        -970       -150         80
##  5 AZ3586611185… ANDERSON  Extended   Wildfire        -970       -150         20
##  6 OR4334712281… WILLIAMS… Extended   Wildfire        -970       -150        110
##  7 FL2685008188… UNNAMED   Initial (… Unknown         9999       9999        400
##  8 FL2763008118… UNNAMED   Initial (… Unknown         9999       9999        350
##  9 UT3779310991… COOPER S… Extended   Wildfire        -970       -150        120
## 10 WA4649412121… DISCOVERY Extended   Wildfire        -970       -150        100
## # ℹ 29,916 more rows
## # ℹ 12 more variables: MODERATE_T <dbl>, HIGH_THRES <dbl>, IG_DATE <chr>,
## #   LATITUDE <dbl>, LONGITUDE <dbl>, ACRES <dbl>, MAP_ID <chr>, MAP_PROG <chr>,
## #   DNBR_OFFST <chr>, DNBR_STDDV <chr>, PRE_ID <chr>, POST_ID <chr>

3. Land Cover

Data Preparation

Research questions and necessary data.

1. Which areas in the U.S. are at most risk of wildfires?

a. us.cities data

b. Fire point locations within the US

2. Has there been an increase in large wildfires since 1984?

a. Wildfire Shapefile

b. Key variables within the dataset are the years that individual wildfires occur and how large they are measured in the “acres”.

4. Brainstorming

1. Frequency of large wildfires:

a. Filter wildfires that are well above the average size and visualize any change in the frequency of large fires overtime.

2. General wildfire trends.

a. Separate the entire dataset into multiple periods of time than compare wildfire change on line charts throughout each time period.

DATA EXPLORATION

5. Summarize key variables of interest.

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	29926	6918	25127	500	1168	1930	4213	1068802
year	29926	2006	10	1984	1999	2008	2014	2021
GREENNESS_	29926	2984	4691	-650	-150	-150	9999	9999
LOW_THRESH	29926	14	1148	-9999	40	75	164	9999
MODERATE_T	29926	23	6612	-9999	-120	277	400	9999
HIGH_THRES	29926	1048	8058	-9999	-9999	550	9999	9999

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	846	8078	26515	502	1184	2050	4606	333571
year	846	2008	8.6	1990	2000	2010	2015	2020
GREENNESS_	846	2464	4442	-320	-150	-150	9999	9999
LOW_THRESH	846	64	936	-9999	46	75	150	740
MODERATE_T	846	848	6121	-9999	150	292	449	9999
HIGH_THRES	846	1622	7727	-9999	-87	580	9999	9999

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	29926	6918	25127	500	1168	1930	4213	1068802
year	29926	2006	10	1984	1999	2008	2014	2021

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	10056	6211	19009	500	1193	1994	4465	563527
year	10056	1994	5.6	1984	1989	1994	1999	2002
GREENNESS_	10056	3581	4895	-500	-150	-150	9999	9999
LOW_THRESH	10056	55	1062	-9999	50	100	220	1100
MODERATE_T	10056	-1285	6060	-9999	-9999	223	335	9999
HIGH_THRES	10056	-696	7518	-9999	-9999	435	710	9999

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	19870	7276	27706	500	1153	1899	4083	1068802
year	19870	2012	5	2003	2008	2011	2016	2021
GREENNESS_	19870	2682	4555	-650	-150	-150	9999	9999
LOW_THRESH	19870	-6	1188	-9999	40	65	150	9999
MODERATE_T	19870	685	6779	-9999	89	300	9999	9999
HIGH_THRES	19870	1931	8177	-9999	-9999	600	9999	9999

Summary Findings:

1. Data collection from 1990-2020 expresses wildfires ranging between 500 to 333,571 acres in size. The median fire size of 2,050 acres more accurately depicts the average size of wildfires than the mean of 8,078 acres which is likely skewed by outlying fires much large in scope than the majority of wildfires.

2. While the minimum and maximum year values were determined by the study period, the positive skew of the data towards later date periods is an indicator of an increasing frequency of wildfires towards the 2nd half of the study period.

6. Create a boxplot to identify outliers.

Boxplot of years 1984-2002 and 2003-2021:

Outlier Findings:

As clearly depicted by the boxplots, there are numerous outliers of large wildfires that skew the data each year. The majority of outlier data will be kept as large scale wildfires are important to overall trends for Wildfires. However, summary statistics were also be analysed with the outliers removed.

When comparing the first half of the study period vs. the 2nd, the total incidents of wildfires increased from 10,056 (1984-2002) to 19,870 wildfires (2003-2021)

7. Compute and visualize statistics for acres burned and year.

Summary statistics for the two most important varables in measuring wildfire size over time: Acres and Years. The study period was separated into two halves and 5 year increments to represent wildfire history from the 1st half of the study period and the 2nd half of the study period.

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	846	8078	26515	502	1184	2050	4606	333571
year	846	2008	8.6	1990	2000	2010	2015	2020

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	10056	6211	19009	500	1193	1994	4465	563527
year	10056	1994	5.6	1984	1989	1994	1999	2002

Summary Statistics
Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 50	Pctl. 75	Max
ACRES	19870	7276	27706	500	1153	1899	4083	1068802
year	19870	2012	5	2003	2008	2011	2016	2021

8. Compute and visualize correlation.

## [1] 0.02212699

## [1] 0.01879452

Correlation between size and time seperated into two time periods; 1984-2002 and 2003-2021.

## [1] 0.02876973

## [1] 0.002651568

The correlation analysis showed little correlation between the size of wildfires and timeframe burns occur. Only a slight positive correlation was found over the years.

9. Create an exploratory and multivariate visualization.

The analysis chart confirmed numerous outliers of wildfires vastly larger than median (2050); specifically in the years 1990, 2010, 2015, and 2020.

The line chart measuring the average size of wildfires from 1984 indicates a gradually increasing trend with a spike in acres burned from 2002-2005.

The line chart comparing prescribed burns vs wildfires indicates prescribed burnes have consistently burned at approximately 1,000 acres since 1984 while wildfires burned at approximately 15,000 acres on average in 1990 increasing to an approximate average of 20,000 acres in 2021.

Trends separated into two 19 year study periods:

The first half of the study period from 1984-2002 experienced an average increase in wildfire size from 5,700 to 8,800 acres . Approximately a 35% increase. The 2nd half of the study period (2003-2021) decreased from 9,200 acres on average to approximately 7,500 in 2021. A near 18% decrease in average size. While there seems to be a general increase in size overtime, a spike in large wildfires between 2003-2005 with a sudden decrease in the following years shows unpredictability in wildfire trends.

Spatial Integration and Visualization

Data Preperation

1. Check Projections

## [1] TRUE

2 Crop wildfires and cities to Utah. Merge two datasets together.

## [1] "SpatialPointsDataFrame"
## attr(,"package")
## [1] "sp"

##  [1] "TX" "OH" "CA" "GA" "NY" "OR" "NM" "LA" "VA" "PA" "FL" "IA" "AK" "IN" "MA"
## [16] "MI" "MD" "MN" "WI" "IL" "CO" "NC" "NJ" "AL" "WA" "ME" "AZ" "TN" "NE" "MT"
## [31] "MS" "ND" "MO" "ID" "UT" "KY" "CT" "OK" "NV" "WY" "SC" "WV" "NH" "AR" "RI"
## [46] "DE" "HI" "KS" "VT" "SD" "DC"

## [1] "UT"

## [1] "character"

## [1] "character"

3: Filter and create new dataset pertaining to wildfires over 150,000 acres.

4. Large Wildfire Distribution ( > 150,000 : United States)

5. Spatial Visualization of Wildfires Greater Than 150,000 Acres.

The line and point plots depict an increasing trend in the frequency of large wildfires over 150,000 acres. While the initial data collected in the 1980s showed more incidents than the 1990s, the point plot shows much greater frequency of large wildfires after 2005.

6. Create a spatial visualization of your data using tmap. Describe the visualization as well as any interesting patterns or trends

## Error in wk_handle.wk_wkb(wkb, s2_geography_writer(oriented = oriented,  : 
##   Loop 0 is not valid: Edge 344 crosses edge 346

The tmap shows the majority of large wildfires occur in the western part of the United States. The majority of fires greater than 800,000 acres occured in California and Oregon with a few of the largest wildfires recorded near the Oklahoma-Kansas border.

Wildfire Trends

Josh Carr

2023-04-01

Data for Analysis

1. Utah Surface Management Agency (SMA)

a. This dataset was collected by the BLM and retrieved from the Department of Interior’s website. The layers in this dataset are polygon features depicting the surface area of Utah and which management agency oversees them.

b. https://data.doi.gov/dataset/blm-ut-surface-management-agency-polygon-3e9c5

2. United States Wildfires

a. This dataset was collected by the US Forest Service and retrieved from the Department of Agriculture’s website. The layers within this vector point shapefile depict points where wildfires have occurred since 1984.

b. https://data.fs.usda.gov/geodata/edw/datasets.php?xmlKeyword=monitoring+trends+in+burn+severity

3. Land Cover

a. This dataset was collected by the SWReGAP project which is a multi-institutional group. The layers within this raster file depict landcover over a five state region to assess biodiversity.

b. https://swregap.org/data/browse/?prefix=public/data/landcover/UT/

Load and Clean Data:

1. Surface Management (SMA) Shapefile

Data Cleaned

Drop column 4 as it was all NA

Drop missing values within rows

Remove duplicate rows

2. Wildfire Shapefile

Data Cleaned

Drop columns with no data: wildfires <- wildfires[, -c(14,17,20)]

Drop missing values within rows: wildfires <- wildfires %>% na.omit()

Remove duplicate rows: wildfires <- (wildfires[!duplicated(wildfires$FIRE_ID), ])

3. Land Cover

Data Preparation

Research questions and necessary data.

1. Which areas in the U.S. are at most risk of wildfires?

a. us.cities data

b. Fire point locations within the US

2. Has there been an increase in large wildfires since 1984?

a. Wildfire Shapefile

b. Key variables within the dataset are the years that individual wildfires occur and how large they are measured in the “acres”.

4. Brainstorming

1. Frequency of large wildfires:

a. Filter wildfires that are well above the average size and visualize any change in the frequency of large fires overtime.

2. General wildfire trends.

a. Separate the entire dataset into multiple periods of time than compare wildfire change on line charts throughout each time period.

DATA EXPLORATION

5. Summarize key variables of interest.

Summary Findings:

2. While the minimum and maximum year values were determined by the study period, the positive skew of the data towards later date periods is an indicator of an increasing frequency of wildfires towards the 2nd half of the study period.

6. Create a boxplot to identify outliers.

Boxplot of years 1984-2002 and 2003-2021:

Outlier Findings:

When comparing the first half of the study period vs. the 2nd, the total incidents of wildfires increased from 10,056 (1984-2002) to 19,870 wildfires (2003-2021)

7. Compute and visualize statistics for acres burned and year.

Summary statistics for the two most important varables in measuring wildfire size over time: Acres and Years. The study period was separated into two halves and 5 year increments to represent wildfire history from the 1st half of the study period and the 2nd half of the study period.

8. Compute and visualize correlation.

Correlation between size and time seperated into two time periods; 1984-2002 and 2003-2021.

The correlation analysis showed little correlation between the size of wildfires and timeframe burns occur. Only a slight positive correlation was found over the years.

9. Create an exploratory and multivariate visualization.

The analysis chart confirmed numerous outliers of wildfires vastly larger than median (2050); specifically in the years 1990, 2010, 2015, and 2020.

The line chart measuring the average size of wildfires from 1984 indicates a gradually increasing trend with a spike in acres burned from 2002-2005.

The line chart comparing prescribed burns vs wildfires indicates prescribed burnes have consistently burned at approximately 1,000 acres since 1984 while wildfires burned at approximately 15,000 acres on average in 1990 increasing to an approximate average of 20,000 acres in 2021.

Trends separated into two 19 year study periods:

Spatial Integration and Visualization

Data Preperation

1. Check Projections

2 Crop wildfires and cities to Utah. Merge two datasets together.

3: Filter and create new dataset pertaining to wildfires over 150,000 acres.

4. Large Wildfire Distribution ( > 150,000 : United States)

5. Spatial Visualization of Wildfires Greater Than 150,000 Acres.

The line and point plots depict an increasing trend in the frequency of large wildfires over 150,000 acres. While the initial data collected in the 1980s showed more incidents than the 1990s, the point plot shows much greater frequency of large wildfires after 2005.

6. Create a spatial visualization of your data using tmap. Describe the visualization as well as any interesting patterns or trends

The tmap shows the majority of large wildfires occur in the western part of the United States. The majority of fires greater than 800,000 acres occured in California and Oregon with a few of the largest wildfires recorded near the Oklahoma-Kansas border.