Oregon Wildfires

Import Data and Libraries

import datetime
from jdcal import jd2gcal
from matplotlib import gridspec
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
import sqlite3
import xarray as xr
import geopandas as gpd
from matplotlib import pyplot as plt

import cdsapi
import sklearn_pandas as skp
import altair as alt
from vega_datasets import data
from shapely.geometry import Point
import plotly.express as px
import geopandas as gpd
from sklearn.preprocessing import MinMaxScaler
import folium
from folium.plugins import MarkerCluster


custom_colors = ['#68A33E', '#A10702', '#FB9E60', '#FFFF82', '#0F0326']

Load in Raw Data

input_filename = 'FPA_FOD_20170508.sqlite'
conn = sqlite3.connect(input_filename)
query = '''
    SELECT
       *
    FROM
        Fires;
'''
df_raw = pd.read_sql_query(query, conn)

Clean Data and Extract Oregon Boundaries

drop_columns = ['NWCG_REPORTING_AGENCY',
                'NWCG_REPORTING_UNIT_ID',
                'NWCG_REPORTING_UNIT_NAME',
                'FIRE_NAME',
                'COMPLEX_NAME', 
                'OWNER_DESCR',
                'OWNER_CODE']
df_US= df_raw.drop(columns= drop_columns)
df_US['MONTH'] = df_US['DISCOVERY_DATE'].apply(lambda x: jd2gcal(x, 0)[1])
df_US['DAY'] = df_US['DISCOVERY_DATE'].apply(lambda x: jd2gcal(x, 0)[2])
geometry = [Point(xy) for xy in zip(df_US['LONGITUDE'], df_US['LATITUDE'])]
df_geo = gpd.GeoDataFrame(df_US, geometry=geometry)
df_geo.crs = "EPSG:4326"
states = alt.topo_feature(data.us_10m.url, feature='states')
us_states = gpd.read_file("cb_2018_us_state_500k/cb_2018_us_state_500k.shp")
oregon = us_states[us_states['NAME'] == 'Oregon']
df_OR = df_geo[df_geo.STATE == 'OR']
df_OR.head()
OBJECTID FOD_ID FPA_ID SOURCE_SYSTEM_TYPE SOURCE_SYSTEM SOURCE_REPORTING_UNIT SOURCE_REPORTING_UNIT_NAME LOCAL_FIRE_REPORT_ID LOCAL_INCIDENT_ID FIRE_CODE ICS_209_INCIDENT_NUMBER ICS_209_NAME MTBS_ID MTBS_FIRE_NAME FIRE_YEAR DISCOVERY_DATE DISCOVERY_DOY DISCOVERY_TIME STAT_CAUSE_CODE STAT_CAUSE_DESCR CONT_DATE CONT_DOY CONT_TIME FIRE_SIZE FIRE_SIZE_CLASS LATITUDE LONGITUDE STATE COUNTY FIPS_CODE FIPS_NAME Shape MONTH DAY geometry
31 32 32 FS-1418907 FED FS-FIRESTAT 0601 Deschutes National Forest 1 1-001 BJ9H None None None None 2005 2453406.5 36 0943 4.0 Campfire 2453407.5 37.0 1226 0.1 A 43.995556 -121.414167 OR 17 017 Deschutes b'\x00\x01\xad\x10\x00\x00\x90\xaf\xeb\xb4\x81... 2 5 POINT (-121.41417 43.99556)
35 36 36 FS-1418916 FED FS-FIRESTAT 0601 Deschutes National Forest 3 1-003 BK2W None None None None 2005 2453434.5 64 1042 4.0 Campfire 2453434.5 64.0 1107 0.1 A 44.043333 -121.386111 OR 17 017 Deschutes b'\x00\x01\xad\x10\x00\x00\x9c\x84_\x0b\xb6X^\... 3 5 POINT (-121.38611 44.04333)
43 44 44 FS-1418924 FED FS-FIRESTAT 0601 Deschutes National Forest 6 1-004 BMP6 None None None None 2005 2453465.5 95 1205 4.0 Campfire 2453465.5 95.0 1210 0.1 A 43.955556 -121.352222 OR 17 017 Deschutes b'\x00\x01\xad\x10\x00\x00\xc4\xf4\x10\xcf\x8a... 4 5 POINT (-121.35222 43.95556)
46 47 47 FS-1418935 FED FS-FIRESTAT 0614 Umatilla National Forest 1 019 BE07 None None None None 2005 2453440.5 70 1000 9.0 Miscellaneous 2453440.5 70.0 1401 5.0 B 44.911111 -119.696111 OR 69 069 Wheeler b'\x00\x01\xad\x10\x00\x00@\xf5\x9c\x15\x8d\xe... 3 11 POINT (-119.69611 44.91111)
48 49 49 FS-1418943 FED FS-FIRESTAT 0601 Deschutes National Forest 2 1-002 BKG2 None None None None 2005 2453417.5 47 1535 4.0 Campfire 2453417.5 47.0 1600 0.1 A 43.725278 -121.574167 OR 17 017 Deschutes b'\x00\x01\xad\x10\x00\x00\x98\x86\x8f%\xbfd^\... 2 16 POINT (-121.57417 43.72528) 
import matplotlib.pyplot as plt
import panel as pn
from ipywidgets import widgets
pn.extension()

years = df_OR['FIRE_YEAR'].unique().tolist()  # Convert to a list

year_dropdown = pn.widgets.Select(options=years, name='Select Year')

def update_map(selected_year):
    df_selected_year = df_OR[df_OR['FIRE_YEAR'] == selected_year]
    
    fig, ax = plt.subplots(figsize=(10, 10))

    oregon.plot(ax=plt.gca(), color='#68A33E', edgecolor='grey')
    df_selected_year.plot(ax=ax, facecolor="#FFFF82", edgecolor="#68A33E")

    # Format
    ax.set_axis_off()
    ax.set_aspect("equal")  # This forces an equal aspect ratio

    plt.show()

@pn.depends(selected_year=year_dropdown)
def update(selected_year):
    update_map(selected_year)

# Create the Panel app
dashboard = pn.Row(year_dropdown, update)

# Display the Panel app
dashboard.servable()

Oregon Wildfires by Year

When discussing wildfires in the US, California is usually front and center of the narrative. However, many other states that experience similar rates of burning, yet are often left off the table. We’d like to change that narrative, focusing on Oregon’s recent wildfire trends. To complete this study, we have chosen to dig into observations taken between 2014-15.

The map above highlights point data for observed wildfires across this span of time. To explore the data, use the toggle on the top to select a year. It is observed that, while wildfires occur in nearly every area of the state, they tend to be concentrated to the west, where a majority of the forests and heavily-forested areas are. However, in order to get better at preventing wildfires, it’s important to understand what factors may most contribute to their ignition. The following exploratory analyses use data that may contribute to predicting fires.

To obtain information about Oregon’s climate, we used the ‘riem’ package (available in R) to pull data from the Automated Surface Observing Systems (ASOS). Originally collected and stored by the National Weather Service, the ‘riem’ package is maintained by the Iowa Environmental Mesonet (IEM). We collected data related to dew point temperature, precipitation, and wind speed - all variables potentially related to wildfire activity. Our findings from these exploratory analyses are detailed below.

Tracking Wildfires Across the Year

custom_colors = ['#68A33E','#FFFF82','#FB9E60','#A10702', '#0F0326']  # Add your desired hex colors

df_freq_mon = df_OR.groupby(['MONTH', 'FIRE_YEAR']).size().unstack()
df_freq_mon.to_csv("wildfireData.csv")
counter_fig = 1
mon_ticks = ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec']
plt.figure(figsize=[11,5])
sns.heatmap(df_freq_mon, cmap=custom_colors, linewidth=.2, linecolor=[.9,.9,.9])
plt.yticks(np.arange(0.5,12.5), labels=mon_ticks, rotation=0, fontsize=12)
plt.xticks(fontsize=12);
plt.xlabel('')
plt.ylabel('Month', fontsize=13)
plt.title('Number of fire events in Oregon', fontsize=13)
plt.tight_layout()

First, we observe the raw number of wildfires that occur in Oregon. There is a clear pattern of the summer months, July and August, being the most common months for wildfire season. This is a break in the pattern seen across the US, where wildfres occur throughout most times of the year. This can lead us to thinking that the causes of wildfires in Oregon may hold different factors than those seen in other areas of the country.

Cause of Wildfires by Time of Year

# plot fire frequency by cause and day of year
cause_by_doy = df_OR.groupby(['STAT_CAUSE_DESCR','DISCOVERY_DOY']).size().unstack()
counter_fig +=1
plt.figure(figsize=[10,5])
ax = sns.heatmap(cause_by_doy,cmap=custom_colors,vmin=0,vmax=500) #'CMRmap_r' <- old color scheme
plt.xticks(np.arange(0.5,366.5,20), labels=range(1,366,20), rotation=0, fontsize=11)
plt.yticks(fontsize=11)
plt.ylabel('Fire Cause', fontsize=12)
plt.xlabel('Day of year', fontsize=12)
for borders in ["top","right","left","bottom"]:
    ax.spines[borders].set_visible(True)
plt.title('Distribution of OR fires by cause & day of year')
plt.tight_layout()

The above visualization describes trends in causes of wildfires in Oregon between 2014 and 2015. Overall, lightning appears to be a main contributor, majorly happening in the summer months. Other factors often play a part, though are less sustained over time. One example lies in the sharp and unsustained uptick in fires caused by fireworks that happen in the middle of the year, presumably the 4th of July celebrations. However, there are a few causes which appear to happen equally across the year, regardless of season - these include burning debris and campfires.

Download and Extract Climate Data

bendWeather = pd.read_csv("BendWeather.csv")

def convert_date(date_str):
    formats_to_try = ['%d/%m/%Y %H:%M:%S', '%m/%d/%Y %H:%M:%S']  # Add more formats if needed
    for fmt in formats_to_try:
        try:
            return pd.to_datetime(date_str, format=fmt)
        except ValueError:
            pass
    return None
bendWeather['interval60'] = pd.to_datetime(bendWeather['interval60'], format='%m/%d/%Y %H:%M')
bendWeather['Day'] = bendWeather['interval60'].dt.day
bendWeather['Month'] = bendWeather['interval60'].dt.month
bendWeather['Year'] = bendWeather['interval60'].dt.year
bendWeather['Hour'] = bendWeather['interval60'].dt.hour

bendWeather.head(2)
interval60 Temperature Precipitation Wind_Speed Dew_Point_Temp Day Month Year Hour
0 2020-01-01 00:00:00 46.4 0.02 15 41.0 1 1 2020 0
1 2020-01-01 01:00:00 44.6 0.01 14 41.0 1 1 2020 1

Analyzing Weather Patterns

Patterns in Wind Speed

# Annual Patterns in Wind Speed
df_freq_mon = bendWeather.groupby(['Month', 'Wind_Speed']).size().unstack()
# plot monthly frequency of fire events
counter_fig = 3
mon_ticks = ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec']
plt.figure(figsize=[11,5])
sns.heatmap(df_freq_mon, cmap=custom_colors, linewidth=.01, linecolor=[.9,.9,.9])
plt.yticks(np.arange(0.5,12.5), labels=mon_ticks, rotation=0, fontsize=12)
plt.xticks(fontsize=12);
plt.xlabel('')
plt.ylabel('Month', fontsize=13)
plt.title('Annual Precipitation Patterns in Oregon', fontsize=13)
plt.tight_layout()

The above visualization details average wind speeds my time of year for 2014-15. While there is typically a light breeze in the air year-round, the cooler months of late fall and early winter occasionally see heavier gusts - October through January may see winds higher than 30 miles per hour. Now below, we will combine the the wildfires for 2014 and 2015 in oregon with the corresponding weather data for the to produce a dahsboard in the analysis tab.

import pandas as pd
from astropy.time import Time
from datetime import datetime
import numpy as np
pd.set_option('display.max_columns', None)

year = [2014, 2015]
df_OR_year = df_OR[df_OR['FIRE_YEAR'].isin(year)]
pd.set_option('display.max_columns', None)

# Filter out non-finite values
df_OR_year = df_OR_year[np.isfinite(df_OR_year['DISCOVERY_DATE'])]

# Convert Julian dates to standard dates
df_OR_year['DISCOVERY_DATE'] = Time(df_OR_year['DISCOVERY_DATE'], format='jd').to_value('iso', subfmt='date')
df_OR_year.head(20)

##################################################

# Drop rows with NA values in the 'DISCOVERY_TIME' column
df_OR_year = df_OR_year.dropna(subset=['DISCOVERY_TIME'])

# Check the datatype of 'DISCOVERY_TIME'
if pd.api.types.is_datetime64_any_dtype(df_OR_year['DISCOVERY_TIME']):
    # Round off to the nearest hour
    df_OR_year['DISCOVERY_TIME'] = df_OR_year['DISCOVERY_TIME'].dt.round('H').dt.strftime('%H:%M:%S')
else:
    # Convert to datetime object
    df_OR_year['DISCOVERY_TIME'] = df_OR_year['DISCOVERY_TIME'].apply(lambda x: datetime.strptime(x, '%H%M'))
    # Round off to the nearest hour
    df_OR_year['DISCOVERY_TIME'] = df_OR_year['DISCOVERY_TIME'].dt.round('H').dt.strftime('%H:%M:%S')

    
################################################   

# Drop rows with NaN values in 'DISCOVERY_DATE' or 'DISCOVERY_TIME'
df_OR_year = df_OR_year.dropna(subset=['DISCOVERY_DATE', 'DISCOVERY_TIME'])

# Combine date and time into a new column
df_OR_year['DISCOVERY_DATETIME'] = pd.to_datetime(df_OR_year['DISCOVERY_DATE'] + ' ' + df_OR_year['DISCOVERY_TIME'], format='%Y-%m-%d %H:%M:%S')

###############################################

df_OR_year['DISCOVERY_DATETIME'] = df_OR_year['DISCOVERY_DATETIME'].dt.floor('H')

df_OR_year.head()
OBJECTID FOD_ID FPA_ID SOURCE_SYSTEM_TYPE SOURCE_SYSTEM SOURCE_REPORTING_UNIT SOURCE_REPORTING_UNIT_NAME LOCAL_FIRE_REPORT_ID LOCAL_INCIDENT_ID FIRE_CODE ICS_209_INCIDENT_NUMBER ICS_209_NAME MTBS_ID MTBS_FIRE_NAME FIRE_YEAR DISCOVERY_DATE DISCOVERY_DOY DISCOVERY_TIME STAT_CAUSE_CODE STAT_CAUSE_DESCR CONT_DATE CONT_DOY CONT_TIME FIRE_SIZE FIRE_SIZE_CLASS LATITUDE LONGITUDE STATE COUNTY FIPS_CODE FIPS_NAME Shape MONTH DAY geometry DISCOVERY_DATETIME
1722247 1722248 300000318 FS-1527910 FED FS-FIRESTAT 0601 Deschutes National Forest 241 1-157 JGK4 None None None None 2014 2014-11-10 314 10:00:00 9.0 Miscellaneous 2456971.5 314.0 1049 0.1 A 43.787778 -121.116667 OR Deschutes 017 Deschutes b'\x00\x01\xad\x10\x00\x00\xb8\x0b{wwG^\xc0\xd... 11 10 POINT (-121.11667 43.78778) 2014-11-10 10:00:00
1722248 1722249 300000319 FS-1527911 FED FS-FIRESTAT 0601 Deschutes National Forest 242 1-158 JG4Y None None None None 2014 2014-12-15 349 11:00:00 9.0 Miscellaneous 2457006.5 349.0 1201 0.1 A 43.820000 -121.321667 OR Deschutes 017 Deschutes b'\x00\x01\xad\x10\x00\x00<\xf7\xcc/\x96T^\xc0... 12 15 POINT (-121.32167 43.82000) 2014-12-15 11:00:00
1722249 1722250 300000320 FS-1527906 FED FS-FIRESTAT 0601 Deschutes National Forest 136 1-086 EK2A None None None None 2014 2014-07-29 210 18:00:00 1.0 Lightning 2456867.5 210.0 1935 0.1 A 43.397778 -121.218611 OR Lake 037 Lake b'\x00\x01\xad\x10\x00\x00\x80\xffs\xb9\xfdM^\... 7 29 POINT (-121.21861 43.39778) 2014-07-29 18:00:00
1722250 1722251 300000321 FS-1527907 FED FS-FIRESTAT 0601 Deschutes National Forest 232 1-149 JFC6 None None None None 2014 2014-10-04 277 13:00:00 9.0 Miscellaneous 2456936.5 279.0 1749 17.0 C 43.652778 -120.971111 OR Deschutes 017 Deschutes b'\x00\x01\xad\x10\x00\x00\xdc\x8e6\xaf&>^\xc0... 10 4 POINT (-120.97111 43.65278) 2014-10-04 13:00:00
1722251 1722252 300000322 FS-1525770 FED FS-FIRESTAT 0601 Deschutes National Forest 007 1-007 H36G None None None None 2014 2014-07-05 186 11:00:00 4.0 Campfire 2456843.5 186.0 1322 0.1 A 43.485278 -121.121389 OR Lake 037 Lake b'\x00\x01\xad\x10\x00\x00t)\xe8\xd5\xc4G^\xc0... 7 5 POINT (-121.12139 43.48528) 2014-07-05 11:00:00 
import pandas as pd
bendWeather_up = pd.read_csv("BendWeather_up.csv")

bendWeather_up = bendWeather_up[bendWeather_up['interval60'].str.len() > 10]
bendWeather_up['interval60'] = pd.to_datetime(bendWeather_up['interval60'], 
                                              format='%Y-%m-%d %H:%M:%S'
                                             )
merged_df = pd.merge(df_OR_year, 
                     bendWeather_up, 
                     left_on='DISCOVERY_DATETIME', 
                     right_on='interval60', 
                     how='inner')