Inducing Factors and Impacts of the October 2017 California Wildfires

The California wildfires of October 2017 were one of the largest wildfires in the state's history. Using surface temperature, surface pressure, cloud liquid and ice water contents, precipitation data, and wind data, we explore possible reasons for the wildfires. It is found that the mean surface temperature in California has increased, while mean cloud water contents and mean precipitation in California has decreased over the past 39 years. Higher temperatures, higher surface pressures, lower cloud water contents, lower precipitation, enhanced surface Santa Ana winds, and enhanced sinking air have set up favorable meteorological conditions for stronger wildfires in California, such as the October 2017 wildfires. Furthermore, the CO2 data from the Orbiting Carbon Observatory 2 satellite have, for the first time, made it possible for us to quantitatively characterize the impact of wildfires on atmospheric CO2 in California, which revealed that atmospheric CO2 increased by 2 ppm after the October 2017 California wildfires. Analyses in this study can help us better understand the causes and impacts of wildfires.


Introduction
As a response to rising greenhouse gases, air temperature increases over the global domain (Intergovernment Panel on Climate Change, 2013), which can further increase the abundance of water vapor in the atmosphere following the Clausius-Clapeyron relationship (Kao et al., 2018;Li et al., 2011;Santer et al., 2007;Trenberth et al., 2005). Unlike the impact on atmospheric moisture, the influence of global warming on precipitation is more complex. It is found that precipitation increases in wet areas and decreases in dry areas, which is called "wet-get-wetter and dry-get-drier" mechanism (e.g., Allan & Soden, 2007;Chou & Neelin, 2004;Kao et al., 2017;Kao et al., 2018;Li et al., 2011;Polson et al., 2013;Su et al., 2017;Trammell et al., 2015;Wang et al., 2016). The "dry-get-drier" mechanism affects the droughts in the southwestern United States. The air is dry in summer over the southwestern United States, for the Pacific high pressure moves moist air away from this region. In recent years, the air has become dryer and more severe droughts have happened over the southwestern States (Crockett & Westerling, 2018;Griffin & Anchukaitis, 2014). It is also found that droughts in the western United States cover a greater area in recent years than earlier years (Crockett & Westerling, 2018).
In principle, severe droughts could favor wildfires (e.g., Balling et al., 1992;Pausas and Fernandez-Munoz, 2012). In this study, we will explore possible meteorological inducing factors for recent wildfires in California. The October 2017 California wildfires were one of largest wildfires in the state's history. It included~9,000 wildfires, 1.2 million burned acres, and more than 10,000 destroyed structures. We will use the 2017 wildfires as an example to examine the factors favoring large wildfires in California.
The relationship between the environment and wildfires is interactive. Wildfires affect the environment in many aspects, for example, air quality (Cai et al., 2016). Here we investigate the impact of such wildfires on the concentration of atmospheric CO 2 , a well-known greenhouse gas. The impacts of fires on atmospheric CO 2 have been investigated in previous studies (e.g., Guyon et al., 2005;Heymann et al., 2017;O'Shea et al., 2013). Guyon et al. (2005) explored CO 2 emission from Amazonian deforestation fires using aircraft and noticed that there were more CO 2 emitted from forest fires compared with surrounding areas. O' Shea et al. (2013) utilized airborne measurements over eastern Canada and found that these fires released 1,512 g/(kg dry matter) of CO 2 to the atmosphere. Recently, Heymann et al. (2017) utilized CO 2 retrievals from a satellite and found that fires released more CO 2 to the atmosphere over Indonesia compared with its surrounding areas. The global data sets of satellite CO 2 can help us understand the impact of fires on ©2019. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. CO 2 in different areas. Here we will utilize satellite CO 2 products to investigate the impact of wildfires on atmospheric CO 2 in California.

Data
Surface temperature, surface pressure, cloud liquid and ice water contents, precipitation, zonal wind, meridional wind, and vertical wind are used to explore the meteorological conditions for the October 2017 California wildfires. Monthly mean surface temperature data, surface pressure data, surface zonal wind data, and surface meridional wind data from European Weather Centre (ECMWF) Interim (Dee et al., 2011) are utilized in this paper. Spatial resolutions of the ECMWF-Interim surface temperature data, surface pressure data, and surface wind data are 0.75°× 0.75°(latitude × longitude). It covers from January 1979 to present. Data can be downloaded at https://apps.ecmwf.int/datasets/data/interim-fulldaily/levtype=sfc/ website. Monthly mean precipitation data from Global Precipitation Climatology Project (GPCP; Adler et al., 2012;Adler et al., 2018) are used to explore the variability of precipitation. Spatial resolutions of the GPCP version 2.3 precipitation data are 2.5°× 2.5°(latitude × longitude). It

10.1029/2019EA000661
Earth and Space Science covers from January 1979 to present. Data can be downloaded at https://www.esrl.noaa.gov/psd/data/gridded/data.gpcp website. Cloud liquid and ice water contents from ECMWF-Interim (Delanoe et al., 2011) are also used in this paper to explore cloud variability. Monthly mean 500-hPa vertical velocity data from ECMWF-Interim (Dee et al., 2011) are used to explore the vertical transport. The spatial resolutions for ECMWF-Interim cloud water contents and vertical velocity are 0.75°× 0.75°(latitude × longitude). Data are available from January 1979 to present.
Monthly mean CO 2 data from Orbiting Carbon Observatory 2 (OCO-2) are used to explore the impact of wildfires on atmospheric CO 2 . The OCO-2 satellite was launched in September 2014 Eldering et al., 2017). Two CO 2 absorption bands at 1.61 and 2.06 μm were used to estimate the column CO 2 dry-air mole fraction, X CO2 , using physical retrieval methods described in O'Dell et al. (2012) and Connor et al. (2008). The difference between OCO-2 X CO2 retrievals with the Total Carbon Column Observing Network measurements is less than 0.5 ppm with a standard deviation of 1.5 ppm (Wunch et al., 2017). The spatial resolutions of the OCO-2 CO 2 data are 1.29 km × 2.25 km. OCO-2 X CO2 retrievals can be downloaded at https://oco.jpl.nasa.gov/science/OCO2DataCenter/ website. We have regridded OCO-2 X CO2 data to 2°× 2°(latitude × longitude).
Moderate Resolution Imaging Spectrometer (MODIS) Climate Modeling Grid Burned Area products (Giglio et al., 2018) are used in this paper to explore burned areas in October 2017. MODIS burned areas are created using an updated burned area mapping algorithm from Giglio et al. (2016). The spatial resolutions of MODIS Climate Modeling Grid Burned Area product are 0.25°× 0.25°(latitude × longitude). Monthly mean MODIS burned area data are available from November 2000 to present. MODIS burned area data can be downloaded at ftp://fuoco.geog.umd.edu/MCD64CMQ/C6/ website.

Results
As a characteristic of the typical Mediterranean climate, there is usually more precipitation in the winter season and less precipitation in the summer season in California. In 2017, there was more precipitation in January to April, which triggered a massive growth of weeds/vegetation. These weeds/vegetation were dried out later on and became fuels for the wildfires in October 2017. There is a severe drought in June to September of 2017, which help produce favorable conditions for the wildfire in October 2017. To better explore possible reasons and favorable conditions for the California wildfire in October 2017, we examine the long-term trends of surface temperature, cloud liquid and ice water contents, and precipitation during the summer season (June-September) from 1979 to 2017. The whole summer season, which is just before the October 2017 wildfires, is chosen because we want to study the cumulative effect of the whole season on the formation of the October 2017 wildfires. The averaged ECMWF-Interim surface temperature (red line) in California in the summer season (June-September) is calculated in Figure 1a. The linear trend of the mean surface temperature is estimated by a multiple regression method (Bevington & Robinson, 2003;Li et al., 2011) and shown as red dashed line in Figure 1a. The linear trend of ECMWF-Interim surface temperature in California is about 0.043 ± 0.015 K/year. Over the past 39 years, the surface temperature of California has increased as a total of~1.7 K in response to increasing greenhouse gases. The uncertainty of the trend is estimated by the standard deviation and degrees of freedom of the data (Bevington & Robinson, 2003;Box et al., 2005;Li et al., 2011). Details for the trend are summarized in Table 1.
During the summer, an increase in surface temperature coupled with changes in cloud water contents and precipitation lead to more severe droughts. To better understand the problem, we investigate ECMWF-Interim cloud liquid and ice water contents in California in the summer season (June-September, JJAS) from 1979 to 2017 (Figure 1b). Cloud water path including both liquid and ice water contents represent the total abundance of water in clouds per unit area. The cloud water path in California exhibits a weak linear trend of −0.07 ± 0.09 g/m 2 /year in the past 39 years. Such a negative trends may explain less rain and therefore more droughts in recent years in California. To characterize the correlation between liquid/ice water contents and precipitation in California, we also explore the variation of precipitation in California in the summer season (JJAS) in California from 1979 to present, which is based on the datasets of GPCP version 2.3 precipitation. As shown in Figure 1c, the linear trend of GPCP summer precipitation in California is −0.12 ± 0.09 mm/year, which suggests that there is less rain in California in recent years.
To better explore the relationships among surface temperature, cloud water content, and precipitation, we compare the detrended time series of these variables in Figure 2. Linear trends have been removed from the raw data. Then the mean value for the raw data is added back to the detrended time series, so it is easier to compare with Figure 1. Time series of detrended surface temperature, detrended cloud water content, and detrended precipitation in summer season (JJAS) are shown in Figure 2. There is a negative correlation between detrended ECMWF-Interim summer surface temperature and detrended ECMWF-Interim summer cloud liquid and ice contents as shown in Figure 2a. The correlation coefficient between detrended ECMWF- Interim summer surface temperature and detrended ECMWF-Interim summer cloud water contents is −0.52 with a significance level of 0.1%. The significance level of the correlation coefficient is estimated using a Monte Carlo method (Jiang et al., 2004;Press et al., 1992). Time series of detrended ECMWF-Interim summer surface temperature and detrended GPCP summer precipitation are shown in Figure 2b. The correlation coefficient between detrended summer surface temperature and detrended summer precipitation is −0.52 (0.1%). The negative correlations between temperature, cloud water content, and precipitation are consistent with dry-get-dryer mechanism from the thermodynamical perspective. There is a positive correlation coefficient of 0.89 (0.1%) between detrended summer precipitation and detrended summer cloud water contents in California, suggesting that precipitation is closely related to the cloud liquid and ice water contents.
Since natural variability, such as El Niño and Pacific Decadal Oscillation (PDO), can influence precipitation as suggested by previous studies (e.g., Ashok et al., 2007;Marvel & Bonfils, 2013;Smith et al., 2006;Trammell et al., 2016), we also explore possible relationships among summer precipitation in California, El Niño, and PDO. Southern Oscillation Index (SOI) and PDO indices are used to represent the strengths of El Niño and PDO. Results between summer precipitation and SOI are shown in Figure S1 in the supporting information. As shown in Fig. S1a, summer precipitation has a negative trend of −0.12 ± 0.09 mm/year and SOI has a weak positive trend of 0.015 ± 0.034/year. Detrended summer precipitation and detrended SOI are shown in Fig. S1b. The correlation coefficient of detrended summer precipitation and detrended summer SOI is −0.04 (58.9%), which suggests that there is no clear relationship between summer precipitation and SOI. This might be explained by the fact that the summer season is the dry season for California and low precipitation can be influenced by different factors, such as cloud, surface temperature, and circulation. We also explore for a possible relationship between summer precipitation and PDO index, and do not notice any significant correlation between summer precipitation and PDO index.

Earth and Space Science
We investigate the spatial patterns of ECMWF-Interim surface temperature anomaly, GPCP precipitation anomaly, ECMWF-Interim cloud water content anomaly, and ECMWF-Interim 500-hPa vertical velocity anomaly at California in October 2017. Climatological values averaged from January 1979 to October 2017 have been removed from these variables. As shown in Figure 3a, there are positive temperature anomalies in California in October 2017, which agree with our previous finding about the increasing decadal trend of surface temperature in California. Meanwhile, there are negative precipitation anomalies in California in Octber 2017 (Figure 3b), especially in northern California. The negative precipitation anomalies are further related to the negative anomalies of cloud liquid and ice water path in California (Figure 3c). Convection and the related vertical motion also play important roles in the formation of precipitation. Hence, we also investigate the ECMWF-Interim 500-hPa vertical velocity, which is an index for the large-scale ascending/ descending motion in the atmosphere and tightly related to precipitation (Kao et al., 2018). The reanalysis data shows that there are negative vertical velocity anomalies in California, suggesting strong sinking air in California in October 2017. The strong sinking air enhances the atmospheric stability and does not favor the formation of precipitation, leading to less precipitation in California in October 2017.
Since Pacific high pressure and horizontal winds can influence the weather in California, we explore the spatial distributions of surface pressure and surface winds in California in Figure 4. Monthly mean surface pressure and surface winds are shown in Figure 4a. The high pressure at Pacific Ocean tends to move moist air away from California. In the meanwhile, the Santa Ana winds tend to blow dry inland air to California. Both the high pressure and Santa Ana winds contribute to the drought in California in October 2017. Surface pressure anomalies and surface wind anomalies are shown in Figure 4b. There are positive pressure anomalies over the northern part and negative pressure anomalies over the southern part, which steepen the pressure gradient force and induces stronger northerly and northeast winds. As shown in Figure 4b, there are anomalies in the northerly winds near the California coast, which blocks the moist air away from California. There are also strong northerly winds and northeast winds over the land, which bring dry air from inland to California and contribute to the drought in this region. There are about 15 destructive fire events in October 2017 with the burnt area exceeding 1,000 acres. Four fire events are among the twenty most destructive wildfires in California's history. These four fire events are Tubbs fire (Napa and Sonoma County), Nuns fire (Sonoma County), Atlas (Napa and Solano County), and Redwood Valley Complex (Mendocino County), which occurred in the northern California. As shown in Figure 4, there are anomalous northerly winds and northeast winds over northern California, which bring dry air from inland to these areas. There are also strong wild fires occurring in southern California, such as Canyon 2 fire (Orange County), Buffalo fire (San Diego County), and Wildomar fire (Riverside County). There are anomalous northerly winds in Southern California, which bring dry air to Southern California. The analyses of time series and spatial patterns both suggest that the positive surface temperature, negative precipitation, negative cloud water contents, enhanced sinking air, Pacific high pressure, and enhanced Santa Ana winds contribute to drought conditions in California, which further induces the October 2017 wildfires in California.
In addition to exploring the favorable conditions for the October 2017 wildfires in California, we also assess the impact of wildfires on atmospheric CO 2 . It is well known that wildfires release CO 2 into the atmosphere through the combustion processes of organic matter. The newly retrieved global CO 2 data sets from the satellite OCO-2  provide us a great opportunity to examine the increase in the atmospheric CO 2 from one of the largest wildfires in California's history. Monthly mean OCO-2 column CO 2 data in October 2017 is shown in Figure 5a, which shows that more than 2-ppm CO 2 is released to the atmosphere in California during the October 2017 wildfires compared with surrounding states (e.g., Nevada). MODIS burned areas in October 2017 are shown in Figure 5b. The major fire events in October 2017 can be identified in Figure 5b. The CO 2 increase due to the largest wildfires is comparable to the magnitude of the long-term CO 2 trend (~2 ppm/year) induced by human activities. Such a significant CO 2 source needs to be considered in the current studies of environment and model simulation. Previous study (Saha et al., 2017) also found that the fire can increase albedo and latent heat, which leads to less convective precipitation. As suggested in Saha et al. (2017), the brightening after fire as a result of drier soils and losing senescent vegetation is responsible for rainfall suppression. It will further lead to more severe droughts and wildfires in California in the future, which will release more CO 2 to the atmosphere.

Conclusion
The October 2017 California wildfires were one of the largest wildfires in the state's history. The averaged surface temperature in the summer season (JJAS) in California has increased by 1.7 K during the past 39 years  as suggested by the ECMWF-Interim Reanalysis data. Meanwhile, both summer precipitation and summer cloud liquid and ice water contents decrease with time in California. The decreasing precipitation contributes to more severe droughts in the region and further favors the wildfires in California.
Our investigation of the anomalies of surface temperature, surface pressure, precipitation, cloud water contents, surface horizontal winds, and 500-hPa vertical velocity in October 2017 suggests significant correlations and possible causality among high temperature, surface pressure, low precipitation, low cloud water contents, enhanced surface Santa Ana winds, and enhanced sinking air in California. All of these factors contribute to more severe droughts, which can contribute to stronger and more devastating wildfires.
We also quantify the release of CO 2 from the largest wildfires in California's history for the first time. As suggested by the OCO-2 satellite CO 2 data, there is~2 ppm more CO 2 released to the atmosphere by the October 2017 wildfires. Considering that CO 2 released from wildfires may keep increasing, we have to carry out more stringent control policy and curb more emissions of CO 2 from other anthropogenic sources (e.g., industrial and automobile; Newman et al., 2016) to mitigate global warming. As a positive feedback, effective control of anthropogenic CO 2 can help us prevent future wildfires in California and other areas.