Fires in croplands, plantations, and rangelands contribute significantly to fire emissions in the United States, yet are often overshadowed by wildland fires in efforts to develop inventories or estimate responses to climate change. Here we quantified decadal trends, interannual variability, and seasonality of Terra Moderate Resolution Imaging Spectroradiometer (MODIS) observations of active fires (thermal anomalies) as a function of management type in the contiguous U.S. during 2001–2010. We used the Monitoring Trends in Burn Severity database to identify active fires within the perimeter of large wildland fires and land cover maps to identify active fires in croplands. A third class of fires defined as prescribed/other included all residual satellite active fire detections. Large wildland fires were the most variable of all three fire types and had no significant annual trend in the contiguous U.S. during 2001–2010. Active fires in croplands, in contrast, increased at a rate of 3.4% per year. Cropland and prescribed/other fire types combined were responsible for 77% of the total active fire detections within the U.S and were most abundant in the south and southeast. In the west, cropland active fires decreased at a rate of 5.9% per year, likely in response to intensive air quality policies. Potential evaporation was a dominant regulator of the interannual variability of large wildland fires, but had a weaker influence on the other two fire types. Our analysis suggests it may be possible to modify landscape fire emissions within the U.S. by influencing the way fires are used in managed ecosystems.Key Points class="enumerated" style="list-style-type:decimal">Wildland, cropland, and prescribed fires had different trends and patternsSensitivity to climate varied with fire typeIntensity of air quality regulation influenced cropland burning trends class="head no_bottom_margin" id="__sec2title">1. IntroductionFor over a century, considerable effort has been invested in systematically monitoring and managing fires in the United States, with the aim of minimizing threats to human health and property and maintaining ecosystem function and biodiversity [; ]. Although fires are now widely used as a tool in land management, they also had an important role in regulating ecosystem processes prior to extensive human settlement [; ]. Climate influences fire on multiple time scales, including by determining distributions of plant functional types and species over a period of decades to centuries [], the amount and characteristics of lightning and other ignition sources [], fuel loads and moisture [], the length of the fire season, and fire weather during individual events []. Fires, in turn, modify climate through aerosol and greenhouse gas emissions [; ; ] and by changing land surface properties [; ; ; ]. These two way climate-fire interactions create the potential for regional and global scale feedbacks, with their magnitude and sign likely varying regionally [e.g., ; href="#b63" rid="b63" class=" bibr popnode">Tosca et al., 2013]. With fires playing an important role in modifying many Earth system and ecosystem processes, an important challenge is to understand the role of management and climate in controlling contemporary changes in fire activity.Multiple fire regimes exist within the U.S. as a consequence of considerable regional variability in ecosystems, climate, and land use. Large wildfires in western U.S. forests and shrublands are dominant contributors to regional and contiguous U.S. (CONUS) burned area [href="#b34" rid="b34" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627156">Littell et al., 2009; href="#b29" rid="b29" class=" bibr popnode">Kasischke et al., 2011]. For this class of fires, interannual and decadal-scale variability is often driven by fire weather controls during summer and by cumulative winter precipitation levels over several years that influence the continuity of surface fuels [href="#b61" rid="b61" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627167">Swetnam and Betancourt, 1990; href="#b69" rid="b69" class=" bibr popnode">Westerling et al., 2003; href="#b14" rid="b14" class=" bibr popnode">Crimmins, 2006]. The use of prescribed fire in forests as a federal policy started in the 1960s, when studies showed that landscape-level changes in ecosystem composition could be attributed to fire suppression [href="#b60" rid="b60" class=" bibr popnode">Stephens and Ruth, 2005]. In cropland areas, fire is frequently applied to clear fields of crop residues and to manage pests and disease [e.g., href="#b58" rid="b58" class=" bibr popnode">Smiley et al., 1996]. As a consequence, cropland fires vary considerably among different crop types [href="#b39" rid="b39" class=" bibr popnode">McCarty et al., 2007, href="#b40" rid="b40" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627164">2009; href="#b64" rid="b64" class=" bibr popnode">Tulbure et al., 2010; href="#b33" rid="b33" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627159">Lin et al., 2012]. Prairies of the Great Plains are burned every 1–2 years to avoid woody encroachment and enhance grazing productivity [href="#b8" rid="b8" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627158">Brockway et al., 2002; href="#b57" rid="b57" class=" bibr popnode">Simmons et al., 2007; href="#b2" rid="b2" class=" bibr popnode">Allen and Palmer, 2011]. Fires are also used for ecosystem management in pine forests in the southeast; prescribed fires, usually at 2–5 year intervals, are used to prepare the site before seeding and planting, to remove logging debris, and to manage understory species [href="#b68" rid="b68" class=" bibr popnode">Waldrop et al., 1992; href="#b10" rid="b10" class=" bibr popnode">Carter and Foster, 2004].Recent changes in climate have contributed to increases in the number of large wildfires in North America, as a consequence of longer fire seasons and warmer, drier conditions during summer [href="#b25" rid="b25" class=" bibr popnode">Gillett et al., 2004; href="#b28" rid="b28" class=" bibr popnode">Kasischke and Turetsky, 2006; href="#b70" rid="b70" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627162">Westerling et al., 2006; href="#b44" rid="b44" class=" bibr popnode">Morton et al., 2013]. With expected changes in climate over the next several decades, burned area for wildfires will likely increase [href="#b59" rid="b59" class=" bibr popnode">Spracklen et al., 2009], with individual fires becoming more intense and severe [href="#b71" rid="b71" class=" bibr popnode">Westerling et al., 2011]. However, much less is known about the climate sensitivity of cropland, rangeland, and plantation fires, or the underlying mechanisms regulating these sensitivities. As a result, it is difficult to estimate how the contributions of managed fires to regional and continental-scale emissions will evolve in the future. For example, while there is evidence showing that drought events decrease the occurrence of cropland fires in Australia [href="#b33" rid="b33" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627166">Lin et al., 2012], it remains unclear if this is caused by drought-induced reductions in crop yields and thus fuel loads, or from farmers igniting fewer fires during warmer and drier periods. This is an example of a critical gap in our understanding, and one that can be partly addressed by combining analysis of satellite imagery with other spatially distributed agricultural data sets.Satellite observations of burned area and active fires (thermal anomalies) provide consistent and systematic coverage, which enables assessment of changes in cropland and prescribed fires at a continental scale [href="#b30" rid="b30" class=" bibr popnode">Korontzi et al., 2006; href="#b40" rid="b40" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627155">McCarty et al., 2009]. Satellite data are suitable for evaluating fire impacts on air quality or surface characteristics [href="#b11" rid="b11" class=" bibr popnode">Chu et al., 2003; href="#b18" rid="b18" class=" bibr popnode">Engel-Cox et al., 2004; href="#b35" rid="b35" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627165">Liu et al., 2005; href="#b46" rid="b46" class=" bibr popnode">Park et al., 2007; href="#b5" rid="b5" class=" bibr popnode">Beck et al., 2011] and complement existing state and federal reports that typically document fire statistics using the number of occurrences or area burned within a given region. While existing state and federal reporting systems provide valuable information for many wildland fires and some prescribed fires, cropland burning has not yet been specifically targeted, and thus emissions and trends for this fire type remain highly uncertain. Given this limitation, satellite observations that provide comprehensive coverage of the U.S. have the potential to improve our understanding of the relative importance of different fire types and their relationship with environmental drivers. However, an understanding of the characteristics of each satellite sensor, including spectrometer sensitivities and orbit characteristics, is needed to properly interpret these observations.Two basic approaches exist for sampling fire patterns and trends with remote sensing data. The first is to map burned area using surface reflectance imagery from pre- and post-burn periods [href="#b62" rid="b62" class=" bibr popnode">Tansey et al., 2004; href="#b24" rid="b24" class=" bibr popnode">Giglio et al., 2009]. The second is to quantify actively burning fire fronts using measurements of surface thermal anomalies [href="#b21" rid="b21" class=" bibr popnode">Giglio et al., 2003a; href="#b4" rid="b4" class=" bibr popnode">Arino et al., 2005; href="#b50" rid="b50" class=" bibr popnode">Roberts et al., 2011]. Both remote sensing data streams have strengths and weaknesses with respect to their use in quantifying fire trends at a continental scale. Burned area more immediately lends itself to computing emissions, if it can be combined with information on fuel loads and combustion completeness [href="#b56" rid="b56" class=" bibr popnode">Seiler and Crutzen, 1980]. However, most global-scale remote sensing products of burned area are derived from surface reflectance imagery with moderate resolution (∼500 m to 1 km) [e.g., href="#b53" rid="b53" class=" bibr popnode">Roy et al., 2005]. This resolution is suitable for mapping large wildfires in savannas and boreal forests but can be insufficient for tracking fires that are much smaller than the spatial resolution of an individual surface reflectance pixel—as is often the case for cropland or plantation fires and many prescribed fires [href="#b41" rid="b41" class=" bibr popnode tag_hotlink tag_tooltip" id="__tag_463627163">McCarty, 2011; href="#b48" rid="b48" class=" bibr popnode">Randerson et al., 2012]. Surface reflectance-based products with a higher spatial resolution, such as Landsat with a 30 m pixel size, have the advantage of capturing pre- and post-fire surface reflectance changes suitable for landscape level analysis [e.g., href="#b19" rid="b19" class=" bibr popnode">Epting and Verbyla, 2005; href="#b17" rid="b17" class=" bibr popnode">Eidenshink et al., 2007]. However, the temporal resolution of Landsat with a 16 day repeat cycle may not be suitable for capturing many cropland or plantation fires that last for a short period and are often immediately followed by other forms of land management such as plowing or harvesting. Here we chose to use the MODIS active fire (thermal anomaly) product for our analysis at a continental level, because this product can detect fire fronts that are an order of magnitude smaller than moderate resolution burned area products [href="#b22" rid="b22" class=" bibr popnode">Giglio et al., 2003b]. The higher resolution of the active fire detections is important for systematically quantifying long-term trends in agricultural and prescribed fires and for quantitatively comparing activity levels across different fire types.In this paper, our goal was to quantify trends in satellite-derived time series of active fires as a function of fire type. The satellite observations we used from MODIS provide a statistical sampling (“daily snapshot”) of the distribution of fire thermal anomalies across the landscape. We divided these MODIS observations for the contiguous U.S. into three classes according to fire type based on a combination of remote sensing burned area and land cover products. The three fire types were large wildland fires, cropland fires, and prescribed/other fires. Although past work provides evidence for strong climate control on wildfires, less is known about relationships with climate for cropland and prescribed/other fires. We hypothesized that climate control of the latter two fire types was weaker than for large wildland fires, as a result of land managers regulating patterns of ignition. Our analysis identified differences in the response of several fire types to climate, thus providing information that is needed for the design of efficient fire management policies that account for projected climate changes during the next few decades. Finally, for cropland fires, we analyzed how long-term trends may respond to differences in the intensity of air quality policies enacted by different states. We hypothesized that states with stronger air quality regulation would have smaller increases (or greater reductions) in cropland fires over the past decade compared to states with fewer controls.
展开▼