skip to main content


The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 11:00 PM ET on Thursday, May 23 until 2:00 AM ET on Friday, May 24 due to maintenance. We apologize for the inconvenience.

Search for: All records

Creators/Authors contains: "Williams, A. Park"

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Abstract. The annual area burned due to wildfires in the western United States (WUS) increased bymore than 300 % between 1984 and 2020. However, accounting for the nonlinear, spatially heterogeneous interactions between climate, vegetation, and human predictors driving the trends in fire frequency and sizes at different spatial scales remains a challenging problem for statistical fire models. Here we introduce a novel stochastic machine learning (SML) framework, SMLFire1.0, to model observed fire frequencies and sizes in 12 km × 12 km grid cells across the WUS. This framework is implemented using mixture density networks trained on a wide suite of input predictors. The modeled WUS fire frequency matches observations at both monthly (r=0.94) and annual (r=0.85) timescales, as do the monthly (r=0.90) and annual (r=0.88) area burned. Moreover, the modeled annual time series of both fire variables exhibit strong correlations (r≥0.6) with observations in 16 out of 18 ecoregions. Our ML model captures the interannual variability and the distinct multidecade increases in annual area burned for both forested and non-forested ecoregions. Evaluating predictor importance with Shapley additive explanations, we find that fire-month vapor pressure deficit (VPD) is the dominant driver of fire frequencies and sizes across the WUS, followed by 1000 h dead fuel moisture (FM1000), total monthly precipitation (Prec), mean daily maximum temperature (Tmax), and fraction of grassland cover in a grid cell. Our findings serve as a promising use case of ML techniques for wildfire prediction in particular and extreme event modeling more broadly. They also highlight the power of ML-driven parameterizations for potential implementation in fire modules of dynamic global vegetation models (DGVMs) and earth system models (ESMs). 
    more » « less
  2. Abstract

    On seasonal time scales, vapor pressure deficit (VPD) is a known predictor of burned area in the southwestern United States (“the Southwest”). VPD increases with atmospheric warming due to the exponential relationship between temperature and saturation vapor pressure. Another control on VPD is specific humidity, such that increases in specific humidity can counteract temperature-driven increases in VPD. Unexpectedly, despite the increased capacity of a warmer atmosphere to hold water vapor, near-surface specific humidity decreased from 1970 to 2019 in much of the Southwest, particularly in spring, summer, and fall. Here, we identify declining near-surface humidity from 1970 to 2019 in the southwestern United States with both reanalysis and in situ station data. Focusing on the interior Southwest in the months preceding the summer forest fire season, we explain the decline in terms of changes in atmospheric circulation and moisture fluxes between the surface and the atmosphere. We find that an early spring decline in precipitation in the interior region induced a decline in soil moisture and evapotranspiration, drying the lower troposphere in summer. This prior season precipitation decline is in turn related to a trend toward a Northern Hemisphere stationary wave pattern. Finally, using fixed humidity scenarios and the observed exponential relationship between VPD and burned forest area, we estimate that with no increase in temperature at all, the humidity decline alone would still lead to nearly one-quarter of the observed VPD-induced increase in burned area over 1984–2019.

    Significance Statement

    Burned forest area has increased significantly in the southwestern United States in recent decades, driven in part by an increase in atmospheric aridity [vapor pressure deficit (VPD)]. Increases in VPD can be caused by a combination of increasing temperature and decreasing specific humidity. As the atmosphere warms with climate change, its capacity to hold moisture increases. Despite this, there is a decrease in near-surface air humidity in the interior southwestern United States over 1970–2019, which during the summer is likely caused by a decline in early spring precipitation leading to limited soil moisture and evaporation in spring and summer. We estimate that this declining humidity alone, without an increase in temperature, would cause about one-quarter of the VPD-induced increase in burned forest area in this region over 1984–2019.

    more » « less
  3. Abstract Recent record-breaking wildfire seasons in California prompt an investigation into the climate patterns that typically precede anomalous summer burned forest area. Using burned-area data from the U.S. Forest Service’s Monitoring Trends in Burn Severity (MTBS) product and climate data from the fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ERA5) over 1984–2018, relationships between the interannual variability of antecedent climate anomalies and July California burned area are spatially and temporally characterized. Lag correlations show that antecedent high vapor pressure deficit (VPD), high temperatures, frequent extreme high temperature days, low precipitation, high subsidence, high geopotential height, low soil moisture, and low snowpack and snowmelt anomalies all correlate significantly with July California burned area as far back as the January before the fire season. Seasonal regression maps indicate that a global midlatitude atmospheric wave train in late winter is associated with anomalous July California burned area. July 2018, a year with especially high burned area, was to some extent consistent with the general patterns revealed by the regressions: low winter precipitation and high spring VPD preceded the extreme burned area. However, geopotential height anomaly patterns were distinct from those in the regressions. Extreme July heat likely contributed to the extent of the fires ignited that month, even though extreme July temperatures do not historically significantly correlate with July burned area. While the 2018 antecedent climate conditions were typical of a high-burned-area year, they were not extreme, demonstrating the likely limits of statistical prediction of extreme fire seasons and the need for individual case studies of extreme years. Significance Statement The purpose of this study is to identify the local and global climate patterns in the preceding seasons that influence how the burned summer forest area in California varies year-to-year. We find that a dry atmosphere, high temperatures, dry soils, less snowpack, low precipitation, subsiding air, and high pressure centered west of California all correlate significantly with large summer burned area as far back as the preceding January. These climate anomalies occur as part of a hemispheric scale pattern with weak connections to the tropical Pacific Ocean. We also describe the climate anomalies preceding the extreme and record-breaking burned-area year of 2018, and how these compared with the more general patterns found. These results give important insight into how well and how early it might be possible to predict the severity of an upcoming summer wildfire season in California. 
    more » « less
  4. Abstract

    Atmospheric rivers (AR) are critically important to water resources management along the US west coast, driving variability in both droughts and floods across the region. Inter‐annual variability of ARs is well documented in the instrumental record back to the mid‐twentieth century, but long‐term variations in the frequency and landfall location of ARs along the US west coast are poorly understood due to limited records. This limitation impedes the ability to contextualize emerging trends and projections of AR activity. Here we use station‐based records of daily precipitation and tree‐ring records to present novel, spatially explicit estimates of daily AR occurrences in the first half of the twentieth century and annual AR counts over the last 600 years. First, we use neural networks and daily precipitation across Western North America to classify the daily occurrence of AR landfalls in three regions along the US west coast during the cold season back to 1916 CE. Then, we reconstruct the annual frequency of AR landfalls in those same regions back to 1400 CE using a gridded, tree‐ring based reconstruction of the standardized precipitation index and a Poisson regression framework. The skillful reconstruction of daily and annual AR occurrences provides previously unavailable estimates of AR landfall variability and highlights new peaks in AR activity and modes of low‐frequency variability prior to the instrumental record. Our reconstructions suggest that the average latitude of AR landfall has varied considerably on multi‐decadal scales over the last 600 years, but without any discernible trends beyond this quasi‐oscillatory behavior.

    more » « less
  5. Abstract

    Inter-annual climate variability (hereafter climate variability) is increasing in many forested regions due to climate change. This variability could have larger near-term impacts on forests than decadal shifts in mean climate, but how forests will respond remains poorly resolved, particularly at broad scales. Individual trees, and even forest communities, often have traits and ecological strategies—the legacies of exposure to past variable conditions—that confer tolerance to subsequent climate variability. However, whether local legacies also shape global forest responses is unknown. Our objective was to assess how past and current climate variability influences global forest productivity. We hypothesized that forests exposed to large climate variability in the past would better tolerate current climate variability than forests for which past climate was relatively stable. We used historical (1950–1969) and contemporary (2000–2019) temperature, precipitation, and vapor pressure deficit (VPD) and the remotely sensed enhanced vegetation index (EVI) to quantify how historical and contemporary climate variability relate to patterns of contemporary forest productivity. Consistent with our hypothesis, forests exposed to large temperature variability in the past were more tolerant of contemporary temperature variability than forests where past temperatures were less variable. Forests were 19-fold times less sensitive to contemporary temperature variability where historical inter-annual temperature variability was 0.66 °C (two standard deviations) greater than the global average historical temperature variability. We also found that larger increases in temperature variability between the two study periods often eroded the tolerance conferred by the legacy effects of historical temperature variability. However, the hypothesis was not supported in the case of precipitation and VPD variability, potentially due to physiological tradeoffs inherent in how trees cope with dry conditions. We conclude that the sensitivity of forest productivity to imminent increases in temperature variability may be partially predictable based on the legacies of past conditions.

    more » « less