Remote sensing and statistical analysis of the effects of hurricane María on the forests of Puerto Rico
Introduction
Hurricanes are major natural disturbances of temperate and tropical forests in the coastal regions of North and Central America, and the Caribbean Islands (Boose et al., 1994; Everham and Brokaw, 1996; Mabry et al., 1998; McNab et al., 2004). The island of Puerto Rico has frequent encounters with hurricanes. Since 1700, Puerto Rico has experienced over 80 hurricanes (Boose et al., 1994). Among those hurricanes, hurricane Hugo (1989) and hurricane Georges (1998) stimulated significant research on the effects of hurricane disturbance on tropical forests (e.g., Walker, 1991a, Walker, 1991b; Brokaw and Grear, 1991; Uriarte et al., 2005). Depending on hurricane intensity and landfall duration, forest impacts vary greatly, including defoliation, small and large branches loss, and the snapping and uprooting of stems (Lugo, 2008).
At the local scale, wind disturbance is influenced by forest type, stand characteristics, and tree species (Boose et al., 2004: Negrón-Juárez et al., 2010), and related to species adaptability, stem density, and collateral effects (Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014). Different tree species vary in their vulnerability to hurricane disturbances (Zimmerman et al., 1994; Canham et al., 2010) and in their recovery pathways following wind disturbance (e.g., Walker, 1991a; Zimmerman et al., 1994; Uriarte et al., 2009; Canham et al., 2010). Tree mortality is mainly due to uprooting and broken stems. Older forest stands often experience greater structural loss and basal area loss than younger stands (Everham and Brokaw, 1996; Foster et al., 1999; Flynn et al., 2010).
At the landscape scale, forest change and structural loss from tropical cyclones is influenced by topography (Boose et al., 1994; Foster et al., 1999; Flynn et al., 2010). Previous studies have shown that forests growing at high elevations and on windward slopes and ridges are more susceptible to wind disturbance (Everham and Brokaw, 1996; Arriaga, 2000; Bellingham and Tanner, 2000; Boose et al., 2004). Valleys provide protection to forests from strong wind, which tends to result in lower levels of disturbance (Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014). However, valleys may also amplify disturbances, as the wind can be constricted and accelerated (Everham and Brokaw, 1996).
At the regional scale, significant correlations have been found between wind speed and forest structural loss (Chambers et al., 2007; Zeng et al., 2009; Negrón-Juárez et al., 2014a; Schwartz et al., 2017). Wind speed decreases as a hurricane moves inland, and it decreases at larger radii from the eyewall of the storm (Boose et al., 2004; Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014). Wind speed is further modified by topography (Philippopoulos and Deligiorgi, 2012). Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014 showed that wind speed explains 20% of the observed variation in disturbance intensity and could be used as a predictor to assess forest disturbance. Boose et al. (1994) developed a simple meteorological model, HURRECON, which combines information on the track, size, and intensity of a hurricane and the corresponding land cover to estimate wind speed. The estimated wind maps from the model supplement limited wind observations and provide high accuracy in reconstructing historical wind maps (see details in Feng et al., 2018).
Most previous studies on forest disturbance intensity were based on repeated field surveys and ground-based measurements. While these traditional approaches provide more detailed and precise local data, they are time-consuming and expensive for investigating the full extent and the magnitude of disturbance at larger scales. Due to the lack of landscape scale forest disturbance maps, most forest disturbance studies can only explain the local disturbance variance using several factors collected from ground data. Few studies have addressed the potential for a number of mapped predictor variable layers and associated statistical analyses on forest impact variability at the landscape scale (Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014). Currently, remote sensing and spatial analysis tools have rapidly developed and emerged as effective methods to investigate large-scale forest disturbance metrics after hurricanes, enabling greater insight into factors that influence spatial and temporal variation in forest disturbance and recovery processes.
Satellite remote sensing approaches can be employed to quantify the effects of forest disturbance from local to global scales and at different temporal resolutions (Frolking et al., 2009, Chambers et al., 2007, Zhu et al., 2012, Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014, Baumann et al., 2014). Landsat imagery with high spatial resolution (30 m) has been successfully applied to detect forest disturbances in a number of studies, and Moderate Resolution Imaging Spectroradiometer (MODIS) data with high temporal resolution (daily) enable time series analyses of hurricane disturbances (Chambers et al., 2007; Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014; Helmer et al., 2010; Kennedy et al., 2010). Landsat–MODIS data fusion has become a useful method, which combines moderate spatial and high temporal resolution, to quantify and explain some of forest disturbance patterns (Chambers et al., 2007; Hilker et al., 2009; Xin et al., 2013; Negrón-Juárez et al., 2014, Negrón-Juárez et al., 2014). However, few researchers have carried out landscape-scale studies on the factors which affect the spatial distribution of these disturbance patterns. Moreover, traditional ground-based field validations are time-consuming and costly. High spatial resolution data, such as Panchromatic IKONOS with 1-m resolution and QuickBird satellite data with 0.7-m resolution, has been used to quantify local tropical forest mortality (Frolking et al., 2009; Clark et al., 2004). In this study, we used high spatial resolution airborne images to assist in validating larger-scale forest disturbance patterns.
Hurricane María made landfall in Puerto Rico on September 20, 2017, as Category 4 hurricane on the Saffir–Simpson scale (Pasch et al., 2018), producing unprecedented forest disturbance in Puerto Rico (Uriarte et al., 2019). In this study, we use satellite imagery to quantify the effects of hurricane María on forests in Puerto Rico, and analyze a large number of landscape-scale factors that affect the spatial variance in forest disturbance. The use of Google Earth Engine (GEE) (Gorelick et al., 2017) allowed us to combine data from different sources, including observations from satellite and aerial imaging systems, topographic, land cover, and GIS datasets, for a comprehensive understanding on the effects of Hurricane María on Puerto Rico forests. The objectives of this study were to:
- 1)
Develop a GEE remote sensing data analysis tool for rapidly quantifying spatial variability in forest disturbance following a hurricane landfall;
- 2)
Study the landscape factors that affect the patterns and severity of forest disturbance intensity;
We also developed a number of user interfaces tools to share our results, which allow a larger community with access to maps and analysis tools for the hurricane affected area.
Section snippets
Study area
The island of Puerto Rico is located in the Caribbean (centered at 18.2°N, 66.4°W). Forests cover about 60% of Puerto Rico, and all the forests fall within the subtropical belt of Holdridge Life Zone System (Holdridge, 1967; Harris et al., 2012). The forest types vary widely, including drought deciduous forests, semi-deciduous forests, seasonal evergreen forests, and evergreen forests. The wettest forest types are found at higher elevations, while the driest forest types are in southern and
Hurricanes
In the past 50 years, two major hurricanes hit Puerto Rico, reaching up to 175 km h−1 maximum wind speed. Past hurricanes and minor storms affected ecosystems, human populations, and infrastructure (see summary of effects in Tanner et al., 1991). Following Hurricane Irma, Hurricane María made landfall near Yabucoa Harbor, Puerto Rico with the maximum sustained wind speed around 250 km h−1. This research focused on the combined effects of these two hurricanes in extreme active 2017 hurricane
Multi-spectral remote sensing data
Landsat 8 Operational Land Imager and Thermal Infrared Sensor (OLI/TIRS) surface reflectance data were used in this study. To represent the pre-disturbance period images from June 1, 2016 to September 30, 2016 were used, and the for the post-disturbance period images from October 1, 2017 to January 30, 2018. Only images with less than 40% cloud cover were used. The two slightly different time periods were chosen to generate cloud-free pre and post hurricane images that cover most of the island
Effects of María on Puerto Rico
An analysis of the MODIS-EVI (Fig. 2) quantified the temporal effects of hurricane María on the forests of Puerto Rico. Comparing to mean EVI since 2000 (black lines in Fig. 2), EVI for Puerto Rico during the hurricane season of 2017 showed a sharp decline in vegetation greenness (brown line in Fig. 2). A rapid decline on September 13 and a much steeper decline in the subsequent interval up to September 29 in the EVI of 2017 (green line in Fig. 2) shows the effect of hurricane Irma and
Tree mortality and recovery
Previous studies have shown that an increase in the Landsat-derived ΔNPV of the same area after tropical cyclones is positively and linearly correlated with field-measured tree mortality and extensive crown effects (Chambers et al., 2007; Negrón-Juárez et al., 2014a). This strong correlation between the shift in the spectral reflectance and tree mortality and canopy disturbance has enabled quantitative measurements of tree mortality using remote sensing images. Chambers et al. (2007) developed
Conclusion
This research showed a remote sensing and statistical analysis of the hurricane disturbance in Puerto Rico, and its association with the landform characteristics and forest structures. It shows the power of remote sensing for large-scale post-hurricane analysis and provides insights impossible to achieve from isolated ground efforts. Large portions of the disturbed forests were clustered in Luquillo Mountains, the lower mid-west forested area of the main island, and Piñones Mangroves in
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This research was supported as part of the Next Generation Ecosystem Experiments-Tropics (NGEE), funded by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research under contract number DE-AC02-05CH11231. We give special thanks to Ariel E. Lugo from International Institute of Tropical Forestry, who provides helpful comments and suggestions to improve and clarify our manuscript. We also thank the G-LiHT team for collecting aerial photos for 2017/2018
References (80)
- et al.
Classification of multispectral images based on fractions of endmembers: application to land-cover change in the Brazilian Amazon
Remote Sens. Environ.
(1995) - et al.
Landsat remote sensing of forest windfall disturbance
Remote Sens. Environ.
(2014) - et al.
Google earth engine: planetary-scale geospatial analysis for everyone
Remote Sens. Environ.
(2017) - et al.
Mapping tropical dry forest height, foliage height profiles and disturbance type and age with a time series of cloud-cleared Landsat and ALI image mosaics to characterize avian habitat
Remote Sensing of Environment
(2010) - et al.
Generation of dense time series synthetic Landsat data through data blending with MODIS using a spatial and temporal adaptive reflectance fusion model
Remote Sensing of Environment
(2009) - et al.
Overview of the radiometric and biophysical performance of the MODIS vegetation indices
Remote Sens. Environ.
(2002) - et al.
Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr—temporal segmentation algorithms
Remote Sens. Environ.
(2010) - et al.
Landscape distribution and characteristics of large hurricane-related canopy gaps in a southern Appalachian watershed
For. Ecol. Manag.
(2004) - et al.
Multi-scale sensitivity of Landsat and MODIS to forest disturbance associated with tropical cyclones
Remote Sens. Environ.
(2014) - et al.
Detection of subpixel treefall gaps with Landsat imagery in Central Amazon forests
Remote Sens. Environ.
(2011)
Relationships between common forest metrics and realized impacts of hurricane Katrina on forest resources in Mississippi
For. Ecol. Manag.
Application of artificial neural networks for the spatial estimation of wind speed in a coastal region with complex topography
Renew. Energy
Green vegetation, nonphotosynthetic vegetation, and soils in AVIRIS data
Remote Sens. Environ.
Disturbance and coastal forests: a strategic approach to forest management in hurricane impact zones
For. Ecol. Manag.
Regional scale variation in forest structure and biomass in the Yucatan peninsula, Mexico: effects of forest disturbance
For. Ecol. Manag.
Toward near real-time monitoring of forest disturbance by fusion of MODIS and Landsat data
Remote Sens. Environ.
Continuous monitoring of forest disturbance using all available Landsat imagery
Remote Sens. Environ.
Remote Sensing of Landscapes with Spectral Images: A Physical Modeling Approach
Types and causes of tree mortality in a tropical montane cloud forest of Tamaulipas, Mexico
J. Trop. Ecol.
Hurricane Georges and vegetation change in Puerto Rico using AVHRR satellite data
Int. J. Remote Sens.
The influence of topography on tree growth, mortality, and recruitment in a tropical montane forest
Biotropica
Hurricane impacts to tropical and temperate forest landscapes
Ecol. Monogr.
Landscape and regional impacts of hurricanes in Puerto Rico
Ecol. Monogr.
Forest structure before and after Hurricane Hugo at three elevations in the Luquillo Mountains, Puerto Rico
Biotropica
Variation in susceptibility to hurricane damage as a function of storm intensity in Puerto Rican tree species
Biotropica
Hurricane Katrina’s carbon footprint on US Gulf Coast forests
Science
Quantifying mortality of tropical rain forest trees using high-spatial-resolution satellite data
Ecol. Lett.
Automatic boosted flood mapping from satellite data
Int. J. Remote Sens.
NASA Goddard’s Lidar, Hyperspectral and Thermal (G-LiHT) airborne imager
Remote Sens.
Forest damage and recovery from catastrophic wind
Bot. Rev.
Rapid remote sensing assessment of impacts from hurricane Maria on forests of Puerto Rico
PeerJ Preprints
Hurricane disturbance alters secondary forest recovery in Puerto Rico
Biotropica
Human or natural disturbance: landscape-scale dynamics of the tropical forests of Puerto Rico
Ecol. Appl.
Generalized collinearity diagnostics
JASA
Forest disturbance and recovery: a general review in the context of spaceborne remote sensing of impacts on aboveground biomass and canopy structure
J. Geophys. Res. Biogeosci.
Reconstructing hurricane disturbance in a tropical montane forest landscape in the cordillera central, Dominican Republic: implications for vegetation patterns and dynamics
Arct. Antarct. Alp. Res.
High-resolution global maps of 21st-century forest cover change
Science
Luquillo Experimental Forest: Research History and Opportunities. Experimental Forest and Range EFR-1
The landscape ecology of tropical secondary forest in montane Costa Rica
Ecosystems
Mapping the forest type and land cover of Puerto Rico, a component of the Caribbean biodiversity hotspot
Caribb. J. Sci.
Cited by (40)
Contribution of environmental factors to post-typhoon litterfall stability in subtropical montane cloud forests of the Asia-Pacific region
2024, Forest Ecology and ManagementCombining contemporary and pre-remote-sensing disturbance events to construct wind disturbance regime in a large forest landscape
2024, Forest Ecology and ManagementGoogle Earth Engine for archaeologists: An updated look at the progress and promise of remotely sensed big data
2023, Journal of Archaeological Science: ReportsPost-typhoon forest damage estimation using multiple vegetation indices and machine learning models
2022, Weather and Climate Extremes