Available online at www.sciencedirect.

com Current Opinion in

ScienceDirect Environmental Science & Health

Remote sensing techniques to assess post-fire vegetation

recovery
Fernando Pérez-Cabello1,2, Raquel Montorio1,2 and Daniel Borini Alves3

Abstract is the total or partial combustion of vegetation cover.

Wildfires substantially disrupt and reshape the structure, The intensity of the subsequent regrowth process
composition and functioning of ecosystems. Monitoring post- (often called recovery [2]) is heavily dependent on the
fire recovery dynamics is crucial for evaluating resilience and following: the effectiveness of the anatomical and
securing the relevant information that will enhance manage- physiological regeneration strategies of the affected
ment and support ecosystem restoration after fires. Compared species, the degree of alteration of the other elements of
to the extensive and labour-intensive field campaigns, remote the soilevegetation complex, and their interactions in
sensing provides a time- and cost-effective tool to monitor light of environmental factors and post-fire temporal
post-fire vegetation recovery (PVR). This concise literature conditions [3]. Monitoring post-fire vegetation recovery
review presents tools and recent advances in remote sensing (PVR) is crucial, as it provides valuable information for
techniques, focusing on the most commonly used sensors and analysing ecosystem resilience, for determining land-
indicators/metrics. It also provides recommendations on the scape dynamics, and for forest management purposes.
use of these tools for assessing vegetation recovery and on
existing gaps regarding technical limitations that could guide Compared to extensive and labour-intensive field cam-
future research. paigns, remote sensing (RS) techniques are a time- and
cost-effective way to monitor post-fire ecosystem re-
Addresses covery [4]. Numerous studies have affirmed the capac-
1
Department of Geography and Spatial Management, University of
ity of satellite imagery to quantify fire impacts over vast
Zaragoza, C/Pedro Cerbuna 12, 50009, Zaragoza, Spain
2
GEOFOREST-IUCA Research Group, Environmental Sciences zones and different ecosystems. In this sense, RS has
Institute (IUCA), University of Zaragoza, C/Pedro Cerbuna 12, 50009, been identified as an effective tool for understanding
Zaragoza, Spain how ecosystems respond to fire, which can provide an
3
Laboratory of Vegetation Ecology, Instituto de Biociências, Universi- enhanced understanding of vegetation recovery patterns
dade Estadual Paulista (UNESP), Avenida 24-A 1515, 13506-900, Rio
and make a positive contribution to sustainable forest
Claro, Brazil
management [5]. Spectral trajectories (considered a
Corresponding author: Pérez-Cabello, Fernando (fcabello@unizar.es) proxy for vegetation recovery) are usually used to anal-
yse the spatialetemporal dynamics of vegetation cover
following wildfires, using different indices, metrics, and
Current Opinion in Environmental Science & Health 2021,
temporal perspectives.
21:100251
This reviews comes from a themed issue on Environmental impact A concise and nonsystematic review of different publi-
assessment: Fire. Prevention, management and challenges
cations on the trends in RS applications for analysis of
Edited by Paulo Pereira and Edivaldo L Thomaz
PVR is presented in this article. The emphasis is on
For a complete overview see the Article Collection contextualizing the use of the different temporal per-
https://doi.org/10.1016/j.coesh.2021.100251 spectives applied, and the most commonly used sensors,
2468-5844/© 2021 The Authors. Published by Elsevier B.V. This is an indicators, and metrics. It is important to highlight that
open access ar ticle under the CC BY-NC-ND license (http:// there are a large number of reviews on change detection
creativecommons.org/licenses/by-nc-nd/4.0/). techniques and trend analysis using RS products. For
example, Gitas et al. [6] and Chu & Guo [5] provided
Keywords one of the first systematic reviews of post-fire moni-
Wildfire, Vegetation regrowth, Satellite imagery, Change detection, toring methods and techniques. Banskota et al. [7], Zhu
Time series, Spectral variables. [8], Tewkesbury et al. [9] and Hirschmugl et al. [10]
addressed a complete revision of the different algo-
rithms developed to analyse imagery time series,
Introduction including applications for monitoring areas affected by
Wildfires are an important modelling agent in terrestrial wildfires. Bartels et al. [11] undertook a quantitative
ecosystems, influencing the dynamics and interaction of review of the literature, to determine recovery times
soil and vegetation components [1]. In the immediate following wildfire, whereas Cohen et al. [12] conducted
aftermath, the most significant impact on the landscape a comparative analysis of the main algorithms used to

www.sciencedirect.com Current Opinion in Environmental Science & Health 2021, 21:100251

2 Environmental impact assessment: Fire. Prevention, management and challenges

map the full range of magnitudes of forest disturbance. (Light Detection and Ranging), are being deployed.
Martı́nez et al. [13] and Szpakowski and Jensen [14] are Unlike optical sensors, active sensors can retrieve in-
two of the most recent and up-to-date studies that deal formation below the tree canopies, allowing access to
with RS techniques for post-fire monitoring. All these variables associated with the vertical structure of the
reviews provide insightful contributions on the topics vegetation recovery after fire [20e22]. Although LIDAR
summarized in the present paper (sensors, indicators data are largely acquired via aircraft missions, especially
and metrics used in PVR). in certain areas, NASA’s Global Ecosystem Dynamics
Investigation (GEDI) programme has delivered signifi-
Sensors and RS products cant advances. GEDI data are derived from an LIDAR
For decades, PVR has been mainly analysed through sensor on board of a satellite platform, providing access
multispectral optical satellite imagery, both with coarse to data on biomass dynamics and diversity of canopy
spatial resolution instrumentsdAdvanced Very High structure [23,24] with outstanding potential for PVR
Resolution Radiometer, Syste`me pour l’observation de la monitoring. At the same time, the growing application of
Terre and Moderate Resolution Imaging Spectroradi- unmanned aerial vehicles (UAVs) for generating RS data
ometer (MODIS)das well as with tools that provide a sets provides more flexible access to multitemporal data
finer resolutiondThematic Mapper, Enhanced The- and allows for the deployment of sensors with ultra-high
matic Mapper Plus, Operational Land Imager and spatial resolution [25e27].
MultiSpectral Instrument (MSI). These sensors
generate images that go beyond simply capturing verti- Furthermore, a growing number of investigations
cal views of the Earth’s surface; they provide informa- combine optical and active RS data sets, both as com-
tion on spectral regions that capture vegetation plementary data and in an integrated way through fusion
conditions, such as near- and mid-infrared, and thereby processes. For example, Bolton et al. [22] fused Landsat
support detailed monitoring of post-fire dynamics. time series data and airborne (LIDAR) to assess changes
in forest structure; Meng et al. [28] linked multispectral
Among the finer spatial resolution RS data sets, the satellite imagery, airborne imaging spectroscopy and
Landsat program (NASA-USGS) is the main source of LIDAR, with the aim to quantify post-fire forest re-
time-series data for monitoring vegetation at regional covery rates by differentiating canopy recovery from
scales because of the following advantages: (1) good understory recovery. In the same way, Voleger et al. [29]
temporal, spectral and spatial characteristics (30 m explored the combination of LIDAR data and multi-
spatial resolution and 16 days of revisiting time); (2) temporal Landsat series (calibrated with field data), to
long time span coverage (from 1980s to today); and (3) produce maps of post-fire wildlife habitats.
the derived data sets can be easily accessed free of
charge [15e18]. Moreover, the RS data sets derived Indicators, metrics and algorithms
from the MSI sensors onboard the Sentinel-2 satellites Spectral variables and indicators
of the European Space Agency have grown in impor- From an RS perspective, the vegetation processes that
tance since the start of the mission in 2015. Their in- follow disturbances can be mainly analysed by reflec-
formation is also freely available and provides greater tance values and spectral indices [30]. In relation to
spatial and temporal resolution than Landsat (10e20 m PVR, these indicators generally rely on greenness mea-
and up to a 5-day revisiting cycle). More recently, the surements of redenear-infrared (ReNIR) vegetation
development of harmonized Landsat and Sentinel-2 indices [31,32], based on different algebraic combina-
products has received special attention in the RS com- tions between original spectral bands. These indices are
munity [19]. By providing more frequent acquisitions, used to determine (1) whether postdisturbance values
they enable the detailed monitoring that is necessary to correspond to the previously recorded state (i.e. post-
uncover the short-term, post-fire dynamics in ecosys- fire resilience) and (2) how long it costs to reach the
tems that recover quickly, like tropical savannas and previous state [33].
grasslands. Regarding the coarse spatial resolution
products, the MODIS series stands out. With an almost Among the NIR-based spectral indices derived from RS
daily temporal resolution (1e2 days, with a spatial res- imagery, the Normalized Difference Vegetation Index
olution of up to 250 m), it allows broader spatial scales, (NDVI), the Enhanced Vegetation Index (EVI) and the
which can explore in more detail the effects of season- Normalized Burn Ratio (NBR) are the ones used most
ality and land surface phenology (LSP) associated with frequently to monitor PVR. NDVI and EVI are very
the vegetation’s response to fire. sensitive to seasonal and biophysical variations of vege-
tation changes and are, thus, used where natural vari-
Concurrently with the recent development and avail- ability is important [34]. Although they tend to saturate
ability of collections of dense time series of optical over dense forests and are not effective in measuring
satellites, other sensors and technologies, such as forest structure or species composition, they still pro-
RADAR (Radio Detection and Ranging) and LIDAR vide a good proxy for vegetative regrowth [35]. NBR is

Current Opinion in Environmental Science & Health 2021, 21:100251 www.sciencedirect.com

 Post-fire vegetation recovery using remote sensing Pérez-Cabello et al. 3

usually used to evaluate burn severity levels; however, it metric integrates different spectral indi-
also delivers good results in assessing long-term vege- cesddifferenced Normalized Burn Ratio, differenced
tation regeneration [36] and is being reconceptualized Normalized Difference Vegetation Index, the Change
as an indicator for the scope of post-fire recovery (e.g. Vector, and the Relative Change Vector Maximumdto
burn recovery ratio) [33] (Table 1). Other indices used identify the magnitude of the spectral changes between
as variables in PVR are the Soil Adjusted Total Vegeta- pre- and postdisturbance events [42]. Similarly, Torres
tion Index [31], an effective way to capture temporal et al. [43] proposed the Cumulative Relative Recovery
changes in grassland vegetation; the Anthocyanin Index, a long-term recovery indicator using the product
Reflectance Index 2 and the Transformed Chlorophyll MOD13Q1 (MODIS) (Table 1); and White et al. [44]
Absorption Reflectance Index [37], very resistant to the readapted the Recovery Indicator of Kennedy et al. [45]
changes of Leaf Area Index and solar zenith angle; the (Table 1). Recently, Du et al. [46] proposed the Tri-
Tasseled Cap Transformationebased indices [38], sen- Temporal Logic-verified Change Vector Analysis, an
sitive for monitoring the canopy moisture and structure; unsupervised method for improving bitemporal
the Forest Recovery Index 2, that is, the reciprocal of methods, which introduces an additional image to form a
Integrated Forest Z-score, a threshold-based index mutual validation logic.
developed as a part of the Vegetation Change Tracker
algorithm [32], or variables derived from texture anal- There is an ongoing shift from bitemporal to continuous
ysis, especially useful in areas of heterogeneous vege- approaches [47], because of the improved access to
tation because they consider the spatial adjacency continuous RS series and the opportunity they provide
relationships of pixels [37]. to assess post-fire dynamics in detail [48]. Long satellite
time series can capture the complexity of vegetation
Variables derived from traditional classification, spectral regeneration processes in fire-affected areas, allowing
mixture analysis (SMA) techniques, Geographic object- the analysis of both short duration phenomena and the
based image analysis (GEOBIA) and active microwave smoothing of long-term trends with high consistency
RS are also important alternatives. SMA considers that [8]. Thus, the open access to satellite image archives,
each type of ground cover is represented by its mean especially MODIS or USGS Landsat, has led to the
spectral signature, deriving endmember proportions development of many techniques and applications to
using spectral unmixing procedures. In this sense, some describing vegetation regrowth patterns in fire-affected
outstanding examples in the generation of indicators are forest [8,18]. One of the most widely used tools is the
the shade normalized green vegetation fraction image Landsat-based detection of Trends in Disturbance and
obtained by applying Multiple Endmember Spectral Recovery (LandTrendr) [49], a trajectory segmentation
Mixture Analysis [39], or the Normalized Degradation method that applies a temporal and spatial normaliza-
Fraction Index for monitoring forest degradation [40], tion process for extracting spectral trajectories of land
both using Landsat time series. GEOBIA are techniques surface change. The technique consists of decomposing
that use both spectral response and contextual infor- the time series curve into a sequence of straight-line
mation to assess post-fire vegetation characteristics in segments [30]. In the case of burnt areas, three seg-
groups of pixels (geographic objects generated by image ments would be obtained: (1) a flat line before the fire
segmentation) [25]. Regarding active microwave RS, event, (2) a declining line following the disturbance and
advanced applications that stand out include the use of (3) a segment line with a positive slope throughout the
L-band HV-polarized SAR backscatter in the monitoring recovery. Recently, this method has been used to map
of post-fire changes (e.g. tree survival in eucalyptus snag hazard for fire responders in disturbed forests [50]
forests of Western Australia [20] or the use of airborne or to record the disturbance and recovery history in pine
LIDAR in the Boreal Shield West Ecozone of Canada forests [30,38,48,51].
[22]).
Other spectral trajectory methods are based on curves
and trajectory fitting (i.e. methods that assume a linear
Metrics and algorithms relationship between time and spectral bands or indices
Within the field of change detection methods, in which [8]). Torres et al. [43] proposed the Half Recovery Time
PVR studies are integrated, two types of temporal ap- (HRT), a post-fire recovery indicator, using nonlinear
proaches are usually considered: (1) bitemporal change model fitting of the post-fire NDVI anomalies to iden-
detection methods, a comparison between states at tify the number of days needed to reach a 50% level of
different moments (i.e. pre- and post-fire) and (2) recovery. Looking through a 5-year window, Frazier et al.
spectral trajectories of land surface change, in which [52] used different indicators based on predisturbance
recovery processes are considered as a continuous pro- NBR to detect trends in post-fire spectral recovery;
cess [13]. An example of the first approach is the Multi- Wang and Zhang [53] calculated LSP trends from
Index Integrated Change Analysis, which was applied in MODIS time series of about 1000 fires that occurred
the context of the Fire and Resource Management from 2002 to 2014 in the western USA. Vogelmann et al.
Planning Tools programme (LANDFIRE) [41]. This [54] proposed the Image Trends from Regression
www.sciencedirect.com Current Opinion in Environmental Science & Health 2021, 21:100251
Current Opinion in Environmental Science & Health 2021, 21:100251

4 Environmental impact assessment: Fire. Prevention, management and challenges

Table 1

Examples of spectral variables recently used for monitoring post-fire changes vegetation.

Index Source Applied or Adapted by Expression

GVShade
ðNPV +Soil Þ

NDFI Souza CM, Roberts DA, Bullock et al. (2020) [40] NDFI =
GVShade +ðNPV +Soil Þ
Cochrane MA: To detect tropical forest canopy damage and
Combining spectral and degradation processes NPV = nonphotosynthetic vegetation
spatial information GV = green vegetation
to map canopy damage from
GV
selective logging GVShade =
1
Shade
and forest fires. Remote Sensing
of Environment 2005,
98:329–343
rNIR
rRed
EVI Huete A, Didan K, Miura T, Vo & Kinoshita (2020) [35] EVI ¼
rNIR + C1 × rRed
C2 × rBlue + 1
Rodriguez EP, Gao X, To assess the effects of post-fire treatment
Ferreira LG: Overview of the (wood and straw mulch) on vegetation
radiometric and biophysical
performance of the MODIS
vegetation indices. Remote
Sensing of Environment 2002,
83:195–213
NIR
SWIR
NDMI Gao BC: NDWI - A normalized Hamunyela et al. (2020) [55] NDMI ¼
NIR + SWIR
difference water index for To determine vegetation water content on
remote sensing of vegetation regeneration monitoring of montane forests
liquid water from space. of Eastern Tanzania
Remote Sensing of Environment
1996, 58:257–266
DNBRregrowth
RI [44] White et al. (2017) [44] RI ¼
DNBRdisturbance
To measure spatial and temporal patterns in
post-disturbance vegetation recovery DNBRregrowth = NBRpostfire-
(harvest and wildfire) in Canada’s forested NBR year of the disturbance
ecosystems)
DNBRdisturbance = NBRprefire-
NBR at the end of disturb.
VRR Lin WT, Chou WC, Lin CY, Adagbasa et al. (2020) [36]
rNIR
rRed
Huang PH, Tsai JS: To validate vegetation response-ability NDVI ¼
rNIR + rRed
Vegetation recovery monitoring models on grassland, integrating environmental
NDVI2
NDVI1
and assessment factor and adaptive vegetation strategies VRR ¼
NDVI0 + NDVI1
at landslides caused by
earthquake in Central NDVI0 ¼ prefire
Taiwan. Forest Ecology NDVI1 ¼ disturbance
and Management 2005, NDVI2 ¼ postfire
www.sciencedirect.com

210:55–66
1 X N
|NDVIpost ;i
minNDVIfire |
CRRI [43] Torres et al. (2018) [43] CRRI ¼
N i ¼1 NDVIpre
An integrative indicator to measure long-term
recovery and to rank the main drivers in northern
Portugal using high-temporal resolution satellites
HRT [43] Torres et al. (2018) [43] Number of days necessary to reach
To measure short-term recovery velocity using the 50% level of recovery from the
high-temporal resolution satellites minimum NDVI value observed
www.sciencedirect.com

during the year of fire to the pre-fire

median.
RTI [43] Torres et al. (2018) [43] The slope of the trend in the
To capture temporal patterns of PVR process NDVI data for the post-fire
after the first phases of regrowth using period using the Theil–Sen
high-temporal resolution satellites estimator.
SATVI [31] Villarreal et al. (2016) [31] SATVI ¼
To characterize long-term recovery trajectories rSWIR5
rRed rSWIR7
of desert grassland ð1 + LÞ

rSWIR5 + rRed +L 2

   
1 1
ARI Gitelson AA, Merzlyak MN, Chivkunova OB: Optical Fernández-Guisuraga et al. (2019) [37] ARI ¼ B7

B3 B6
Properties and Nondestructive Estimation of Anthocyanin To assess quantitative variables of vegetation
Content in Plant Leaves. Photochemistry and recovery (density seedlings and woody
Photobiology 2001, 74:38–45 species cover) in fire-prone ecosystems using
fine-grained satellite imagery   
B6
TCARI Haboudane D, Miller JR, Pattey E, Zarco-Tejada PJ, Fernández-Guisuraga et al. (2019) [37] TCARI ¼ 3 ðB6
B5Þ
0:2ðB6
B3Þ
B5
Strachan IB: Hyperspectral vegetation indices and To assess variable quantifying vegetation
novel algorithms for predicting green LAI of crop canopies: recovery (density seedlings and woody B3 = green (510–580 nm)
Modeling and validation in the context of species cover) in fire-prone ecosystems B5 = red (630–690 nm)
precision agriculture. using fine-grained satellite imagery B6 = red edge (705–745 nm)
Remote Sensing of Environment 2002, 90:337–352
NBRta
NBRtd
BRR ¼

Post-fire vegetation recovery using remote sensing Pérez-Cabello et al.

BRR Lin WT, Lin CY, Chou, WC: Assessment of vegetation recovery Chompuchan and Lin (2017) [33]
NBRto + NBRtd
and soil erosion at landslides caused by a catastrophic earthquake: To evaluate the forest recovery considering the
a case study in Central Taiwan. Ecol. Eng. 2006, 28: 79–89. concept of resilience and the magnitude of fire to = prefire event
damage td = time when delay
Current Opinion in Environmental Science & Health 2021, 21:100251

mortality existed
ta = time of assessment
TCT-TCAPowell SL, Cohen WB, Healey SP, Kennedy RE, Moisen GG, Pierce KB, Viana-soto et al. (2020) [38]
TCG
Ohmann JL: Quantification of live aboveground forest biomass dynamics To characterize postfire TCA ¼ tan
1
TCB
with Landsat time-series and field inventory data: A comparison of empirical trajectories in
modeling approaches. Remote Sensing of Environment 2010, 114:1053–1068 Mediterranean pine forests TCB = Tasseled Cap: Brightness

TCG = Tasseled Cap: Greenness

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
FRI2 Huang, C.; Goward, S$N.; Masek, J.G.; Thomas, N.; Zhu, Z.; Morresi et al. (2019) [32] u
u 1 X N
ðbi + bi Þ
Vogelmann, J.E. An automated approach for reconstructing recent To track long-term forest regeneration and to IFZ ¼ t
NB i ¼ 1 SDi
forest disturbance history using dense Landsat time series stacks. monitor the development of tree canopy cover
1
Remote Sens. Environ. 2010, 114, 183–198 FRI2 ¼
ðIFZ + 1Þ
bi = spectral value of the pixel in
bandi,bi and SDi = mean and
standard deviation obtained from
forest samples
NB = number of spectral bands

ARI, Anthocyanin Reflectance Index; BRR, Burn Recovery Ratio; CRRI, Cumulative Relative Recovery Index; EVI, Enhanced Vegetation Index; FRI2, Forest Recovery Index 2; HRT, Half Recovery Time index;
NDFI, Normalized Difference Fraction Index; NDMI, Normalized difference moisture (water) index; RI, Recovery Indicator at short-term; RTI, Recovery Trend Index; SATVI, Soil Adjusted Total Vegetation Index;
TCARI, Transformed Chlorophyll Absorption Reflectance Index; TCT-TCA, Tasseled Cap Transformations –Tasseled Cap Angle; VRR, Vegetation Recovery Rate.

5
6 Environmental impact assessment: Fire. Prevention, management and challenges

Analysis (ITRA), which is based on NDVI and SWIR/ management strategies, especially in the context of
NIR index, to assess gradual changes; Chompuchan and global climate change. The most practical way to
Lin [33] identified time of recovery using a curve-fitting monitor changes over large areas and periods is through
of forest recovery trajectories to the exponential decay image-processing techniques based on change detection
function. Hamunyela et al. [55] used the STEF algo- or classification techniques [40]. These approaches are
rithm (Space-Time Extremes and Features) based on especially fitting because wildfires, and the following
space-time featuresdsuch as magnitude of change, vegetation recovery processes, substantially alter the
temporal linear trend in spatial variability (using NDWI land surface’s spectral signature. Metrics and tech-
as indicator)dto track forest disturbances and detected niques to track vegetation changes and trends following
forest gains (i.e. regeneration) in Tanzania. Cunha et al. fire using satellite time series provide information at
[56] applied the Time Series Segmentation and RE- different spatiotemporal and spectral resolutions. The
Sidual TREND method [57], implemented in the last years have seen a proliferation of research efforts,
Breaks For Additive Seasonal and Trend method [58], to published in various specialized journals; however,
differentiate structural change (breakpoint) and trends several key gaps concerning the use of RS for the anal-
happening over a longer period in seasonally tropical dry ysis of PVR remain. From the literature review presen-
forests. Furthermore, nonparametric tests such as ted in this paper, the following key considerations have
ManneKendall and the TheileSen slope estimator have emerged:
also been widely used. For example, Morresi et al. [32]
assessed the significance of the SVIs trends and calcu- 1) According to Pickell et al. [2], along with the
lated the rate of change and the direction of NDVI numerous studies in which post-fire recovery trends
trends. Torres et al. [43] proposed the Recovery Trend are analysed, it is necessary to highlight the impor-
Index, computed as the slope of the trend in the NDVI tance of properly interpreting the use of spectral
data for the post-fire period using the TheileSen indicators with the recovery in ecological terms,
estimator. taking into account data on the structure, composi-
tion and ecological functions of the colonizing plant
Alternatively, metrics based on digital classification communities. In this context, Bartels et al. [11] point
processes applied to time series data have also been out that there is a lack of clarity in the definition of
used. For example, Cardille and Fortin (2016) used the term ‘recovery’ and that a connection should be
Bayesian Updating of Land Cover, an algorithm set up between spectral indicators and ecological
designed to allow continuous updating of classifications understanding of forest recovery. For example, the
using image collections, applied for tracking a fast- classification of plant associations, combined with
growing forest fire from Landsat-8 images [59,60]; measures such as canopy cover, tree height or stand
Savage et al. [61] used Landsat imagery to predict basal area, are fundamental for understanding the
species composition of vegetation growing from a effects of fire on vegetation recovery and elaborating
disturbance ecology point of view, using the zero- fitting conservation strategies [66]. In this sense,
inflated regression to map percent canopy cover by according to Szpakowski & Jensen [14], owing to the
species and subcanopy species. However, despite the specific nature of the vegetation’s spectral response
large number of algorithms and changes detection and the different components of forest recovery,
techniques, Heiley et al. [62] demonstrated that an using a single method may not be the best option,
ensemble of change detection algorithms could be more requiring multiple methods to be deployed in each
effective and accurate than maps from any single auto- ecosystem. Moreover, ground-based validations are
mated algorithm. In this sense, examples of other more necessary, to determine how a recovery component is
complex algorithms from the literature on detecting and being displayed by the metrics.
monitoring land disturbance using Landsat time series 2) The effects of temporal mismatch issues should be
include COntinuous monitoring of Land minimized; it is essential to secure adequate imaging
Disturbance providing large-scale detection of land data sets in relation to the fire occurrence date.
disturbance [63]; Vegetation Regeneration and Distur- Image date discrepancies may introduce more noise
bance Estimates Through Time, a segmentation algo- than algorithms or metrics [62], thereby hindering
rithm to track forest changes (patch-based approach) the establishment of consistent connections between
[64]; or Ecosystem Disturbance and Recovery Tracker, a ecological meaning and RS information. Each type of
highly automated system to detect disturbances such us ecosystem has specific post-fire vegetation cover re-
wildfire burn, tree mortality or forest treatments, sponses, requiring an adequate time scale approach
processing Landsat images time series [65]. that can strike a compromise between immediate and
long-term fire effects. Moreover, several factors can
Final remarks hinder post-fire RS monitoring, such as phenology,
A deep understanding of PVR is critical for elucidating topography, vegetation characteristics and the con-
ecosystem processes and for the elaboration of effective sistency of spectral responses. All these issues still

Current Opinion in Environmental Science & Health 2021, 21:100251 www.sciencedirect.com

 Post-fire vegetation recovery using remote sensing Pérez-Cabello et al. 7

need to be carefully considered and addressed to 7. Banskota A, Kayastha N, Falkowski MJ, Wulder MA, Froese RE,
White JC: Forest monitoring using landsat time series data: a
generate a consistent PVR monitoring. review. Can J Rem Sens 2014, 40:362–384.
3) The use of RS to analyse vegetation recovery is ex-
8. Zhu Z: Change detection using landsat time series: a review
pected to grow even further in application and * of frequencies, preprocessing, algorithms, and applications.
prominence as new sensors become available (i.e. ISPRS J Photogrammetry Remote Sens 2017, 130:370–384.
Although it does not specifically focus on burnt areas, this article was
UAVs and new satellites) and bring enhanced spatial, highlighted because it proposes a comprehensive review of change
spectral and temporal resolutions to the observations. detection existing algorithms, based on Landsat time series.
In this sense, the great challenge lies in the devel- 9. Tewkesbury AP, Comber AJ, Tate NJ, Lamb A, Fisher PF:
opment of methodologies that combine the potential A critical synthesis of remotely sensed optical image change
detection techniques. Remote Sens Environ 2015, 160:1–14.
of different sensors, with particular emphasis on
studies that integrate data from active and passive 10. Hirschmugl M, Gallaun H, Dees M, Datta P, Deutscher J,
Koutsias N, Schardt M: Methods for mapping forest distur-
sensors (e.g. integrating LIDAR with UAS imagery bance and degradation from optical Earth observation data: a
[14] or GEDI data with time-series optical imagery to review. Current Forestry Reports 2017, 3:32–45.
enable historic analysis of forest height [67]). The 11. Bartels SF, Chen HYH, Wulder MA, White JC: Trends in post-
availability of open high-capacity analysis software disturbance recovery rates of Canada’s forests following
wildfire and harvest. For Ecol Manag 2016, 361:194–207.
has enhanced the potential to access and analyse
combined data sets (e.g. via Google Earth Engine, a 12. Cohen WB, Healey SP, Yang Z, Stehman SV, Brewer CK,
Brooks EB, Gorelick N, Huang C, Hughes MJ, Kennedy RE, et al.:
cloud-based storage and processing platform [68]). How similar are forest disturbance maps derived from
Moreover, the growing expansion of sensors has to go different landsat time series algorithms? Forests 2017, 8:
1–19.
hand-in-hand with the development of algorithms
that can monitor changes through time irrespective 13. Martínez S, Aguado I, Chuvieco E: Remote sensing techniques
* * applied to post-fire regeneration. In Fire effects on soil prop-
of the characteristics of each platform [69]. erties. Edited by Pereira Paulo, Solera Jorge Mataix,
Ubeda Xavier, Reain Guillermo, E Artemi Cerda, CSIRO Pub-
lishing; 2019:659.
This manuscript was particularly interesting because it is a very recent
Declaration of competing interest study of remote sensing techniques used in the framework of change
The authors declare that they have no known competing analyses in areas affected by forest fires.
financial interests or personal relationships that could 14. Szpakowski David M, Jensen JLR: A review of the applications
have appeared to influence the work reported in this * * of remote sensing in fire ecology. Rem Sens 2019, 11:2638.
This article is highlighted as it reviews remote sensing technologies
paper. connected to fire ecology studies, taking into account different scales
and quantification methods.
Acknowledgements 15. Hislop S, Haywood A, Jones S, Soto-Berelov M, Skidmore A,
This work was made possible by the institutional support of the Environ- Nguyen TH: A satellite data driven approach to monitoring
mental Sciences Institute (IUCA) of the University of Zaragoza, the and reporting fire disturbance and recovery across boreal
research project HARMO-LS2 (UZCUD2020-HUM-02): “Harmonización and temperate forests. Int J Appl Earth Obs Geoinf 2020, 87:
de imágenes Landsat/Sentinel-2 para el seguimiento de la dinámica veg- 102034.
etal”, and the São Paulo Research Foundation (FAPESP) (grant #2019/ 16. Hislop S, Jones S, Soto-Berelov M, Skidmore A, Haywood A,
07357e8). Nguyen TH: Using landsat spectral indices in time-series to
assess wildfire disturbance and recovery. Rem Sens 2018, 10:
References 1–17.
Papers of particular interest, published within the period of review, 17. Nguyen TH, Jones SD, Soto-Berelov M, Haywood A, Hislop S:
have been highlighted as: A spatial and temporal analysis of forest dynamics using
Landsat time-series. Remote Sens Environ 2018, 217:461–475.
* of special interest
* * of outstanding interest 18. Wulder MA, Loveland TR, Roy DP, Crawford CJ, Masek JG,
Woodcock CE, Allen RG, Anderson MC, Belward AS, Cohen WB,
et al.: Current status of Landsat program, science, and ap-
1. Pausas JG, Bond WJ: On the three major recycling pathways plications. Remote Sens Environ 2019, 225:127–147.
in terrestrial ecosystems. Trends Ecol Evol 2020, 35:767–775.
19. Claverie M, Ju J, Masek JG, Dungan JL, Vermote EF, Roger J-
2. Pickell PD, Hermosilla T, Frazier R J, Coops NC, Wulder MA: Forest CC, Skakun SV, Justice C: The Harmonized Landsat and
recovery trends derived from Landsat time series for North
Sentinel-2 surface reflectance data set. Remote Sens Environ
American boreal forests. Int J Rem Sens 2016, 37:138–149. 2018, 219:145–161.
3. Pereira P, Francos M, Brevik EC, Ubeda X, Bogunovic I: Scien- 20. Fernandez-Carrillo A, McCaw L, Tanase MA: Estimating pre-
ceDirect Post-fire soil management. Current Opin Environ-
scribed fire impacts and post-fire tree survival in eucalyptus
mental Science Health 2018, 5:26–32. forests of Western Australia with L-band SAR data. Remote
4. Fernández-Guisuraga JM, Suárez-Seoane S, Calvo L: Modeling Sens Environ 2019, 224:133–144.
Pinus pinaster forest structure after a large wildfire using
21. Bohlin I, Olsson H, Bohlin J, Granström A: Quantifying post-fire
remote sensing data at high spatial resolution. For Ecol
fallen trees using multi-temporal lidar. Int J Appl Earth Obs
Manag 2019, 446:257–271. Geoinf 2017, 63:186–195.
5. Chu T, Guo X: Remote sensing techniques in monitoring post- 22. Bolton DK, Coops NC, Wulder MA: Characterizing residual
fire effects and patterns of forest recovery in boreal forest
structure and forest recovery following high-severity fire in
regions: a review. Rem Sens 2013, 6:470–520. the western boreal of Canada using Landsat time-series
6. Gitas I, Mitri G, Veraverbeke S, Polychronaki A: Advances in and airborne lidar data. Remote Sens Environ 2015, 163:
remote sensing of post-fire vegetation recovery monitoring - 48–60.
a review. Remote Sensing Biomass Principles Applications 2012,
23. Duncanson L, Neuenschwander A, Hancock S, Thomas N,
https://doi.org/10.5772/20571. Fatoyinbo T, Simard M, Silva CA, Armston J, Luthcke SB, Hofton M,

www.sciencedirect.com Current Opinion in Environmental Science & Health 2021, 21:100251

8 Environmental impact assessment: Fire. Prevention, management and challenges

et al.: Biomass estimation from simulated GEDI, ICESat-2 and 39. Fernandez-Manso A, Quintano C, Roberts DA: Burn severity in-
NISAR across environmental gradients in Sonoma County, fluence on post-fire vegetation cover resilience from Landsat
California. Remote Sens Environ 2020, 242:111779. MESMA fraction images time series in Mediterranean forest
ecosystems. Remote Sens Environ 2016, 184:112–123.
24. Schneider FD, Ferraz A, Hancock S, Duncanson LI, Dubayah RO,
Pavlick RP, Schimel DS: Towards mapping the diversity of 40. Bullock EL, Woodcock CE, Olofsson P: Monitoring tropical
canopy structure from space with GEDI. Environ Res Lett forest degradation using spectral unmixing and Landsat time
2020, 15:115006. series analysis. Remote Sens Environ 2020, 238:110968.
25. Samiappan S, Hathcock L, Turnage G, McCraine C, Pitchford J, 41. Jin S, Yang L, Danielson P, Homer C, Fry J, Xian G:
Moorhead R: Remote sensing of wildfire using a small un- A comprehensive change detection method for updating the
manned aerial system: post-fire mapping, vegetation recov- National Land Cover Database to circa 2011. Remote Sens
ery and damage analysis in grand bay, Mississippi/Alabama, Environ 2013, 132:159–175.
USA. Drones 2019, 3:1–18.
42. Palaiologou P, Essen M, Hogland J, Kalabokidis K: Locating
26. Fernández-Guisuraga JM, Sanz-Ablanedo E, Suárez-Seoane S, forest management units using remote sensing and
* Calvo L: Using unmanned aerial vehicles in postfire vegeta- geostatistical tools in north-central Washington, USA. Sen-
tion survey campaigns through large and heterogeneous sors 2020, 20:2454.
areas: opportunities and challenges. Sensors 2018, 18:586.
This study is of outstanding interest because it presents the opportu- 43. Torres J, Gonçalves J, Marcos B, Honrado J: Indicator-based
nities and challenges of using drones for monitoring large and het- * * assessment of post-fire recovery dynamics using satellite
erogeneous burned areas. NDVI time-series. Ecol Indicat 2018, 89:199–212.
This study is especially interesting as it combines different indicators of
27. Talucci AC, Forbath E, Kropp H, Alexander HD, DeMarco J, recovery dynamics using satellite NDVI time series, in order to assess
Paulson AK, Zimov NS, Zimov S, Loranty MM: Evaluating post- changes in ecosystem and landscape functioning after a fire. More-
fire vegetation recovery in Cajander Larch Forests in North- over, the authors highlight the lack of satellite-based indicators that can
eastern Siberia using UAV derived vegetation indices. Rem describe specific facets of post-fire vegetation recovery, such as rate
Sens 2020, 12:2970. and completeness.
28. Meng R, Wu J, Zhao F, Cook BD, Hanavan RP, Serbin SP: 44. White JC, Wulder MA, Hermosilla T, Coops NC, Hobart GW:
* Measuring short-term post-fire forest recovery across a burn A nationwide annual characterization of 25 years of forest
severity gradient in a mixed pine-oak forest using multi- disturbance and recovery for Canada using Landsat time
sensor remote sensing techniques. Remote Sens Environ series. Remote Sens Environ 2017, 194:303–321.
2018, 210:282–296.
This publication is of outstanding interest because it reveals the 45. Kennedy RE, Yang Z, Cohen WB, Pfaff E, Braaten J, Nelson P:
importance of linking multi-sensor remote sensing techniques (e.g., Spatial and temporal patterns of forest disturbance and
NASA Goddard’s LIDAR and Hyperspectral and Thermal [G-LiHT]) to regrowth within the area of the Northwest Forest Plan.
monitor forest dynamics by taking into account different vegetation Remote Sens Environ 2012, 122:117–133.
layers (i.e., canopy, understory).
46. Du P, Wang X, Chen D, Liu S, Lin C, Meng Y: An improved
29. Vogeler JC, Yang Z, Cohen WB: Mapping post-fire habitat change detection approach using tri-temporal logic-verified
characteristics through the fusion of remote sensing tools. change vector analysis. ISPRS J Photogrammetry Remote
Remote Sens Environ 2016, 173:294–303. Sens 2020, 161:278–293.

30. Liang L, Hawbaker TJ, Zhu Z, Li X, Gong P: Forest disturbance 47. Woodcock CE, Loveland TR, Herold M, Bauer ME: Transitioning
interactions and successional pathways in the Southern * * from change detection to monitoring with remote sensing: a
Rocky Mountains. For Ecol Manag 2016, 375:35–45. paradigm shift. Remote Sens Environ 2020, 238:111558.
This article reveals and clarifies the change of paradigm from bi-
31. Villarreal ML, Norman LM, Buckley S, Wallace CSA, Coe MA: temporal approaches to characterizing trends using time-series. Be-
Multi-index time series monitoring of drought and fire effects sides, it identifies several trends in using remote sensing that may
on desert grasslands. Remote Sens Environ 2016, 183: provide improvements in PVR studies (multi-sensor fusion or devel-
186–197. opment of Operational Applications, among others).
32. Morresi D, Vitali A, Urbinati C, Garbarino M: Forest spectral re- 48. Cohen WB, Yang Z, Healey SP, Kennedy RE, Gorelick N:
covery and regeneration dynamics in stand-replacing wild- A LandTrendr multispectral ensemble for forest
fires of central Apennines derived from Landsat time series. disturbance detection. Remote Sens Environ 2018, 205:131–140.
Rem Sens 2019, 11:308.
49. Kennedy RE, Yang Z, Cohen WB: Detecting trends in forest
33. Chompuchan C, Lin C-YY: Assessment of forest recovery at disturbance and recovery using yearly Landsat time series:
Wu-Ling fire scars in Taiwan using multi-temporal Landsat 1. LandTrendr - temporal segmentation algorithms. Remote
imagery. Ecol Indicat 2017, 79:196–206. Sens Environ 2010, 114:2897–2910.
34. Veraverbeke S, Gitas I, Katagis T, Polychronaki A, Somers B, 50. Dunn CJ, O’Connor CD, Reilly MJ, Calkin DE, Thompson MP:
Goossens R: Assessing post-fire vegetation recovery using Spatial and temporal assessment of responder exposure to
red-near infrared vegetation indices: accounting for back- snag hazards in post-fire environments. For Ecol Manag 2019,
ground and vegetation variability. ISPRS J Photogrammetry 441:202–214.
Remote Sens 2012, 68:28–39.
51. Bright BC, Hudak AT, Kennedy RE, Braaten JD, Henareh
35. Vo VD, Kinoshita AM: Remote sensing of vegetation condi- Khalyani A: Examining post-fire vegetation recovery with
tions after post-fire mulch treatments. J Environ Manag 2020: Landsat time series analysis in three western North American
260. forest types. Fire Ecology 2019, 15:8.
36. Adagbasa EG, Adelabu SA, Okello TW: Development of post- 52. Frazier RJ, Coops NC, Wulder MA, Hermosilla T, White JC:
fire vegetation response-ability model in grassland moun- Analyzing spatial and temporal variability in short-term rates
tainous ecosystem using GIS and remote sensing. ISPRS J of post-fire vegetation return from Landsat time series.
Photogrammetry Remote Sens 2020, 164:173–183. Remote Sens Environ 2018, 205:32–45.
37. Fernández-Guisuraga JM, Calvo L, Fernández-García V, Marcos- 53. Wang J, Zhang X: Investigation of wildfire impacts on land
Porras E, Taboada Á, Suárez-Seoane S: Efficiency of remote surface phenology from MODIS time series in the western US
sensing tools for post-fire management along a climatic forests. ISPRS J Photogrammetry Remote Sens 2020, 159:
gradient. For Ecol Manag 2019, 433:553–562. 281–295.
38. Viana-Soto A, Aguado I, Salas J, García M: Identifying post-fire 54. Vogelmann JE, Xian G, Homer C, Tolk B: Monitoring gradual
recovery trajectories and driving factors using landsat time ecosystem change using Landsat time series analyses: case
series in fire-prone mediterranean pine forests. Rem Sens studies in selected forest and rangeland ecosystems. Remote
2020, 12:1499. Sens Environ 2012, 122:92–105.

Current Opinion in Environmental Science & Health 2021, 21:100251 www.sciencedirect.com

 Post-fire vegetation recovery using remote sensing Pérez-Cabello et al. 9

55. Hamunyela E, Brandt P, Shirima D, Do HTT, Herold M, Roman- Kennedy RE, et al.: Mapping forest change using stacked
Cuesta RM: Space-time detection of deforestation, forest generalization: an ensemble approach. Remote Sens Environ
degradation and regeneration in montane forests of Eastern 2018, 204:717–728.
Tanzania. Int J Appl Earth Obs Geoinf 2020, 88:102063.
63. Zhu Z, Zhang J, Yang Z, Aljaddani AH, Cohen WB, Qiu S,
56. Cunha J, Nóbrega RLB, Rufino I, Erasmi S, Galvão C, Valente F: Zhou C: Continuous monitoring of land disturbance based on
Surface albedo as a proxy for land-cover clearing in Landsat time series. Remote Sens Environ 2020, 244:111116.
seasonally dry forests: evidence from the Brazilian Caatinga.
Remote Sens Environ 2020:238. 64. Joseph Hughes M, Douglas Kaylor S, Hayes DJ: Patch-based
forest change detection from Landsat time series. Forests
57. Burrell AL, Evans JP, Liu Y: Detecting dryland degradation 2017, 8:1–22.
using time series segmentation and residual trend analysis
(TSS-RESTREND). Remote Sens Environ 2017, 197:43–57. 65. Koltunov A, Ramirez CM, Ustin SL, Slaton M, Haunreiter E,
eDaRT: The Ecosystem Disturbance and Recovery
58. Verbesselt J, Hyndman R, Newnham G, Culvenor D: Detecting Tracker system for monitoring landscape disturbances
trend and seasonal changes in satellite image time series. and their cumulative effects. Remote Sens Environ 2020,
Remote Sens Environ 2010, 114:106–115. 238:111482.
59. Cardille JA, Fortin JA: Remote Sensing of Environment 66. Pesaresi S, Mancini A, Quattrini G, Casavecchia S: Mapping
Bayesian updating of land-cover estimates in a data-rich mediterranean forest plant associations and habitats with
environment 2016, 186:234–249. functional principal component analysis using Landsat 8
NDVI time series. Rem Sens 2020, 12:1132.
60. Fortin JA, Cardille JA, Perez E: Multi-sensor detection of forest-
* cover change across 45 years in Mato Grosso, Brazil. Remote 67. Potapov P, Li X, Hernandez-Serna A, Tyukavina A, Hansen MC,
Sens Environ 2020:238. Kommareddy A, Pickens A, Turubanova S, Tang H, Silva CE,
This work is of interest because it utilizes new algorithms, such as et al.: Mapping global forest canopy height through integra-
BULC (Bayesian Updating of Land Cover) and GEE, that allow the tion of GEDI and Landsat data. Remote Sens Environ 2021,
fusion of different spaceborne images to monitor changes at sub- 253:112165.
annual time scales using discrete variables.
68. Kennedy RE, Yang Z, Gorelick N, Braaten J, Cavalcante L,
61. Savage SL, Lawrence RL, Squires JR: Mapping post- Cohen WB, Healey S: Implementation of the LandTrendr al-
disturbance forest landscape composition with Landsat sat- gorithm on Google Earth engine. Rem Sens 2018, 10:1–10.
ellite imagery. For Ecol Manag 2017, 399:9–23.
69. Crowley MA, Cardille JA: Remote sensing ’ s recent and future
62. Healey SP, Cohen WB, Yang Z, Kenneth Brewer C, Brooks EB, contributions to landscape ecology. Current Landscape Ecol-
Gorelick N, Hernandez AJ, Huang C, Joseph Hughes M, ogy Reports 2020, 5:45–57.

www.sciencedirect.com Current Opinion in Environmental Science & Health 2021, 21:100251