Content uploaded by J.P. Nunes
Author content
All content in this area was uploaded by J.P. Nunes on Aug 19, 2022
Content may be subject to copyright.
RESEARCH ARTICLE
How do large wildfires impact sediment redistribution over
multiple decades?
Dante Follmi
1
| Jantiene Baartman
1
| Akli Benali
2
| Joao Pedro Nunes
1,3
1
Soil Physics and Land Management Group,
Wageningen University, Wageningen, The
Netherlands
2
Forest Research Centre, School of
Agriculture, University of Lisbon, Lisbon,
Portugal
3
CE3C: Centre for Ecology, Evolution and
Environmental Changes, School of Science,
University of Lisbon, Lisbon, Portugal
Correspondence
Dr. ir. Jantiene EM Baartman, Soil Physics and
Land Management Group (SLM), Wageningen
University, PO Box 47, Wageningen 6700 AA,
The Netherlands.
Email: jantiene.baartman@wur.nl
Funding information
cE3c research center, Grant/Award Number:
UIDB/00329/2020; Erasmus+travel
fellowship; Portuguese Foundation for Science
and Technology, Grant/Award Numbers:
FRISCO project (PCIF/MPG/0044/2018),
individual grant to A Benali
(CEECIND/03799/2018/C, individual grant to
JP Nunes (IF/00586/2015)
Abstract
Wildfires have become an increasing threat for Mediterranean ecosystems, due to
increasing climate change-induced wildfire activity and changing land management
practices. In addition to the initial risk, wildfires can alter the soil in various ways—
depending on fire severity—and cause enhanced post-fire erosion. Usually, post-fire
erosion studies focus on a short time window and lack the attention for sediment
dynamics at larger spatial scales. Yet, these large spatial and temporal scales are fun-
damental for a better understanding of long-term destructive effects of multiple
recurring wildfires on post-fire erosion processes and catchment sediment dynamics.
In this study the landscape evolution model LAPSUS was used to simulate erosion
and deposition in the 404 km
2
´
Agueda catchment in north-central Portugal over a
41-year (1979–2020) timespan, including eight wildfires each burning >1000 ha. To
include variation in fire severity and its impact on the soil, four burn severity classes,
represented by the difference normalized burn ratio (dNBR), were parameterized.
Although model calibration was difficult due to lack of spatial and temporal measured
data, the results show that long-term post-fire net erosion rates were significantly
higher in the wildfire scenarios (5.95 ton ha
1
yr
1
) compared to those of a non-
wildfire scenario (0.58 ton ha
1
yr
1
). Furthermore, erosion values increased with
burn severity and multiple wildfires increased the overall catchment sediment build-
up. Simulated erosion patterns showed great spatial variability, with large deposition
and erosion rates inside streams. This variability made it difficult to identify land uses
that were most sensitive for post-fire erosion, because some land uses were located
in more erosion-sensitive areas (e.g. streams, gullies) or were more affected by high
burn severity levels than others. Despite these limitations, LAPSUS performed well
on addressing spatial sediment processes and can contribute to pre-fire management
strategies, by identifying locations at risk of post-fire erosion.
KEYWORDS
burn severity, land and wildfire management, long-term modelling, post-fire erosion, sediment
connectivity, wildfires
1|INTRODUCTION
In recent decades, wildfires have become an increasing threat for
Mediterranean ecosystems as a consequence of increasing frequency
of weather conditions conducive for wildfires, which is likely to
increase further due to climate change (Moriondo et al., 2006). This
could lead to an increase in the number of years with high wildfire
risk, an extension of the wildfire season, and an increase of extreme
wildfire events. Subsequent impacts on vegetation and soils play an
important role in land degradation (e.g. Malvar et al., 2011; Nunes,
Naranjo Quintanilla, et al., 2018; Pausas et al., 2008; Shakesby, 2011;
Shakesby et al., 1996) As a consequence of: (i) a reduction of
Received: 20 March 2022 Revised: 24 June 2022 Accepted: 27 June 2022
DOI: 10.1002/esp.5441
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.
© 2022 The Authors. Earth Surface Processes and Landforms published by John Wiley & Sons Ltd.
Earth Surf. Process. Landforms. 2022;1–18. wileyonlinelibrary.com/journal/esp 1
evapotranspiration; (ii) a decrease in soil water retention due to
hydrophobicity; and (iii) a reduction in obstacles, the likelihood of
overland flow is increased, favouring erosion (Shakesby &
Doerr, 2006). Also, soil structure may deteriorate after a wildfire by
the combustion of soil organic matter (Shakesby & Doerr, 2006),
promoting soil erosion (Mataix-Solera et al., 2011). Post-fire erosion is
often linked to fire severity (Borrelli et al., 2017; de Vente &
Poesen, 2005; Delestre et al., 2017; Foster & Meyer, 1975; Ortíz-
Rodríguez et al., 2019; Vieira et al., 2015). In addition, recurrence of
multiple wildfires at the same site is perceived to slow down vegeta-
tion recovery, increasing runoff, erosion, and nutrient removal
(Hosseini et al., 2016; McGuire & Youberg, 2019).
Wildfires are a common threat in Portugal. Mateus and Fernandes
(2014) estimated that the accumulated burned area was 4.2 10
6
ha
between 1975 and 2012 (45% of Portugal’s land surface), the
highest of all southern European countries. Also, Portuguese wildfires
had the second largest mean fire size compared to other southern
European countries (24.5 ha between 2000 and 2011). Besides an
increasing frequency of weather conditions conducive for wildfires
(Calheiros et al., 2020), anthropogenic activity such as land use
change, induced by socio-economic change and urbanization, also
makes an important contribution to wildfire occurrence (Pausas
et al., 2008; Shakesby, 2011). In the 20th century in Portugal, part of
the agricultural land and shrublands was converted first into Maritime
Pine plantations, and towards the end of the century into Eucalypt
plantations (Hawtree et al., 2015; Jones et al., 2011; Vasconcelos
Ferreira et al., 2010). Initially, the objective of this conversion was to
increase the provision of hydrological services such as erosion protec-
tion and flood mitigation, however the runoff and erosion increase
caused by wildfire disturbances can, in the long run, negate the
intended benefits (Carvalho-Santos et al., 2019; Nunes, Naranjo
Quintanilla, et al., 2018).
Several studies in north-central Portugal have shown an increase
in erosion in burnt areas (Campo et al., 2006; Ferreira et al., 2008;
Hosseini et al., 2016; Hyde et al., 2007; Shakesby et al., 1993,1996).
However, these studies mainly focus on short-term processes during
a few post-fire years, and on smaller plot or hillslope scales
(Shakesby, 2011; Shakesby & Doerr, 2006). Nevertheless, several
studies already formulate the need for investigation of the long-term
effects, including a consecutive number of wildfires (e.g. McGuire &
Youberg, 2019), which could provide insights into the relationship
between soil degradation, vegetation change, and other landscape
processes that generally occur over a much wider time window. In
particular, a better understanding is needed of the importance of
occasional severe post-fire erosion events in comparison with long-
term background erosion processes (Nunes, Doerr, et al., 2018;
Shakesby, 2011; Shakesby & Doerr, 2006). Furthermore, investigating
a larger catchment scale can help identify locations with post-fire ero-
sion risk to help delineate intervention strategies, as well as examine
the transport pathways of sediments from burnt areas which can have
important negative impacts on water quality (Nunes, Bernard-Jannin,
et al., 2018,2020; Shakesby, 2011).
Due to the difficulty of conducting field assessments, numerical
modelling studies focusing on the Iberian Peninsula have been
conducted to investigate post-fire erosion response. However, these
have focused on rather small spatial (patch, hillslope, or small
catchment) and temporal (the first few post-fire years) scales
(Fernández et al., 2010; Hosseini et al., 2018; Soto &
Díaz-Fierros, 1998; Vieira et al., 2015,2018; Wu, Baartman &
Nunes, 2021; Wu et al., 2021; Zema et al., 2020). Models used include
the empirical Revised Universal Soil Loss Equation (RUSLE;
Fernández & Vega, 2016; Fernández et al., 2010), the semi-empirical
Morgan–Morgan–Finney (MMF; Fernández et al., 2010) model, and
the physically based Pan-European Soil Erosion Risk Assessment
(PESERA; Esteves et al., 2012), Water Erosion Prediction Project
(WEPP; Soto & Díaz-Fierros, 1998), and Limburg Soil Erosion Model
(OpenLISEM; Wu, Baartman & Nunes, 2021; Wu et al., 2021). Esteves
et al. (2012) applied the PESERA model to two headwater catchments
(9.7 and 100 ha) for a 50-year simulation period; however, the model
only simulated on-site erosion, not accounting for off-site effects cau-
sed by sediment transport and deposition. Further studies conducted
at the headwater catchment scale (1 km) using LandSoil (Pastor
et al., 2019), the Soil and Water Assessment Tool (SWAT; Nunes,
Bernard-Jannin, et al., 2018, Nunes, Naranjo Quintanilla, et al., 2018),
and OpenLISEM (Wu et al., 2021) show the importance of wildfires
for erosion and sediment yield even at longer time scales (20 and
10 years, respectively). However, the complexity of these models—
and the large number of parameters required to appropriately simu-
late post-fire impacts—has limited their applicability with sufficient
spatial resolution for larger areas, preventing an analysis of the long-
term impacts of recurring wildfires for larger landscapes (>100 km
2
),
which are typically the units at which forest planning is made.
Reduced-complexity landscape evolution models (LEMs) are capable
of investigating long-term landscape sediment dynamics (Baartman,
van Gorp, et al., 2012; Tucker & Hancock, 2010). Thus, LEMs can be
applied to investigate long-term (historical) landscape evolution and
sediment behaviour in a wildfire-affected catchment under limited
data availability conditions.
The aim of this study was to investigate how multiple wildfires
have affected spatial and temporal erosion and deposition dynamics
in the
´
Agueda catchment (north-central Portugal, 404 km
2
) using a
long-term modelling approach. We applied LEM Landscape Process
Modelling at Multi-dimensions and Scales (LAPSUS; Schoorl
et al., 2000,2002) to the study area for a 41-year (1979–2020) time
period using a wildfire and no-wildfire scenario. In the wildfire
scenario, eight major wildfires which occurred in this period were
parameterized, with varying severity and spatial extent; while in the
no-wildfire scenario, no wildfires were assumed to have taken place.
Results were evaluated in terms of (spatially explicit) erosion and
deposition rates over time and how these were affected by (1) land
use and (2) multiple wildfire occurrence.
2|METHODS
2.1 |Study area
The
´
Agueda catchment, located in north-central Portugal (40.62
latitude, 8.27longitude), is commonly affected by wildfires. The
catchment covers 404 km
2
and is situated in the Caramulo mountain
formation with peaks of 1100 m above sea level (Figure 1). Slope
steepness ranges between 10 and 25, with an average of 13.4. The
catchment has a humid Mediterranean climate with annual precipita-
tion of 1000–2500 mm (increasing with altitude), falling mostly in the
2FOLLMI ET AL.
FIGURE 1 Location and topography of the
´
Agueda catchment.
FIGURE 2 Land use in 1995 (a) and 2018 (b) and bar graph (c) showing the percentage cover of each land use type within the catchment
(1995: dashed bars, 2018: clear bars). COS is ‘carta de Uso e Ocupaç˜ao do solo’.
FOLLMI ET AL.3
autumn and winter months (Hawtree et al., 2015; Nunes, Naranjo
Quintanilla, et al., 2018). Mean annual rainfall of 1787 mm was
estimated for the period between 1936 and 2010 (Hawtree
et al., 2015).
The geology of the Caramulo mountain range is characterized by
schist on lower elevations and granites on higher elevated areas
(Hawtree et al., 2015; Tavares Wahren et al., 2016). Soil formation in
the region generally resulted in shallow soils with low organic matter
content and coarse texture properties (Nunes, Naranjo Quintanilla,
et al., 2018). Soils are mostly classified as Leptosols between 40 and
75 cm deep, and Cambisols with a depth between 40 and 100 cm
(Tavares Wahren et al., 2016; WRB, 2015).
Figure 2shows the land use in 1995 and 2018, with the percent-
ages for each land use type. Land use evolution (where Pine has been
changed for Eucalypt between 1995 and 2018) can be seen, indicating
the afforestation practices in the late 20th and beginning 21st century
(Ferreira, 1997; Hawtree et al., 2015). Recent recorded major wildfires
occurred in 1985, 1986, 1991, 1995, 2005, 2013 (Hawtree
et al., 2015), 2016 and 2017. A large wildfire, therefore, took place
every 5 years in the catchment and had a size ranging between
1500 and 9200 ha. Most wildfires occurred during summer (July,
August and September), except for the wildfire in 1986 that happened
in April.
2.2 |LAPSUS model description
LEM LAPSUS is a physically based model that is able to investigate
long-term and large-scale spatial landscape evolution (Baartman, van
Gorp, et al., 2012; Schoorl et al., 2000,2002). The original model
includes topography, soil depth, changing land use, climate variability
(annual rainfall), and soil and vegetation characteristics (annual
quantities as well as spatial evapotranspiration and infiltration)
(e.g. Baartman, Temme, et al., 2012); in this study spatial burn severity
was also included. Considering the multi-decade timespan of this
study, weathering of parent material was not included
(Alexander, 1985). Figure 3shows an overview of the model
procedure. Model output consists of annual maps of erosion, deposi-
tion, soil depth, and runoff, from which temporal and spatial sediment
yield was derived.
LAPSUS is a cellular automata model. Routing of runoff and
sediment influx towards neighbouring cells is determined using a
multiple flow algorithm based on Holmgren (1994):
fi¼ΛðÞ
p
i
Pmax 8
j¼1ΛðÞ
p
j
ð1Þ
where f
i
is the fraction of runoff out of a cell going in direction
i(equal to the gradient of the slope Λin the direction i, powered by
the pconvergence factor) divided by the total sum of slopes of all
lower elevated neighbouring cells j, powered by the factor p.
Incoming water flow Q(m
2
yr
1
) for a particular cell equals the
incoming water flow (from upslope cells) plus the rainfall on the cell.
From this, infiltration and evaporation are subtracted and the remain-
der is used in the calculation for erosion and deposition and trans-
ported to the next (downslope) cell.
For sediment transport, the continuity of transport and conserva-
tion of mass principles apply. Equal bulk density of eroded and
deposited material is assumed. For each transition from cell to cell
along length dx (m) of a finite element, sediment transport capacity
C(m
2
yr
1
) is calculated (Equation (2)) as a function of fractional
discharge Qand slope tangent Λ(Kirkby, 1971):
C¼γQmΛnð2Þ
with discharge exponent mand slope exponent n, and γa constant for
unit conversion with value 1. This is the well-known stream-power
equation (Lague, 2014). Transport capacity is compared to the
incoming amount of sediment in transport S
0
(m
2
yr
1
) to calculate
the amount of sediment S(m
2
yr
1
) that will be transported:
S¼CþS0CðÞedx=hð3Þ
FIGURE 3 Model procedure for LAPSUS showing input data, parameterization, output, and calibration.
4FOLLMI ET AL.
with dx the cell size (m). Term h(m) refers to the transport capacity
divided by the detachment capacity (D,myr
1
) in case of erosion, and
transport capacity divided by settlement capacity (T,myr
1
) in case
of sedimentation:
D¼Kes QΛð4Þ
T¼Pes QΛð5Þ
where K
es
(m
1
) is an aggregated surface factor representing the erod-
ibility of the surface, while P
es
(m
1
) is a similar factor for sedimenta-
tion potential. The potential for erosion or deposition depends on
lumped parameters K
es
and P
es
, aggregating different landscape char-
acteristics (vegetation, soil properties, etc.), which are parameterized
for different land uses and in this case for burnt areas, both immedi-
ately after the wildfire and for the recovery period.
2.3 |Data collection and model parameterization
The main input parameters needed by LAPSUS include a digital eleva-
tion model (DEM), land use and soil depth maps, and annual time
series of precipitation, evapotranspiration, and infiltration. In this
study, a cell size of 25 m and an annual time step were used. The
DEM of the Vouga basin was obtained from the Instituto Geográfico
do Exército. Soil depth data were based on maps from Tavares
Wahren et al. (2016), created using a neural network and soil depth
surveys at 11 locations throughout the catchment. The discharge and
slope exponents (mand n, respectively; Equation 2) were initially set
at 2, based on the work of Kirkby (1987). The initial convergence
factor p(Equation 1) was set to 4.0, based on the proposed value in
the LAPSUS user guide (van Gorp, 2015).
To analyse the multi-decadal effect of wildfires on erosion, two
scenarios were formulated. (i) Wildfires: model parameterization
accounts for large multiple wildfires (>1000 ha) covering the majority
of burnt areas between 1979 and 2020. (ii) No-wildfires: wildfires were
excluded, thus forming a ‘baseline’run. Both scenarios used the same
land uses and climate variability (i.e. rainfall, evapotranspiration, and
infiltration).
Five land use maps for 1995, 2007, 2010, 2015, and 2018 were
used to cover land use change during 41 years (Carta de Uso e
Ocupaç˜ao do Solo [COS] scale: 1:100 000) (see Figure 2for 1995 and
2018; all years given in Supplementary Material S1). These land use
maps were adapted to use a less detailed classification and, for the
wildfire scenario, to include the effect of the wildfire in the first, sec-
ond, and third post-fire years based on burn severity. Only major wild-
fires >1000 ha were included, which covered 80% of the total burnt
area in the 41-year period considered. These wildfires happened in
1985, 1986, 1991, 1995, 2005, 2013, 2016, and 2017 (Figure 4). Burn
severity was estimated by deriving the normalized burn ratio (NBR)
using Landsat satellite imagery (ESPA-USGS, 2020), which was used
to calculate the different normalized burn ratio (dNBR) by subtracting
the pre-fire NBR image from the post-fire NBR image:
dNBR ¼NBRPREFIRE NBRPOSTFIRE ð6Þ
2.3.1 | Erodibility and sediment potential
parameterization
The dNBR maps were classified to derive areas of different burn
severity (Table 1) and combined with the land use maps. For each land
use–burn severity combination, parameterization of lumped K
es
and
P
es
parameters was based on the USLE vegetation cover factor
(USLE_C; Table 2), as vegetation cover is assumed to affect erodibility
(i.e. more vegetation cover leads to lower erodibility) and sedimenta-
tion potential (i.e. more vegetation cover is assumed to increase sedi-
mentation of transported material). Since LAPSUS does not account
for parameterization of the USLE_C value directly, USLE_C values
were used as a proxy for fractional change in K
es
and P
es
values which
are used in the model (Equations 4and 5). Thus, the absolute value of
the USLE_C factor was not used, but the relative differences in
USLE_C factors between land use–burn severity combinations
(Table 2) were transferred to the K
es
and P
es
values for these land
FIGURE 4 Timeline with input land use maps and indication of fire events and post-fire years.
FOLLMI ET AL.5
use–burn severity combinations. As specific USLE_C data is scarce, a
distinction was only made between agriculture, urban, and a combined
group of forest (Eucalypt, Pine, other broadleaf forest) and shrub
vegetation. For unburnt land uses, USLE_C factors were derived from
Nunes, Naranjo Quintanilla, et al. (2018) and were set to 0.001 for the
forest/shrub vegetation land uses (Table 2). Urban USLE_C factors
were set to be two times lower, due to paved and impermeable
conditions of infrastructure and build-up. For agriculture, the value
was set at two times higher than the other land uses based on
Carvalho-Santos et al. (2016), which simultaneously incorporates the
presence of agricultural terraces in the study area. Burn severity was
parameterized for each land use (Table 2) based on Fernández and
Vega (2016), who identified a power relationship between RUSLE’s
C-factor and burn severity (R
2
=0.77) at 87 in-situ burnt plot
measurements in Galicia, Spain.
2.3.2 | Climate data
Total annual quantities of rainfall data for the
´
Aqueda catchment are
based on rain gauge weather station data of the ‘Campia’station,
derived from the ‘Sistema Nacional de Informaç˜ao de Recursos
Hídricos’(SNIRH, 2020). Daily totals (mm/day) were transformed to
annual totals, resulting in a time series of hydrological years 1979/80
until 2019/20. A hydrological year is set from 1 October to
30 September. Gaps in rainfall data were filled with data from the
nearby Bouç˜a station.
To derive the total annual time series of infiltration and actual
evapotranspiration, the catchment outlet streamflow was used. One
of the reasons to focus on the
´
Agueda catchment is the availability of
the (quite unique) historical daily streamflow data (1936–2012),
recorded at the streamflow gauging point ‘Ponte de
´
Agueda’at the
outlet of the catchment (Hawtree et al., 2015). To derive actual
evapotranspiration, the following equation was used:
ETactual ¼PQð7Þ
where ET
actual
is the annual actual evapotranspiration (mm), Pis the
annual rainfall (m), and Qis the annual streamflow data at the gauging
station (mm); deep percolation losses were assumed negligible due to
the relatively impermeable geology.
During the dry months (June to September) there were impound-
ments in the streamflow data due to the closure of a sluice in the
river. Missing streamflow or suspicious data from 2004 until 2020
was corrected based on powered regression of streamflow and rainfall
data (R
2
=0.91), based on the whole rainfall and streamflow dataset
(1936–2012). By using this regression curve, known rainfall values
were translated into predicted streamflow values.
To estimate time series of annual infiltration, the concept of base-
flow (mm/yr) was used, defined as the fraction of the corrected
streamflow data that is not directly generated after a rainfall event in
the form of runoff. As this is the lateral flow through the soil column
and not the runoff transported over the surface, it therefore serves as
a proxy for infiltration. Estimation of baseflow was done by the
recursive digital filter method (Arnold et al., 1995; Nathan &
McMahon, 1990). This method is considered robust no matter the
stream variabilities or catchment size (Nathan & McMahon, 1990).
Apart from including time series data, spatial variability of infiltra-
tion and evapotranspiration was included. Differences in actual
evapotranspiration (ET
A
) values (mm) (shown in Table 3, values
without brackets) between land uses were based on multiplication by
the potential evapotranspiration (PET), equal for the whole catchment,
and the crop coefficient (K
c
), different for each land use. For
calculation of the PET, the Hargreaves method was used, using
monthly temperature variation plus altitude and latitude
(Hargreaves & Samani, 1985). For estimation of K
c
, an approach from
Maselli et al. (2014) was used, that includes seasonality of vegetation
cover and soil moisture to account for dry summer months with lower
K
c
values (see Supplementary Material S2 for an explanatory descrip-
tion about spatial ET).
To estimate spatial variable infiltration values between land uses,
the inverse curve number (CN) (Boonstra, 1994; USDA, 2019) was
estimated (shown as ‘Infiltration’in Table 3, values without brackets).
Inverse CNs were determined based on: land use classes (COS
classes); soil hydrological groups (4 classes), slope classes (5 classes);
the antecedent soil moisture condition (3 classes) and the hydrological
condition related to management (3 classes) (Boonstra, 1994). For
further explanation of the use of the CN and the parameterization of
spatial infiltration, see Supplementary Material S2.
2.4 |Model calibration
To calibrate LAPSUS, the lumped K
es
and P
es
factors (Equations 4
and 5) for the whole catchment and the m,n, and pparameters
(Equations 1and 2) were adjusted. Unfortunately, the measured
TABLE 2 USLE_C factors for each land use–burn severity class
Land use and burn severity class USLE_C
Urban
Unburnt, low, moderate and high severity 5.0 10
4
Agriculture (winter pasture)
Unburnt 2.0 10
3
Low severity and regrowth 4.4 10
3
Moderate severity 4.5 10
2
High severity 1.832 10
1
Eucalypt, Pine, Other broadleaf forest, Shrub vegetation
Unburnt 1.0 10
3
Low severity and regrowth 4.4 10
3
Moderate severity 4.5 10
2
High severity 1.832 10
1
TABLE 1 dNBR intervals for each burn severity class
Burn severity class dNBR range
Unburnt 0.1 ≤dNBR < 0.1
Low severity and regrowth 0.1 ≤dNBR < 0.27
Moderate severity 0.27 ≤dNBR < 0.66
High severity 0.66 ≤dNBR < 1.33
6FOLLMI ET AL.
discharge data at the outlet of the
´
Agueda catchment was not
suitable to calibrate the annual net erosion simulated by LAPSUS, as
it was derived from an uncertain (especially for heavy storms, that
carry most sediment) sediment–discharge relation. Therefore, the
simulated erosion by LAPSUS was calibrated for a headwater catch-
ment located within the
´
Agueda catchment (i.e. Macieira de Alcôba),
which experienced a small fire (10 ha) in 2011 (Nunes et al., 2020).
This smaller headwater catchment contains similar land use, slope,
and soil conditions as the
´
Agueda catchment (Nunes, Naranjo
Quintanilla, et al., 2018,2020; Tavares Wahren et al., 2016). There-
fore, sediment yields were considered representative for the larger
´
Agueda catchment. Measurements at a burnt hillslope (5.9 ha) (two
post-fire years), two agricultural areas (0.63 ha), and 2 min-interval
turbidity measurements at a hydrometric station positioned at the
outlet (1 pre-fire year and 3 post-fire years) were used to calibrate
model outputs (Nunes, Naranjo Quintanilla, et al., 2018,2020)
(see Supplementary Material S3). In addition, plot (8 2 m) studies
from Shakesby et al. (1996) in the north-west of the
´
Agueda catch-
ment (see Shakesby et al., 1996 for coordinates) and Ferreira (1997)
close to Macieira de Alcôba were used to calibrate erosion rates for
the unburnt scenario. Because these typical 8 2 m bounded plots
have limited catchment area, only erosion values in cells in LAPSUS
with close to zero flow accumulation were included (<0.228 m
3
for
second post-fire year and <0.3 m
3
for third post-fire year).
Subsequently, the values included a topographic correction
(Supplementary Material S3, values in square brackets), based on
Cerdan et al. (2010), so that values are representative for the larger
25 25 m cells in LAPSUS. Calibration was done based on the
average of multiple (post-fire) years and not for single years, since
time-lumped K
es
and P
es
factors were used. Supplementary Material
S3 lists the observed erosion values, those with topographic correc-
tion, and the erosion values as simulated by calibrated LAPSUS.
TABLE 3 Infiltration curve number values and evapotranspiration values (mm) per land use–burn severity combination (values with no
brackets); weight factors used in LAPSUS between brackets. A distinction is made between the soil hydrological groups A, B and C, D. For more
information, see Supplementary Material S2
Soil hydrological group
A, B C, D
Infiltration ET
A
Infiltration ET
A
Urban
Unburnt (unburnt maps) 2.0 (0.092) 147.47 (0.215) ––
Unburnt (fire maps) 2.0 (0.032) 147.47 (0.224) ––
Low severity and regrowth 2.0 (0.032) 147.47 (0.225) ––
Moderate and high severity 2.0 (0.032) 145.84 (0.224) ––
Agriculture
Unburnt (unburnt maps) 68.7 (1.001) 792.72 (1.158) 63.0 (0.918) 792.72 (1.158)
Unburnt (fire maps) 68.7 (1.092) 792.72 (1.207) 63.0 (1.002) 792.72 (1.207)
Low severity and regrowth 35.0 (0.561) 792.72 (1.207) 16.0 (0.257) 792.72 (1.209)
Moderate and high severity 35.0 (0.561) 326.86 (0.499) 16.0 (0.257) 326.86 (0.502)
Eucalypt
Unburnt (unburnt maps) 78.9 (1.150) 742.12 (1.084) 64.6 (0.942) 742.12 (1.084)
Unburnt (fire maps) 78.9 (1.254) 742.12 (1.130) 64.6 (1.027) 742.12 (1.130)
Low severity and regrowth 35.0 (0.560) 742.12 (1.130) 16.0 (0.255) 742.12 (1.130)
Moderate and high severity 35.0 (0.560) 308.76 (0.472) 16.0 (0.255) 308.76 (0.471)
Pine
Unburnt (unburnt maps) 78.9 (1.150) 700.76 (1.023) 64.6 (0.942) 700.76 (1.023)
Unburnt (fire maps) 78.9 (1.254) 700.76 (1.067) 64.6 (1.027) 700.76 (1.067)
Low severity and regrowth 35.0 (0.559) 700.76 (1.067) 16.0 (0.256) 700.76 (1.067)
Moderate and high severity 35.0 (0.559) 290.66 (0.444) 16.0 (0.256) 290.66 (0.443)
Other broadleaf forest
Unburnt (unburnt maps) 81.3 (1.185) 579.47 (0.846) 66.9 (0.975) 579.47 (0.846)
Unburnt (fire maps) 81.3 (1.292) 579.47 (0.882) 66.9 (1.064) 579.47 (0.882)
Low severity and regrowth 35.0 (0.563) 579.47 (0.883) 16.0 (0.258) 579.47 (0.883)
Moderate and high severity 35.0 (0.563) 272.55 (0.417) 16.0 (0.258) 272.55 (0.418)
Shrub vegetation
Unburnt (unburnt maps) 81.3 (1.185) 538.58 (0.787) 66.9 (0.975) 538.58 (0.787)
Unburnt (fire maps) 81.3 (1.292) 538.58 (0.820) 66.9 (1.064) 538.58 (0.820)
Low severity and regrowth 35.0 (0.561) 538.58 (0.820) 16.0 (0.256) 538.58 (0.820)
Moderate and high severity 35.0 (0.561) 245.40 (0.375) 16.0 (0.256) 245.40 (0.375)
FOLLMI ET AL.7
3|RESULTS
3.1 |Calibration results
Despite the time-averaged calibration, calibrated model results per
year were in all cases of the same order of magnitude as observed
data, except for the years 2012/13 and 2013/14 for agricultural
slopes in the unburnt scenario: 0.063 vs 0.4 ton ha
1
and 1.13 vs 0.44
ton ha
1
, respectively. The calibration of all observed values was very
good (R
2
=0.94) (Figure 5).
Calibrated values for the model parameter K
es
and P
es
factors are
given in Table 4. Especially higher P
es
values were needed to calibrate
for the erosion rates in the catchment of Macieira de Alcôba. For the
observed values of the burnt scenario, three post-fire years for the
outlet (average of 0.19 ton ha
1
yr
1
) and a relatively high value for
the hillslope erosion (average 27.40 ton ha
1
yr
1
) for the average of
two post-fire years could explain the differences between the lower
K
es
and higher P
es
(Table 4): sediments are deposited before they
reach the stream. For the unburnt scenario a similar trend is shown,
although K
es
values were slightly lower. Model parameters
m(discharge exponent), n(slope exponent), and p(convergence factor)
were adapted to 1.35, 2, and 6, respectively.
3.2 |Effects of wildfires on erosion and sediment
deposition
Figure 6a shows the total cumulative erosion and deposition at the
end of the 41-year model simulation. Eroded sediments are mostly
deposited in larger streams and do not reach the catchment outlet. As
soon as valley bottoms start to get wider and slopes become less
steep, an increase in sedimentation occurs. However, a gradual
decrease of deposition can be seen (e.g. at the lower elevated area in
the western part of the catchment; the colour in Figure 6a shifts from
FIGURE 5 Calibrated versus
observed erosion values. Red dotted line
is 1:1 line.
TABLE 4 Calibrated LAPSUS parameter values
Burnt scenario Unburnt scenario
m=discharge exponent 1.35 1.35
n=slope exponent 2 2
p=convergence factor 6 6
K
es
P
es
K
es
P
es
Lumped K
es
and P
es
factor 6.5 10
6
1.0 10
1
4.0 10
6
1.0 10
1
Urban
Unburnt, low, moderate and high severity 3.25 10
6
5.0 10
2
NA NA
Unburnt NA NA 2.0 10
6
5.0 10
2
Agriculture (winter pasture)
Unburnt 9.75 10
6
1.5 10
1
2.0 10
6
1.5 10
1
Low severity and regrowth 4.241 10
5
1.5 10
1
NA NA
Moderate severity 4.3875 10
4
1.5 10
1
NA NA
High severity 1.78571 10
3
1.5 10
1
NA NA
Eucalypt, Pine, Other broadleaf forest, Shrub vegetation
Unburnt 6.5 10
6
1.0 10
1
4.0 10
6
1.0 10
1
Low severity and regrowth 2.828 10
5
1.0 10
1
NA NA
Moderate severity 2.925 10
4
1.0 10
1
NA NA
High severity 1.19048 10
3
1.0 10
1
NA NA
8FOLLMI ET AL.
green to yellow at the end of the main streams). The highest erosion
rates were simulated in and along the stream network; however, the
largest areas with high erosion rates occur in the north-eastern and
north-western parts of the catchment, which were the areas with the
highest fire frequency (Figure 6b). These are also the areas with the
shallower soil types in the catchment, and hence more vulnerable to
erosion (Tavares Wahren et al., 2016).
Figures 6c and dshow the difference between the burnt and
unburnt scenarios for erosion and deposition, respectively, and thus
show the potential impact of wildfire on erosion and deposition.
Increased erosion due to wildfires is most evident in the central south-
ern area of the catchment and on the catchment borders in the north-
west and north-east: locations that exhibited high wildfire recurrence
(Figure 6b). These are also the areas which showed higher erosion
values. Increased deposition due to wildfires was especially simulated
in the main streams (Figure 6d), although there were small spots with
lower sedimentation values for the burnt scenario, mostly in locations
where several streams merge.
The net erosion simulated by LAPSUS for the burnt scenario was
5.95 ton ha
1
yr
1
on average, whereas for the unburnt scenario the
overall net erosion is one order of magnitude lower: 0.58 ton
ha
1
yr
1
. This is mostly due to much higher values of total erosion
for the burnt scenario: 16.26 ton ha
1
yr
1
for burnt vs 3.46 ton
ha
1
yr
1
for unburnt. There was also an increase in deposition:
10.31 ton ha
1
yr
1
for burnt vs 2.88 ton ha
1
yr
1
for unburnt,
meaning that a large part of the additional erosion due to wildfire
occurrence redeposited downstream, in non-burnt areas. However,
the sediment delivery ratio (SDR =net erosion/total erosion) for the
burnt scenario was much higher (36.6%) than for the unburnt scenario
(16.8%). Thus, more sediment reaches a stream and finally the outlet
in a wildfire-affected catchment.
Figure 7a shows the time series of cumulative sediment yield
(i.e. net erosion) for all 41 hydrological years in the
´
Agueda catchment,
combined with wildfire size and rainfall. Figure 7b shows, for each fire
year, the contribution of fire severity in terms of area burned. In gen-
eral, fire occurrence led to rapid increases of simulated net erosion
and total erosion. This increase was particularly evident in the first
post-fire year, when compared with the second and third post-fire
years.
Model results suggest that the main erosion driver in this region
was wildfire occurrence, with rainfall variability as secondary driver.
For the unburnt scenario a clear positive relationship between rainfall
and erosion rates was found (R
2
=0.85, p< 0.01), where for the burnt
scenario a positive relationship also existed, but with a weak and not
significant correlation (R
2
=0.06, p> 0.1; Figure 8). The relatively
large extent of (especially) moderate and high burn-severity fires in
2013, 2016, and 2017 (Figure 7b) leads to relatively large increase in
net erosion (Figure 7a), except for 2016, which can be explained by
2016 being an exceptionally dry post-fire year. Vice-versa, in 1995 a
steep increase in simulated net erosion was simulated (Figure 7a),
FIGURE 6 (a) Total cumulative erosion and deposition (m) after the 41-year LAPSUS simulation. Note that erosion is given in negative values
and deposition in positive values. (b) Fire recurrence based on the dNBR analysis. (c) Difference in total erosion (burnt minus unburnt scenarios).
(d) Difference in total deposition (burnt minus unburnt scenarios). For both (c) and (d) red colours implicate an increase in erosion or decrease in
deposition, green colours implicate an increase in deposition or a decrease in erosion.
FOLLMI ET AL.9
while the extent of the fire was not so large. This can be explained by
1995 being a very wet post-fire year, still causing relatively much
erosion. Thus, primarily fire extent and severity, with in addition
rainfall, drive episodes of increased catchment erosion.
3.3 |Spatial patterns of erosion and deposition
For the burnt scenario the land use with the largest contribution to
erosion in the catchment was ‘Eucalypt’(36.9%; 1.78 ton ha
1
yr
1
),
FIGURE 7 (a) Time series of simulated net
erosion for the burnt and unburnt scenarios,
including rainfall (top blue bars) and fire
occurrence (vertical lines). (b) Fire extent (ha) for
each fire year, divided into low, moderate, and
high severity.
FIGURE 8 Rainfall versus total
erosion relationship for the burnt and
unburnt scenarios. A weaker correlation
for the burnt scenario can be seen, as
post-fire years do not follow the trendline.
10 FOLLMI ET AL.
followed by ‘Other Broadleaf’(19.6%; 1.64 ton ha
1
yr
1
), ‘Shrubs’
(19.5%, 5.46 ton ha
1
yr
1
), ‘Pine’(16.8%, 1.76 ton ha
1
yr
1
), and
‘Agriculture’(7.1%, 0.20 ton ha
1
yr
1
) (Figures 9a and b). High values
for Eucalypt can be associated with a high land occupation (44.8%).
By contrast, for ‘Shrubs’a low occupation (12.0%) still resulted in a
relatively high contribution for erosion (19.5%). An explanation for
this is that natural vegetation is twice as often burnt as other vegeta-
tion types. Average wildfire recurrences for the 41-year simulation
were 0.42, 1.00, 1.08–1.11, and 2.02 times for ‘Agriculture’,‘Pine’,
‘Other Broadleaf’,‘Eucalypt’, and ‘Shrubs’, respectively.
In addition, the high contribution to erosion of ‘Other Broadleaf’
(19.6%), but low catchment occupation (7.8%), can be related to
broadleaf tree species growing in riparian areas, where localized large
erosion rates are caused by channel erosion processes (Figure 9b). To
verify this, we analysed the contribution of the ‘Other Broadleaf’
occurring on hillslopes and the ‘Other Broadleaf’occurring in riparian
zones separately: ‘Other Broadleaf (Riparian)’contributed almost
three times more to erosion than ‘Other Broadleaf (Hillslope)’(14.5%
vs 5.2%) (Figure 9a). For similar reasons, ‘Other Broadleaf’
contributed to a large extent (35.5%) to the total deposition in the
catchment, where especially riparian areas deposited six times more
than hillslope areas (30.7% vs 4.9%).
Median erosion rates per land use for the burnt scenario are
shown in Figures 9b and cfor the 41-year simulation period and only
for the post-fire years, respectively. Figure 9b (and to a lesser extent
Figure 9c) highlights that median erosion rates are particularly high for
‘Other Broadleaf (Hillslope)’,‘Shrubs’, and to a lesser extent ‘Euca-
lypt’and ‘Pine’. This could be linked to the fact that both these
landcovers normally had higher burn severity than the others, and also
the higher burn frequency of ‘Shrubs’. A smaller factor could be the
higher runoff values, depicted in Figure 9b (blue bars), and computed
as the median runoff per land use type over the 41-year simulation
period. Higher values thus mean that more runoff occurs in these land
use types, which could contribute to higher erosion rates. The erosion
rates show a high variability, as shown by a large difference between
the mean and the median, and a high standard deviation and skew-
ness (Table 5). This variability is particularly high for ‘Other Broadleaf’,
again determined by their presence in the riparian area.
FIGURE 9 (a) Total percentages of
erosion (red bars), deposition (blue bars),
and percentage of total catchment area
per land use (grey bars) for the burnt
scenario. (b) Median erosion rates per
land use for the burnt scenario (red bars)
and median accumulated runoff per land
use type, over the 41-year modelled
timespan (blue bars). (c) Median erosion
rates per land use for the pre-fire year
and three post-fire years.
FOLLMI ET AL.11
In addition to this, for all land uses within the first post-fire year,
erosion values were generally two orders of magnitude higher than
pre-fire erosion rates (Figure 9c) and erosion values generally dropped
by 50% and 70% in the second and third post-fire years compared
to the first post-fire year; this decline was more evident in ‘Eucalypt’
and ‘Shrubs’than in other land covers. This indicates that especially
the first post-fire year has the most impact.
Wildfire recurrency also had a large impact on simulated erosion
rates. The simulated median erosion rate in almost all land uses was
one order of magnitude larger in areas that burnt one or two times
compared to areas that did not burn (Table 5). For three or four times
burnt areas, the median erosion rates might even be two orders of
magnitude higher. For deposition, median values were almost zero, as
it was concentrated in small areas. Mean deposition increased with
burn frequency for ‘Other Broadleaf (Riparian)’and urban areas
(i.e. infrastructure), and to a lesser degree for ‘Agriculture’until a limit
of two times burnt. Note that there was a large variation in the
surface area affected by a certain fire recurrence and its related
characteristics (e.g. slope steepness).
4|DISCUSSION
This study investigated the impact of multiple wildfires on erosion
patterns over multiple decades in a large catchment (404 km
2
). To our
knowledge this is the first study investigating post-fire erosion at such
a large scale, both spatially and temporally in a Mediterranean
country. The results indicate that eight major wildfires in the
´
Agueda
catchment during 41 years led to a significant increase in erosion,
compared to a situation without wildfire occurrence.
4.1 |Model performance
First of all, model calibration was limited by the scarcity of spatially
and temporally explicit measured data. The dataset used for calibra-
tion consisted only of plot studies, hillslopes, (unburnt) agricultural
slopes, and the headwater catchment of Macieira de Alcôba (94 ha),
which is much smaller than the
´
Agueda catchment (404 km
2
)
(see Supplementary Material S3). Moreover, the few observed years
in the dataset made a split-sample approach difficult (i.e. execute both
calibration and validation), so that only calibration was possible. This
data scarcity is quite common for long-term (historical) time scales
and is a problem in many long-term modelling studies (Batista
et al., 2019). It is also common in post-fire erosion modelling (Lopes
et al., 2021), where the unpredictable occurrence and short nature of
wildfire disturbances limits data collection immediately after the wild-
fire. Nevertheless, both studies emphasize the importance of spatial
and temporal data for calibration and validation, and therefore future
studies of this nature might focus on areas encompassing a larger
number of studies in plots, hillslopes, and small catchments, or where
sediment yield data is available.
Furthermore, LAPSUS is an annual-based model that does not
include individual storm events. This might be a disadvantage, since
sediment yield can be highly variable in space and time and can be
dependent on a few single intra-annual rainfall events (Wu, Baartman
& Nunes, 2021; Wu et al., 2021) (time compression, as discussed by
Smetanová et al., 2019). However, it should be noted that excessive
model complexity in the description of temporal patterns might com-
plicate the prediction of long-term erosion patterns due to data scar-
city and accumulation of errors; Baartman et al. (2020) stress that
model complexity should be adequate to the complexity of the
TABLE 5 Erosion and deposition rates per land use for areas that encountered different fire recurrences during the 41-year investigation
period. The two rightmost columns indicate the standard deviation and skewness per land use
Mean (median) erosion (ton ha
1
yr
1
)
Not burnt 1burnt 2burnt 3burnt 4burnt Standard deviation Skewness
Urban 0.31 (0.10) 0.54 (0.05) 1.33 (0.09) 0.18 (0.00) 2.28 (0.00) 2.44 0.313
Agricultural 2.29 (0.11) 9.12 (0.58) 13.22 (1.84) 16.14 (2.41) 33.49 (7.74) 30.50 0.507
Eucalypt 1.02 (0.19) 10.39 (2.21) 13.60 (3.74) 24.69 (8.02) 15.98 (8.26) 40.44 0.593
Pine 0.75 (0.14) 9.53 (2.42) 21.69 (7.07) 33.32 (10.94) 17.52 (6.54) 42.84 0.615
Shrub vegetation 2.36 (0.57) 12.55 (3.16) 18.88 (5.94) 27.10 (8.80) 43.18 (12.10) 67.53 0.713
Other Broadleaf 5.15 (0.25) 23.84 (2.51) 30.85 (4.62) 40.32 (11.68) 130.25 (21.53) 83.87 0.740
Broadleaf (Riparian) 5.96 (0.23) 28.04 (1.79) 35.39 (1.70) 46.20 (4.79) 229.08 (37.40) 87.04 0.744
Broadleaf (Hillslope) 1.02 (0.28) 10.85 (4.04) 15.18 (8.52) 20.28 (12.68) 27.20 (12.98) 67.48 0.667
Mean deposition (ton ha
1
yr
1
)
No fire 1burnt 2burnt 3burnt 4burnt Standard deviation Skewness
Urban 0.99 0.58 15.88 44.22 34.93 19.16 0.183
Agriculture 7.83 9.67 17.85 6.05 3.60 58.56 0.459
Eucalypt 2.02 3.64 3.20 3.33 0.00 41.20 0.232
Pine 1.61 2.18 2.44 2.93 2.12 28.37 0.223
Shrub vegetation 2.13 2.79 3.93 3.64 3.36 36.68 0.273
Other broadleaf 24.03 20.45 24.21 11.65 4.61 108.17 0.591
Broadleaf (Riparian) 28.47 27.30 33.54 48.45 0.74 126.14 0.678
Broadleaf (Hillslope) 2.08 2.82 8.24 3.09 7.44 46.19 0.271
12 FOLLMI ET AL.
systems under study. The limited validation and lumped temporal
representation make the LAPSUS model results more suitable to show
the spatial patterns of erosion and deposition, and give an indication
of the order of magnitude of their rates, than to make precise
predictions.
Despite these limitations, almost all calibrated erosion rates were
of the same order of magnitude as those observed in the
´
Agueda
catchment at the plot (Ferreira, 1997; Shakesby et al., 1994) and small
catchment scales (Nunes, Doerr, et al., 2018), as shown in Supplemen-
tary Material S3 and Figure 5. Furthermore, our results were within
the range of multiple Mediterranean plot studies reviewed by
Shakesby (2011), although slightly at the ‘high end’of observations:
the median first post-fire year erosion rates of 39.2 ton ha
1
yr
1
in
this study fall inside the range 1–50 ton ha
1
yr
1
for moderate to
high burn-severity wildfires, with a median of 3.7 ton ha
1
yr
1
. Local
characteristics of the
´
Agueda catchment could be behind the higher
values, especially the higher rainfall compared with other Mediterra-
nean regions, and the large cell sizes of the LAPSUS model (25 25
m) compared to the shorter length of typical plot studies.
Moreover, the median long-term erosion rates found in this study
for shrublands (5.46 ton ha
1
yr
1
) correspond with those found in
north-west Spain using
137
Cs inventories to assess erosion at burnt
shrub-covered hillslopes for a 50-year period, ranging from 6.6 to 6.7
ton ha
1
yr
1
(Menéndez-Duarte et al., 2009). The erosion values
found in this study seem therefore to broadly agree with those
reported elsewhere in terms of order of magnitude.
It should be noted that the model might also be limited due to the
poor representation of several processes in the
´
Agueda catchment;
streambank erosion, soil renewal, and tillage erosion could all be rele-
vant. For streambank erosion, the lack of sediment yield data at the
catchment scale prevented the determination of different erosivity
parameters for streams in LAPSUS (Baartman, van Gorp, et al., 2012),
and existing stone walls in streams were not accounted for. Soil
renewal was also not accounted for, but it can be argued that these
values are negligible at the 41-year time scale due to the low soil
renewal rates in the Mediterranean region (around 0.1 ton ha
1
yr
1
estimated by Alexander, 1985). Finally, tillage in the managed planta-
tion forests of Eucalypt and Pine might be significant due to the short
rotation cycles of 10–12 years (Kardell et al., 1986; Shakesby, 2011),
estimated at 10 ton ha
1
after each treatment in central Portugal
(Govers et al., 1996; Shakesby, 2011); again, in the long-term these
values are relatively low compared with water erosion, which
concurs with the findings of Baartman, Temme, et al. (2012) for
millennial time scales.
4.2 |Sediment dynamics at catchment scale
One of the main findings is the high spatial variability of erosion rates,
with a high skewness (and a mean one or two orders of magnitude
higher than the median) indicating a concentration of erosion in small
areas. This can be attributed to different processes operating at
different scales, and concurs with the large difference between
measurements at the point, plot, or catchment scale mentioned by
Shakesby (2011).
The areas of concentrated erosion in the
´
Agueda catchment often
have high runoff accumulation, being therefore subjected to gully
formation or, at larger scales, streambank erosion (Figure 6a). This
concentration of erosion is common to many large catchments, as
described for example by de Vente and Poesen (2005) and Smetanová
et al. (2019). In burnt areas, topographic features determining runoff
concentration overlap with the spatial distribution of wildfire severity
to determine erosion ‘hotspots’, as observed by Fernández et al.
(2020), who found a good relation between sediment yield and topo-
graphic connectivity modified by burn severity. However, at larger
scales, the role of sediment sinks (footslope and floodplain sediment
storage) tends to become increasingly important, lowering the sedi-
ment yield (de Vente & Poesen, 2005). This has also been proposed
for burnt areas by Ferreira et al. (2008), who associated it with a
decreasing sediment transport capacity by runoff and streamflow. In
the
´
Agueda catchment, LAPSUS indicates that erosion is particularly
high in some part of the streams or at locations where streams are
originating and the formation of rills, gullies, and streams starts, while
deposition occurs particularly in and around the larger streams/rivers
(Figure 6a). Model outputs therefore fit within the current knowledge
on catchment sediment dynamics.
In general terms, LAPSUS indicates that burnt gullies and streams
function as a source of sediment, while footslopes and floodplains
function as sinks. Sediments deposited at footslopes and floodplains
can subsequently be re-entrained and transported further down-
stream. Model results indicate that wildfires can change the balance
between these factors, as the increase in erosion is larger than the
increase in deposition, resulting in a higher sediment delivery ratio
and more sediments reaching the outlet.
There is, however, variability in erosion and sediment deposition
within the stream network of the
´
Agueda catchment (Figure 6),
possibly caused by differences in the connectivity of the stream
with upslope locations (Borselli et al., 2008; Heckmann &
Schwanghart, 2013). Despite this variability, deposition tends to
decrease towards the outlet of the catchment, suggesting a sediment
build-up in the larger streams. This process could foster ‘sediment
cascades’that can behave as ‘jerky conveyer belts’towards the
stream outlet (Cossart et al., 2018; Schoorl et al., 2014), which has
also been observed in other burned catchments (Inbar et al., 1998;
Keizer et al., 2015; Mayor et al., 2007).
4.3 |The significance of post-fire erosion in the
´
Agueda catchment
The results indicate that wildfires cause long-term erosion and
sediment yield in the
´
Agueda catchment to increase by at least one
order of magnitude when compared to natural (i.e. non-disturbed)
erosion and sediment yield (Table 5and Figures 6c and d,7).
As shown in Figure 8, the relation between rainfall and simulated
total erosion was relatively strong, but decreased when wildfires were
included in the simulations. This indicates that, when a wildfire occurs,
it clearly dominates the effects of rainfall on erosion. This relationship
is influenced by wildfire size, and especially burn severity for the fires
in 2013 and 2017. When compared with other fires, the larger extent
of high burn severity associated with these wildfires corresponds to a
larger increase of net erosion, total erosion, and deposition rates for
the whole catchment, especially in the first post-fire year (Figure 7).
There were relatively high erosion values even in the second and third
FOLLMI ET AL.13
post-fire years, which did not occur after smaller and less severe
wildfires, which was possibly caused by a slower regeneration of the
vegetation or soil (Maia et al., 2012). Interestingly, this difference in
sediment processes persists after the wildfire disturbance is negated
by vegetation regeneration, as sediment eroded from burnt areas
and deposited in the first post-fire years can be available for re-
entrainment in years without wildfire-enhanced erosion. This process
corresponds to earlier research (Inbar et al., 1998; Mayor et al., 2007;
Wittenberg & Inbar, 2009), although in the
´
Agueda headwater catch-
ments Nunes et al. (2020) found that the large rainfall rates can flush
these sediments in a few years.
Wildfire-enhanced erosion rates in the
´
Agueda can lead to soil
degradation. Median values in pre-fire years are under a tolerable rate
of 1 ton ha
1
yr
1
(Verheijen et al., 2012), but estimated erosion rates
for all three years after each fire surpass this threshold by one or two
orders of magnitude. Soil depth loss can be especially severe given
the generally shallow soils in the
´
Agueda catchment (Tavares Wahren
et al., 2016); roughly 8% of the catchment showed a reduction of
10 cm soil depth after 41 years. Moreover, erosion can lead to a
decrease in soil fertility (Bakker et al., 2004) through the selective loss
of carbon and nutrients (Hosseini et al., 2017; Serpa et al., 2020). Even
though our simulated erosion rates are uncertain due to calibration
difficulties (see Section 4.1), this concurs with the generally degraded
condition of mountain soils in this study area (Tavares Wahren
et al., 2016), many of which are already unsuited for Eucalypt planta-
tion and support less profitable Pine plantations. The effects of
repeated wildfires on soil quality have also been observed elsewhere
in the Mediterranean (Carreira et al., 1996; Hosseini et al., 2016).
Furthermore, the higher connectivity caused by the wildfires can
also have implications for water quality. Ashes and fine sediments
produced by wildfires can impact ecosystems and limit human uses,
due to associated contaminants such as polycyclic aromatic hydrocar-
bons, heavy metals, and nutrients (Nunes, Naranjo Quintanilla,
et al., 2018; Serpa et al., 2020). While the increase in net erosion can
indicate potential problems in the immediate years after wildfires, the
increase in background sediment concentrations can indicate persis-
tent contamination problems, especially where related to toxic com-
pounds which can be problematic in small amounts and accumulate
over time.
4.4 |Contribution of land use to erosion patterns
The results indicate that the most important plantation tree species,
Eucalypt and Pine, are the main contributors to the overall sediment
budget in the
´
Agueda catchment (Figure 9a). However, mean erosion
rates for natural vegetation (predominantly consisting of shrub vege-
tation) are much higher than those for other species (Table 5and
Figures 9b and c). Several factors contribute to these differences.
First, natural vegetation is normally located at the ridges of hills or
summits of small mountains, with comparatively shallow soils with
lower water-holding capacity (Tavares Wahren et al., 2016) and
steeper slopes. Even without wildfire, simulated median erosion rates
under natural vegetation are 3 and 4.1 times higher than those of
Eucalypt and Pine, respectively (0.57 vs 0.19 and 0.14 ton ha
1
yr
1
).
In addition, shrub areas tend to be more affected by high wildfire
severity: 39%, compared with 17% for Eucalypt and 26% for Pine,
respectively, making these areas more susceptible to post-fire erosion.
This would expose natural vegetation to post-fire erosion more fre-
quently, although it tends to recover faster than other vegetation
types: while erosion rates in the second and third post-fire years
decreased by 71.1 and 95.4% compared with the first post-fire year,
the decrease rates are 56.7/84.2% for Pine and 64.3/84.2% for Other
Broadleaf. Eucalypt also seems to recover relatively fast, as rates
decrease by 82.8/95.3% in the second and third post-fire years.
Areas with other broadleaf species showed much higher spatial
variability, with erosion rates being determined by a low number of
cells with relatively high erosion values. This can be explained by the
location of part of these forests in riparian areas with higher runoff
accumulation, and therefore susceptible to channel bank erosion but
also to streambed deposition. This spatial variability is therefore a
result of the association between these species and channel
processes.
Finally, ‘Agriculture’and ‘Urban’had lower erosion rates than
forest land uses, even though ‘Agriculture’is parameterized as more
susceptible to erosion (higher K
es
in Table 4). Note, however, that
erosion rates in agricultural land could only be calibrated for unburnt
situations, not for burnt situations (see Supplementary Material S3).
Agricultural areas are usually located on footslopes, while forests are
located on steeper slopes (Tavares Wahren et al., 2016), as is common
in most catchments; crop fields located on steeper slopes are usually
protected by terraces (Nunes, Bernard-Jannin, et al., 2018). Therefore,
even though agriculture is commonly perceived to have a larger
impact on erosion (e.g. Cerdan et al., 2010), our results indicate that
for the
´
Agueda catchment, high erosion rates predominantly occurred
in forest or naturally vegetated areas, thus contradicting the assertion
by Shakesby (2011) that in the long-term, erosion rates in burnt
Mediterranean areas are below those of croplands.
These results complicate a direct assessment of the land use most
responsible for erosion based solely on its characteristics, as the
relation between both variables at the
´
Agueda catchment scale seems
to be related to wildfire occurrence and recurrence, or their topo-
graphic position, rather than local-scale processes such as soil water
repellency or soil cover by needle cast (Benito et al., 2003; Shakesby
et al., 1994; Walsh et al., 1994). This concurs with the conclusions of
Ferreira et al. (2008) that hydrological and sediment connectivity,
and their impact by wildfire, drive burnt area erosion response at
wider scales.
4.5 |Challenges for land management in the future
Increasing global change could further impact ecosystem services and
intensify soil degradation and biodiversity loss, which is detrimental to
food and water security (Certini, 2005). In fact, post-fire erosion in
the region of north-central Portugal is likely to increase, due to inten-
sified rainfall during the winter months associated with a more intense
wildfire regime (Carvalho-Santos et al., 2019; Pastor et al., 2019).
Likewise, the wildfires in the last decade (2013, 2016, and 2017) are
larger in size and had a larger impact due to higher burn severity. This
is partly due to socio-economic drivers such as rural depopulation and
(agricultural) land abandonment, leading to encroachment and foster-
ing fuel load build-up, which is perceived to have led to more wildfires
(Llovet et al., 2009; Shakesby, 2011).
14 FOLLMI ET AL.
Areas of natural vegetation were burnt to a larger extent, higher
severity, and shorter recurrence times than other forest land uses.
However, burnt natural vegetation areas in the north-east of the
´
Agueda catchment received less attention in terms of pre- and post-
fire management. Unlike for natural vegetation, plantations have
economic value and are therefore better managed, which fits into the
ongoing management paradigm to organize and support productive
forests and leave natural areas outside the management structure
(Shakesby, 2011). Managing natural areas is, however, important to
limit wildfire occurrence, soil erosion, and further degradation. More-
over, better management can also decrease toxic or contaminated
ashes that are transported downstream (Nunes, Naranjo Quintanilla,
et al., 2018). Hence, managing scrub encroachment in the north-east
of the
´
Agueda catchment can have positive consequences for natural
areas, as well as for downstream areas. Approaches can include low
investment options such as prescribed burning to lower the fuel load
(Khabarov et al., 2016; Shakesby, 2011), or more conventional but
expensive options such as grazing or wildfire breaks (Raftoyannis
et al., 2014). Similar options can support managing fuel loads in
plantations, together with decreasing tree density (Raftoyannis
et al., 2014), although plantation densities in the
´
Agueda are already
relatively low. The results of this work can help identify priority
management areas and can support post-fire management with
emergency post-fire control by mulching (Keizer et al., 2018; Prats
et al., 2012), allowing the direction of these measures to areas with
greater susceptibility to erosion.
5|CONCLUSIONS
To the best of our knowledge, this is the first study to quantify post-
fire erosion and map sediment transport and deposition patterns at
long temporal (41 years) and large spatial (404 km
2
) scales. Model
calibration was considered good, but lack of data—common in post-
fire erosion assessments—limited a full validation of the LAPSUS
model. We found that wildfires significantly increased erosion in the
´
Agueda catchment in the last 41 years compared to general back-
ground erosion processes by at least one order of magnitude. The first
post-fire year had a substantial contribution to erosion, and the post-
fire erosion increased with burn severity and wildfire recurrence.
Wildfires also amplified the background sediment yield in non-post-
fire years, due to the increased availability of sediment build-up in the
catchment. Simulated post-fire erosion rates varied spatially, and
showed a large skewness, indicating that erosion is concentrated in
well-connected areas, either due to topographic or burn severity
patterns. Median erosion values were within the range of those
reported in the literature for plot studies, although close to the higher
end: 5.95 ton ha
1
yr
1
in the long term, which surpasses a sustain-
able threshold of 1 ton ha
1
yr
1
, indicating the potential for long-
term soil degradation. While there was differentiation of erosion rates
between land uses, this was more related to their topographic
position and susceptibility to wildfire (both wildfire severity and recur-
rence) than their intrinsic characteristics, pointing out the importance
of sediment connectivity—and in this case, the changes to connectiv-
ity caused by wildfire disturbances—for erosion patterns at larger
scales. Finally, we think that due to their spatial outputs, reduced-
complexity models such as LAPSUS can function as a tool to identify
locations at erosion risk that can be adopted by land managers in
pre-fire and post-fire management strategies.
ACKNOWLEDGEMENTS
This work was co-funded by a travel fellowship attributed to D. Follmi
through the Erasmus+programme of the European Union. Additional
funding was provided by the Portuguese Foundation for Science and
Technology, through project FRISCO (PCIF/MPG/0044/2018),
individual grants attributed to J.P. Nunes (IF/00586/2015) and
A. Benali (CEECIND/03799/2018/CP1563/CT0003), and the CE3C
research centre (UIDB/00329/2020).
AUTHOR CONTRIBUTIONS
Dante Follmi: conceptualization; methodology; investigation (e.g. data
collection); writing –initial draft; writing –reviewing and editing.
Jantiene Baartman: conceptualization; methodology; software (its
provision and development); supervision; writing –reviewing and
editing. Akli Benali: conceptualization; methodology; resources
(provision of data, etc.); supervision; writing –reviewing and editing.
Joao Pedro Nunes: conceptualization; funding acquisition; methodol-
ogy; resources (provision of data, etc.); supervision; writing –
reviewing and editing.
DATA AVAILABILITY STATEMENT
The input and calibration data used in this study is not our own and
the source is referred to in the appropriate sections. The output data
is presented in maps and tables in the text. We do not intend to
share these results further. Interested readers may contact the
corresponding author for enquiries.
ORCID
Jantiene Baartman https://orcid.org/0000-0001-6051-8619
Joao Pedro Nunes https://orcid.org/0000-0002-0164-249X
REFERENCES
Alexander, E.B. (1985) Rates of soil formation from bedrock or consoli-
dated sediments. Physical Geography, 6(1), 25–42. Available from:
https://doi.org/10.1080/02723646.1985.10642261
Arnold, J.G., Allen, P.M., Muttiah, R. & Bernhardt, G. (1995) Automated
base flow separation and recession analysis techniques. Groundwater,
33(6), 1010–1018. Available from: https://doi.org/10.1111/j.1745-
6584.1995.tb00046.x
Baartman, J.E.M., Melsen, L.A., Moore, D. & van der Ploeg, M.J. (2020) On
the complexity of model complexity: Viewpoints across the
geosciences. Catena, 186, 104261. Available from: https://doi.org/
10.1016/j.catena.2019.104261
Baartman, J.E.M., Temme, A.J.A.M., Schoorl, J.M., Braakhekke, M.H.A. &
Veldkamp, T.A. (2012) Did tillage erosion play a role in millennial
scale landscape development? Earth Surface Processes and Landforms,
37(15), 1615–1626. Available from: https://doi.org/10.1002/esp.
3262
Baartman, J.E.M., van Gorp, W., Temme, A.J.A.M. & Schoorl, J.M. (2012)
Modelling sediment dynamics due to hillslope–river interactions:
Incorporating fluvial behaviour in landscape evolution model
LAPSUS. Earth Surface Processes and Landforms, 37(9), 923–935.
Available from: https://doi.org/10.1002/esp.3208
Bakker, M.M., Govers, G. & Rounsevell, M.D.A. (2004) The crop
productivity–erosion relationship: An analysis based on experimental
work. Catena, 57(1), 55–76. Available from: https://doi.org/10.
1016/j.catena.2003.07.002
Batista, P.V.G., Davies, J., Silva, M.L.N. & Quinton, J.N. (2019) On the
evaluation of soil erosion models: Are we doing enough?
FOLLMI ET AL.15
Earth-Science Reviews, 197, 102898. Available from: https://doi.org/
10.1016/j.earscirev.2019.102898
Benito, E., Santiago, J.L., de Blas, E. & Varela, M.E. (2003) Deforestation of
water-repellent soils in Galicia (NW Spain): Effects on surface runoff
and erosion under simulated rainfall. Earth Surface Processes and
Landforms, 28(2), 145–155. Available from: https://doi.org/10.1002/
esp.431
Boonstra, J. (1994) Estimating peak runoff rates. In: Ritzema, H.P. (Ed.)
Drainage Principles and Applications. Wageningen: International
Institute for Land Reclamation and Improvement.
Borrelli, P., Robinson, D.A., Fleischer, L.R., Lugato, E., Ballabio, C.,
Alewell, C. et al. (2017) An assessment of the global impact of 21st
century land use change on soil erosion. Nature Communications,
8(1), 2013. Available from: https://doi.org/10.1038/s41467-017-
02142-7
Borselli, L., Cassi, P. & Torri, D. (2008) Prolegomena to sediment and flow
connectivity in the landscape: A GIS and field numerical assessment.
Catena, 75(3), 268–277. Available from: https://doi.org/10.1016/j.
catena.2008.07.006
Calheiros, T., Nunes, J.P. & Pereira, M.G. (2020) Recent evolution of spatial
and temporal patterns of burnt areas and fire weather risk in the Ibe-
rian Peninsula. Agricultural and Forest Meteorology, 287, 107923.
Available from: https://doi.org/10.1016/j.agrformet.2020.107923
Campo, J., Andreu, V., Gimeno-García, E., González, O. & Rubio, J.L. (2006)
Occurrence of soil erosion after repeated experimental fires in a
Mediterranean environment. Geomorphology, 82(3-4), 376-387.
Available from: https://doi.org/10.1016/j.geomorph.2006.05.014
Carreira, J.A., Arevalo, J.R. & Niell, F.X. (1996) Soil degradation and
nutrient availability in fire-prone Mediterranean shrublands of south-
eastern Spain. Arid Soil Research and Rehabilitation, 10(1), 53–64.
Available from: https://doi.org/10.1080/15324989609381420
Carvalho-Santos, C., Marcos, B., Nunes, J., Regos, A., Palazzi, E.,
Terzago, S. et al. (2019) Hydrological impacts of large fires and future
climate: Modeling approach supported by satellite data. Remote
Sensing, 11(23), 2832. Available from: https://doi.org/10.3390/
rs11232832
Carvalho-Santos, C., Nunes, J.P., Monteiro, A.T., Hein, L. & Honrado, J.P.
(2016) Assessing the effects of land cover and future climate condi-
tions on the provision of hydrological services in a medium-sized
watershed of Portugal. Hydrological Processes, 30(5), 720–738.
Available from: https://doi.org/10.1002/hyp.10621
Cerdan, O., Govers, G., le Bissonnais, Y., van Oost, K., Poesen, J., Saby, N.
et al. (2010) Rates and spatial variations of soil erosion in Europe: A
study based on erosion plot data. Geomorphology, 122(1), 167–177.
Available from: https://doi.org/10.1016/j.geomorph.2010.06.011
Certini, G. (2005) Effects of fire on properties of forest soils: A review.
Oecologia, 143(1), 1–10. Available from: https://doi.org/10.1007/
s00442-004-1788-8
Cossart, E., Viel, V., Lissak, C., Reulier, R., Fressard, M. & Delahaye, D.
(2018) How might sediment connectivity change in space and time?
Land Degradation & Development, 29(8), 2595–2613. Available from:
https://doi.org/10.1002/ldr.3022
de Vente, J. & Poesen, J. (2005) Predicting soil erosion and sediment yield
at the basin scale: Scale issues and semi-quantitative models. Earth-
Science Reviews, 71(1), 95–125. Available from: https://doi.org/10.
1016/j.earscirev.2005.02.002
Delestre, O., Darboux, F., James, F., Lucas, C., Laguerre, C. & Cordier, S.
(2017) FullSWOF: Full shallow-water equations for overland flow.
The Journal of Open Source Software, 2(20), 448. Available from:
https://doi.org/10.21105/joss.00448
ESPA-USGS. (2020) EROS Science Processing Architecture on Demand
Interface. United States Geological Survey. espa.cr.usgs.gov/
ordering/new/
Esteves, T.C.J., Kirkby, M.J., Shakesby, R.A., Ferreira, A.J.D., Soares, J.A.A.,
Irvine, B.J. et al. (2012) Mitigating land degradation caused by
wildfire: Application of the PESERA model to fire-affected sites in
Central Portugal. Geoderma, 191, 40–50. Available from: https://doi.
org/10.1016/j.geoderma.2012.01.001
Fernández, C., Fernández-Alonso, J.M. & Vega, J.A. (2020) Exploring the
effect of hydrological connectivity and soil burn severity on sediment
yield after wildfire and mulching. Land Degradation & Development,
31(13), 1611–1621. Available from: https://doi.org/10.1002/ldr.3539
Fernández, C. & Vega, J.A. (2016) Evaluation of RUSLE and PESERA
models for predicting soil erosion losses in the first year after wildfire
in NW Spain. Geoderma, 273, 64–72. Available from: https://doi.org/
10.1016/j.geoderma.2016.03.016
Fernández, C., Vega, J.A. & Vieira, D.C.S. (2010) Assessing soil erosion
after fire and rehabilitation treatments in NW Spain: Performance of
RUSLE and revised Morgan–Morgan–Finney models. Land Degrada-
tion & Development, 21(1), 58–67. Available from: https://doi.org/10.
1002/ldr.965
Ferreira, A.J.D., Coelho, C.O.A., Ritsema, C.J., Boulet, A.K. & Keizer, J.J.
(2008) Soil and water degradation processes in burned areas:
Lessons learned from a nested approach. Catena, 74(3), 273–285.
Available from: https://doi.org/10.1016/j.catena.2008.05.007
Ferreira, C. (1997) Eros˜ao hídrica em solos florestais: estudo em
povoamentos de Pinus Pinaster e Eucalyptus Globulus em macieira
de Alcôba-
´
Agueda. Revista da Faculdade de Letras: Geografia, 13,
145–244.
Foster, G., & Meyer, L. (1975) Mathematical simulation of upland erosion
by fundamental erosion mechanics. Present and prospective technol-
ogy for predicting sediment yields and sources. In Proceedings of the
Sediment-Yield Workshop, USDA Sedimentation Laboratory, Oxford,
MI. Agricultural Research Service, U.S. Department of Agriculture,
pp. 190–207.
Govers, G., Quine, T.A., Desmet, P.J.J. & Walling, D.E. (1996) The relative
contribution of soil tillage and overland flow erosion to soil redistri-
bution on agricultural land. Earth Surface Processes and Landforms,
21(10), 929–946. Available from: https://doi.org/10.1002/(SICI)
1096-9837(199610)21:10<929::AID-ESP631>3.0.CO;2-C
Hargreaves, G.H. & Samani, Z.A. (1985) Reference crop evapotranspiration
from temperature. Applied Engineering in Agriculture, 1(2), 96–99.
Available from: https://doi.org/10.13031/2013.26773
Hawtree, D., Nunes, J.P., Keizer, J.J., Jacinto, R., Santos, J., Rial-Rivas, M.E.
et al. (2015) Time series analysis of the long-term hydrologic impacts
of afforestation in the
´
Agueda watershed of north-central Portugal.
Hydrology and Earth System Sciences, 19(7), 3033–3045. Available
from: https://doi.org/10.5194/hess-19-3033-2015
Heckmann, T. & Schwanghart, W. (2013) Geomorphic coupling and
sediment connectivity in an alpine catchment –exploring sediment
cascades using graph theory. Geomorphology, 182, 89–103. Available
from: https://doi.org/10.1016/j.geomorph.2012.10.033
Holmgren, P. (1994) Multiple flow direction algorithms for runoff
modelling in grid based elevation models: An empirical evaluation.
Hydrological Processes, 8(4), 327–334. Available from: https://doi.
org/10.1002/hyp.3360080405
Hosseini, M., Geissen, V., González-Pelayo, O., Serpa, D., Machado, A.I.,
Ritsema, C. et al. (2017) Effects of fire occurrence and recurrence on
nitrogen and phosphorus losses by overland flow in maritime pine
plantations in north-central Portugal. Geoderma,289,97–106.
Available from: https://doi.org/10.1016/j.geoderma.2016.11.033
Hosseini, M., Keizer, J.J., Pelayo, O.G., Prats, S.A., Ritsema, C. &
Geissen, V. (2016) Effect of fire frequency on runoff, soil erosion,
and loss of organic matter at the micro-plot scale in north-central
Portugal. Geoderma, 269, 126–137. Available from: https://doi.org/
10.1016/j.geoderma.2016.02.004
Hosseini, M., Nunes, J.P., Pelayo, O.G., Keizer, J.J., Ritsema, C. &
Geissen, V. (2018) Developing generalized parameters for post-fire
erosion risk assessment using the revised Morgan–Morgan–Finney
model: A test for north-central Portuguese pine stands. Catena, 165,
358–368. Available from: https://doi.org/10.1016/j.catena.2018.
02.019
Hyde, K., Woods, S.W. & Donahue, J. (2007) Predicting gully rejuvenation
after wildfire using remotely sensed burn severity data. Geomorphol-
ogy, 86(3-4), 496-511. Available from: https://doi.org/10.1016/j.
geomorph.2006.10.012
Inbar, M., Tamir, M. & Wittenberg, L. (1998) Runoff and erosion processes
after a forest fire in Mount Carmel, a Mediterranean area. Geomor-
phology, 24(1), 17–33. Available from: https://doi.org/10.1016/
S0169-555X(97)00098-6
16 FOLLMI ET AL.
Jones, N., de Graaff, J., Rodrigo, I. & Duarte, F. (2011) Historical review of
land use changes in Portugal (before and after EU integration in
1986) and their implications for land degradation and conservation,
with a focus on Centro and Alentejo regions. Applied Geography,
31(3), 1036–1048. Available from: https://doi.org/10.1016/j.apgeog.
2011.01.024
Kardell, L., Steen, E. & Fabiao, A. (1986) Eucalyptus in Portugal: A threat or
a promise? Ambio, 15, 6–13.
Keizer, J.J., Martins, M.A.S., Prats, S.A., Faria, S.R., González-Pelayo, O.,
Machado, A.I. et al. (2015) Within-in flume sediment deposition in a
forested catchment following wildfire and post-fire bench terracing,
north-central Portugal. Cuadernos de Investigaci
on Geográfica, 41(1),
149–164. Available from: https://doi.org/10.18172/cig.2700
Keizer, J.J., Silva, F.C., Vieira, D.C.S., González-Pelayo, O., Campos, I.,
Vieira, A.M.D. et al. (2018) The effectiveness of two contrasting
mulch application rates to reduce post-fire erosion in a Portuguese
eucalypt plantation. Catena, 169, 21–30. Available from: https://doi.
org/10.1016/j.catena.2018.05.029
Khabarov, N., Krasovskii, A., Obersteiner, M., Swart, R., Dosio, A.,
San-Miguel-Ayanz, J. et al. (2016) Forest fires and adaptation options
in Europe. Regional Environmental Change, 16(1), 21–30. Available
from: https://doi.org/10.1007/s10113-014-0621-0
Kirkby, M. (1971) Hillslope process-response models based on the conti-
nuity equation. Transactions of the Institute of British Geographers,3,
15–30.
Kirkby, M. (1987) Modelling some influences of soil erosion, landslides and
valley gradient on drainage density and hollow development. Catena
Supplement, 10, 1–11.
Lague, D. (2014) The stream power river incision model: Evidence, theory
and beyond. Earth Surface Processes and Landforms, 39(1), 38–61.
Available from: https://doi.org/10.1002/esp.3462
Llovet, J., Ruiz-Valera, M., Josa, R. & Vallejo, V.R. (2009) Soil responses to
fire in Mediterranean forest landscapes in relation to the previous
stage of land abandonment. International Journal of Wildland Fire,
18(2), 222–232. Available from: https://doi.org/10.1071/WF07089
Lopes, A.R., Girona-García, A., Corticeiro, S., Martins, R., Keizer, J.J. &
Vieira, D.C.S. (2021) What is wrong with post-fire soil erosion model-
ling? A meta-analysis on current approaches, research gaps, and
future directions. Earth Surface Processes and Landforms, 46(1),
205–219. Available from: https://doi.org/10.1002/esp.5020
Maia, P., Pausas, J.G., Vasques, A. & Keizer, J.J. (2012) Fire severity as a
key factor in post-fire regeneration of Pinus pinaster (Ait.) in Central
Portugal. Annals of Forest Science, 69(4), 489–498. Available from:
https://doi.org/10.1007/s13595-012-0203-6
Malvar, M.C., Prats, S.A., Nunes, J.P. & Keizer, J.J. (2011) Post-fire over-
land flow generation and inter-rill erosion under simulated rainfall in
two eucalypt stands in north-central Portugal. Environmental
Research, 111(2), 222–236. Available from: https://doi.org/10.1016/
j.envres.2010.09.003
Maselli, F., Papale, D., Chiesi, M., Matteucci, G., Angeli, L., Raschi, A. &
Seufert, G. (2014) Operational monitoring of daily evapotranspiration
by the combination of MODIS NDVI and ground meteorological
data: Application and evaluation in Central Italy. Remote Sensing of
Environment, 152, 279–290. Available from: https://doi.org/10.
1016/j.rse.2014.06.021
Mataix-Solera, J., Cerdà, A., Arcenegui, V., Jordán, A. & Zavala, L.M. (2011)
Fire effects on soil aggregation: A review. Earth-Science Reviews,
109(1–2), 44–60. Available from: https://doi.org/10.1016/j.
earscirev.2011.08.002
Mateus, P. & Fernandes, P.M. (2014) Forest fires in Portugal: Dynamics,
causes and policies. In: Reboredo, F. (Ed.) Forest Context and Policies
in Portugal: Present and Future Challenges. Cham: Springer Interna-
tional, pp. 97–115.
Mayor, A.G., Bautista, S., Llovet, J. & Bellot, J. (2007) Post-fire hydrological
and erosional responses of a Mediterranean landscape: Seven years
of catchment-scale dynamics. Catena, 71(1), 68–75. Available from:
https://doi.org/10.1016/j.catena.2006.10.006
McGuire, L.A. & Youberg, A.M. (2019) Impacts of successive wildfire on
soil hydraulic properties: Implications for debris flow hazards and
system resilience. Earth Surface Processes and Landforms, 44(11),
2236–2250. Available from: https://doi.org/10.1002/esp.4632
Menéndez-Duarte, R., Fernández, S. & Soto, J. (2009) The application of
137
Cs to post-fire erosion in north-west Spain. Geoderma, 150(1),
54–63. Available from: https://doi.org/10.1016/j.geoderma.2009.
01.012
Moriondo, M., Good, P., Durao, R., Bindi, M., Giannakopoulos, C. &
Corte-Real, J. (2006) Potential impact of climate change on fire risk
in the Mediterranean area. Climate Research, 31, 85–95. Available
from: https://doi.org/10.3354/cr031085
Nathan, R.J. & McMahon, T.A. (1990) Evaluation of automated techniques
for base flow and recession analyses. Water Resources Research,
26(7), 1465–1473. Available from: https://doi.org/10.1029/
WR026i007p01465
Nunes, J.P., Bernard-Jannin, L., Rodríguez-Blanco, M.L., Boulet, A.K.,
Santos, J.M. & Keizer, J.J. (2020) Impacts of wildfire and post-fire
land management on hydrological and sediment processes in a humid
Mediterranean headwater catchment. Hydrological Processes, 34(26),
5210–5228. Available from: https://doi.org/10.1002/hyp.13926
Nunes, J.P., Doerr, S.H., Sheridan, G., Neris, J., Santín, C., Emelko, M.B.
et al. (2018) Assessing water contamination risk from vegetation
fires: Challenges, opportunities and a framework for progress. Hydro-
logical Processes, 32(5), 687–694. Available from: https://doi.org/10.
1002/hyp.11434
Nunes, J.P., Bernard-Jannin, L., Rodríguez Blanco, M.L., Santos, J.M.,
Coelho, C.O.A. & Keizer, J.J. (2018) Hydrological and erosion pro-
cesses in terraced fields: Observations from a humid Mediterranean
region in northern Portugal. Land Degradation & Development, 29(3),
596–606. Available from: https://doi.org/10.1002/ldr.2550
Nunes, J.P., Naranjo Quintanilla, P., Santos, J.M., Serpa, D., Carvalho-
Santos, C. et al. (2018) Afforestation, subsequent forest fires and
provision of hydrological services: A model-based analysis for a Med-
iterranean mountainous catchment. Land Degradation & Development,
29(3), 776–788. Available from: https://doi.org/10.1002/ldr.2776
Ortíz-Rodríguez, A.J., Muñoz-Robles, C. & Borselli, L. (2019) Changes in
connectivity and hydrological efficiency following wildland fires in
Sierra Madre Oriental, Mexico. Science of the Total Environment, 655,
112–128. Available from: https://doi.org/10.1016/j.scitotenv.2018.
11.236
Pastor, A.V., Nunes, J.P., Ciampalini, R., Koopmans, M., Baartman, J.,
Huard, F. et al. (2019) Projecting future impacts of global change
including fires on soil erosion to anticipate better land management
in the forests of NW Portugal. Water, 11(12), 122617. Available
from: https://doi.org/10.3390/w11122617
Pausas, J.G., Llovet, J., Rodrigo, A. & Vallejo, R. (2008) Are wildfires a
disaster in the Mediterranean basin? A review. International Journal
of Wildland Fire, 17(6), 713–723. Available from: https://doi.org/10.
1071/WF07151
Prats, S.A., MacDonald, L.H., Monteiro, M., Ferreira, A.J.D., Coelho, C.O.
A. & Keizer, J.J. (2012) Effectiveness of forest residue mulching in
reducing post-fire runoff and erosion in a pine and a eucalypt planta-
tion in north-central Portugal. Geoderma, 191, 115–124. Available
from: https://doi.org/10.1016/j.geoderma.2012.02.009
Raftoyannis, Y., Nocentini, S., Marchi, E., Calama Sainz, R., Garcia
Guemes, C., Pilas, I. et al. (2014) Perceptions of forest experts on
climate change and fire management in European Mediterranean
forests. iForest –Biogeosciences and Forestry, 7(1), 33–41. Available
from: https://doi.org/10.3832/ifor0817-006
Schoorl, J.M., Sonneveld, M.P.W. & Veldkamp, A. (2000) Three-
dimensional landscape process modelling: The effect of DEM
resolution. Earth Surface Processes and Landforms, 25(9), 1025–1034.
Available from: https://doi.org/10.1002/1096-9837(200008)25:
9<1025::AID-ESP116>3.0.CO;2-Z
Schoorl, J.M., Temme, A.J.A.M. & Veldkamp, T. (2014) Modelling centen-
nial sediment waves in an eroding landscape –catchment complex-
ity. Earth Surface Processes and Landforms, 39(11), 1526–1537.
Available from: https://doi.org/10.1002/esp.3605
Schoorl, J.M., Veldkamp, A. & Bouma, J. (2002) Modeling water and soil
redistribution in a dynamic landscape context. Soil Science Society of
FOLLMI ET AL.17
America Journal, 66(5), 1610–1619. Available from: https://doi.org/
10.2136/sssaj2002.1610
Serpa, D., Ferreira, R.V., Machado, A.I., Cerqueira, M.A. & Keizer, J.J.
(2020) Mid-term post-fire losses of nitrogen and phosphorus by
overland flow in two contrasting eucalypt stands in north-central
Portugal. The Science of the Total Environment, 705, 135843. Avail-
able from: https://doi.org/10.1016/j.scitotenv.2019.135843
Shakesby, R.A. (2011) Post-wildfire soil erosion in the Mediterranean:
Review and future research directions. Earth-Science Reviews, 105(3),
71–100. Available from: https://doi.org/10.1016/j.earscirev.2011.
01.001
Shakesby, R.A., Boakes, D.J., Coelho, C.O.A., Gonçalves, A.J.B. & Walsh, R.
P.D. (1996) Limiting the soil degradational impacts of wildfire in pine
and eucalyptus forests in Portugal: A comparison of alternative post-
fire management practices. Applied Geography, 16(4), 337–355.
Available from: https://doi.org/10.1016/0143-6228(96)00022-7
Shakesby, R.A., Coelho, C.D.A., Ferreira, A.D., Terry, J.P. & Walsh, R.P.D.
(1993) Wildfire impacts on soil-erosion and hydrology in wet Medi-
terranean forest, Portugal. International Journal of Wildland Fire, 3(2),
95-110. Available from: https://doi.org/10.1071/WF9930095
Shakesby, R.A., Coelho, C., Ferreira, A.D., Terry, J.P. & Walsh, R.P.D.
(1994) Fire, post-burn land management practice and soil erosion
response curves in eucalyptus and pine forests, north-central
Portugal. In: Soil Erosion as a consequence of Forest fires. Logroño,
Spain: Geoforma Ediciones, pp. 111–132.
Shakesby, R.A. & Doerr, S.H. (2006) Wildfire as a hydrological and geomor-
phological agent. Earth-Science Reviews, 74(3), 269–307. Available
from: https://doi.org/10.1016/j.earscirev.2005.10.006
Smetanová, A., Follain, S., David, M., Ciampalini, R., Raclot, D., Crabit, A. &
le Bissonnais, Y. (2019) Landscaping compromises for land degrada-
tion neutrality: The case of soil erosion in a Mediterranean agricul-
tural landscape. Journal of Environmental Management, 235, 282–
292. Available from: https://doi.org/10.1016/j.jenvman.2019.01.063
SNIRH. (2020) Sistema Nacional de Informaç˜ao de Recursos Hídricos,
Agência Portuguesa do Ambiente.https://snirh.apambiente.pt/ (last
accessed March 2020).
Soto, B. & Díaz-Fierros, F. (1998) Runoff and soil erosion from areas of
burnt scrub: Comparison of experimental results with those
predicted by the WEPP model. Catena, 31(4), 257–270. Available
from: https://doi.org/10.1016/S0341-8162(97)00047-7
Tavares Wahren, F., Julich, S., Nunes, J.P., Gonzalez-Pelayo, O.,
Hawtree, D., Feger, K.H. & Keizer, J.J. (2016) Combining digital soil
mapping and hydrological modeling in a data scarce watershed in
north-central Portugal. Geoderma, 264, 350–362. Available from:
https://doi.org/10.1016/j.geoderma.2015.08.023
Tucker, G.E. & Hancock, G.R. (2010) Modelling landscape evolution. Earth
Surface Processes and Landforms, 35(1), 28–50. Available from:
https://doi.org/10.1002/esp.1952
USDA. (2019) National Engineering Handbook Part 630: Hydrology.
Washington DC: U.S. Department of Agriculture. https://directives.
sc.egov.usda.gov/viewerFS.aspx?hid=21422
van Gorp, W. (2015) LAPSUS User Guide (v0.97) –LAndscape ProcesS
modelling at mUlti dimensions and scaleS. Model Version 5.0.
Vasconcelos Ferreira, R., Azevedo Cerqueira, M., Condesso de Melo, M.T.,
Rebelo de Figueiredo, D. & Keizer, J.J. (2010) Spatial patterns of
surface water quality in the Cértima River basin, Central Portugal.
Journal of Environmental Monitoring: JEM, 12(1), 189–199. Available
from: https://doi.org/10.1039/b914409a
Verheijen, F.G.A., Jones, R.J.A., Rickson, R.J., Smith, C.J., Bastos, A.C.,
Nunes, J.P. & Keizer, J.J. (2012) Concise overview of European soil
erosion research and evaluation. Acta Agriculturae Scandinavica
Section B: Soil and Plant Science, 62(suppl. 2), 185–190. Available
from: https://doi.org/10.1080/09064710.2012.697573
Vieira, D.C.S., Fernández, C., Vega, J.A. & Keizer, J.J. (2015) Does soil burn
severity affect the post-fire runoff and interrill erosion response? A
review based on meta-analysis of field rainfall simulation data.
Journal of Hydrology, 523, 452–464. Available from: https://doi.org/
10.1016/j.jhydrol.2015.01.071
Vieira, D.C.S., Serpa, D., Nunes, J.P.C., Prats, S.A., Neves, R. & Keizer, J.J.
(2018) Predicting the effectiveness of different mulching techniques
in reducing post-fire runoff and erosion at plot scale with the RUSLE,
MMF and PESERA models. Environmental Research, 165, 365–378.
Available from: https://doi.org/10.1016/j.envres.2018.04.029
Walsh, R.P.D., Bento-Gonalves, A.J., Boakes, D.J., Coelho, C., Shakesby, R.
A., & Thomas, A.D. (1994) Impact of fire-induced water repellency
and post-fire forest litter on overland flow in northern and central
Portugal. In Proceedings of the Second International Conference on
Forest Fire Research, vol. II, Coimbra, Portugal, pp. 1149–1159.
Wittenberg, L.E.A. & Inbar, M. (2009) The role of fire disturbance on runoff
and erosion processes –a long-term approach, Mt. Carmel case
study, Israel. Geographical Research, 47(1), 46–56. Available from:
https://doi.org/10.1111/j.1745-5871.2008.00554.x
WRB. (2015) World Reference Base for Soil Resources 2014, update 2015:
International soil classification system for naming soils and creating
legends for soil maps. World Soil Resources Reports No. 106. FAO:
Rome.
Wu, J., Baartman, J.E.M. & Nunes, J.P. (2021) Testing the impacts of
wildfire on hydrological and sediment response using the
OpenLISEM model. Part 2: Analyzing the effects of storm return
period and extreme events. Catena, 207, 105620. Available from:
https://doi.org/10.1016/j.catena.2021.105620
Wu, J., Nunes, J.P., Baartman, J.E.M. & Faúndez Urbina, C.A. (2021)
Testing the impacts of wildfire on hydrological and sediment
response using the OpenLISEM model. Part 1: Calibration and
evaluation for a burned Mediterranean forest catchment. Catena,
207, 105658. Available from: https://doi.org/10.1016/j.catena.
2021.105658
Zema, D.A., Nunes, J.P. & Lucas-Borja, M.E. (2020) Improvement of
seasonal runoff and soil loss predictions by the MMF
(Morgan–Morgan–Finney) model after wildfire and soil treatment in
Mediterranean forest ecosystems. Catena, 188, 104415. Available
from: https://doi.org/10.1016/j.catena.2019.104415
SUPPORTING INFORMATION
Additional supporting information can be found online in the
Supporting Information section at the end of this article.
How to cite this article: Follmi, D., Baartman, J., Benali, A. &
Nunes, J.P. (2022) How do large wildfires impact sediment
redistribution over multiple decades? Earth Surface Processes
and Landforms,1–18. Available from: https://doi.org/10.1002/
esp.5441
18 FOLLMI ET AL.
