You are on page 1of 10

International Soil and Water Conservation Research 7 (2019) 325e334

Contents lists available at ScienceDirect

International Soil and Water Conservation Research


journal homepage: www.elsevier.com/locate/iswcr

Original Research Article

Improving cover and management factor (C-factor) estimation using


remote sensing approaches for tropical regions
Andre Almagro, Thais Caregnatto Thome , Carina Barbosa Colman, Rodrigo Bahia Pereira,
 Marcato Junior, Dulce Buchala Bicca Rodrigues, Paulo Tarso Sanches Oliveira*
Jose
Federal University of Mato Grosso do Sul, CxP 549, Campo Grande, MS, 79070-900, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: The Revised Universal Soil Loss Equation (RUSLE)'s cover and management factor (C-factor) is one of the
Received 1 May 2019 most difficult factors to obtain, mainly because long-term monitoring soil erosion plots under natural
Received in revised form rainfall are needed. Therefore, remote sensing approaches have been used as an alternative for obtaining
5 August 2019
this factor. However, there is a lack of studies comparing values of this factor computed from remote
Accepted 16 August 2019
sensing approaches with measured data. In this study, we compare two widely used remote sensing
Available online 20 August 2019
approaches (CrA and CVK) to estimate the C-factor based on the Normalized Difference Vegetation Index
(NDVI) with the literature (CLIT) and field experimental data. We also investigated the influence of C-
Keywords:
Soil erosion
factor methods on the prediction of soil loss and sediment yield (SY) using measured data in the
RUSLE Guariroba basin, Central-West Brazil. We obtained mean C-factor values of 0.032, 0.023 and 0.137 for CLIT,
NDVI CrA and CVK, respectively. We found an average annual soil loss of 2.20 t ha1 yr1, 2.02 t ha1 yr1 and
Landsat 8 10.07 t ha1 yr1 and SY values of 6875 t yr1, 6468 t yr1 and 33,435 t yr1, for CLIT, CrA and CVK, respec-
Land use/land cover tively. Our results indicated a significant improvement in soil loss and SY estimations by using the CrA
approach developed for tropical regions, with a bias of 13% to the measured SY (5709 t yr1). We
conclude that the CrA method present the most suitable alternative to compute soil loss and SY in tropical
regions. Furthermore, this approach allows large-scale evaluation and temporal monitoring, therefore
enhancing multi spatial and temporal assessment of soil erosion processes.
© 2019 International Research and Training Center on Erosion and Sedimentation and China Water and
Power Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction plant growth and rainfall dynamics (Nearing et al., 2005). This
factor is defined as a non-dimensional number between zero and
The Universal Soil Loss Equation (USLE) (Wischmeier & Smith, one that represents a rainfall erosivity-weighted ratio of soil loss
1978) and its revised version (RUSLE) (Renard, Foster, Weesies, from land under specified, vegetated conditions to the corre-
McCool, & Yoder, 1997) are the most widely used models for esti- sponding loss from continuous bare fallow (Oliveira et al., 2015b;
mating long-term average annual soil loss and for soil and water Wischmeier & Smith, 1978).
conservation planning (Kinnell, 2010). The RUSLE is composed of For computing the C-factor is necessary to use experimental soil
six factors: rainfall erosivity e R, soil erodibility e K, slope length e erosion plots under natural rainfall, however these studies are
L, slope steepness e S, cover and management e C, and conserva- expensive and time-consuming (Nearing et al., 2000). In many
tion practices e P. Among these factors, the cover and management countries, available measured data in experimental plots are scarce
(C-factor) is one of the most spatiotemporal sensitive as it follows or inexistent (Morgan, 2005), therefore several authors have used
the C-factor developed for the United State of America (Dissmeyer,
George E., Foster, 1981; Wischmeier & Smith, 1978) to map soil
* Corresponding author. Federal University of Mato Grosso do Sul, Campo Grande, erosion, by using geographic information systems (Ganasri &
MS, Brazil. Ramesh, 2016; Oliveira, Alves Sobrinho, Rodrigues, & Panachuki,
E-mail addresses: andre.almagro@gmail.com (A. Almagro), thaisthome04@
), carinabcolman@gmail.com (C.B. Colman), rodrigoeamb@
2011). However, this approach produces uncertainties and may
gmail.com (T.C. Thome
gmail.com (R.B. Pereira), jrmarcato@gmail.com (J. Marcato Junior), dulcebbr@ affect the quality of soil erosion predictions (Oliveira et al., 2015b).
gmail.com (D.B.B. Rodrigues), paulotarsoms@gmail.com, paulo.t.oliveira@ufms.br Another widely used approach to estimate the C-factor is based on
(P.T.S. Oliveira).

https://doi.org/10.1016/j.iswcr.2019.08.005
2095-6339/© 2019 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. Production and Hosting by Elsevier B.V. This
is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
326 A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334

the use of remotely sensed data (Borrelli, Meusburger, Ballabio, of the Guariroba basin, with an area of ~360 km2, located in Campo
Panagos, & Alewell, 2018; Durigon, Carvalho, Antunes, Oliveira, & Grande, Central-West Brazil, in the Cerrado biome (Fig. 1). This
Fernandes, 2014; Panagos et al., 2015; Schmidt, Alewell, & basin is crucial for Campo Grande as it is the main water supply
Meusburger, 2018; Vrieling, 2006; Zhang, Xie, Liu, & Lu, 2011). In source, providing about 40% of the water consumed in the urban
general, the estimated C-factor values from remote sensing data area (Sone et al., 2019). According to the Ko €ppen climate classifi-
have not been compared with C-factor values obtained from cation, the predominant climate is Aw, tropical savanna (Peel,
measured data and, therefore, may also generate uncertainties Finlayson, & McMahon, 2007). The summer is hot and rainy
concerning soil erosion predictions (Oliveira et al., 2015b). (October through March) and winter is dry (April through March).
To determine the C-factor by remote sensing, have been often The mean annual rainfall and temperature are 1500 mm and 23  C,
used vegetation indices such as the Normalized Difference Vegeta- respectively. The soil is classified in the Brazilian soil classification
tion Index (NDVI) (Durigon et al., 2014; Lunetta, Knight, system (SiBCS) as (% of total basin area): Orthic Quartzarenic Neosol
Ediriwickrema, Lyon, & Worthy, 2006). Usually, NDVI is directly - RQo with sandy texture (94.1%); Dystrophic Red Latosol - LVd with
correlated to the C-factor by a linear (Durigon et al., 2014; Lin, Lin, & sandy loam texture (2.4%); and Hydromorphic Quartzarenic Neosol
Chou, 2002) or exponential regression (van der Knijff, Jones, & e RQg with sandy texture (3.5%). The basin is composed by (% of
Montanarella, 1999). C-factor mapping by remote sensing can pro- total basin area): pastures (74.3%); eucalyptus (6.1%); undisturbed
vide suitable information for improving the spatial and temporal cerrado vegetation (Brazilian savanna) (14.4%); riparian zones
€nninger,
modeling of soil erosion (Borrelli et al., 2018; Meusburger, Ba (3.2%); and scarce vegetation areas (1.3%).
& Alewell, 2010). In addition, this approach enables researchers to The study was delineated to first obtain the cover and man-
compare soil loss estimates (Ostovari, Ghorbani-Dashtaki, Bahrami, agement factor (C-factor) from three methods and then to compute
Naderi, & Dematte, 2017) and sediment yield (Zhou & Wu, 2008) in the average annual soil loss from the RUSLE equation. Therefore,
which values of soil erosion and sediment delivery were obtained using the average annual soil loss resulting from the different
from experimental data (Didone , Minella, & Merten, 2015). values of the C-factor approaches, three values of the sediment
The NDVI was used to evaluate the impact of land-use and land- yield (SY) were estimated. Then, to verify which methodology for
cover changes on soil erosion in a developing area in China (Li, obtaining the C-factor is more suitable for sediment yield estima-
Wang, & Liu, 2014). Scho € nbrodt, Saumer, Behrens, Seeber, and tions, we compared these estimation values to the measured
Scholten (2010) used a derivation of the NDVI, the fractional sediment yield from a monitored section in the Guariroba basin
vegetation cover (FVC), to estimate the C-factor for a large moun- (see Fig. 2).
tainous area from the Yangtze River catchment in China and
concluded that the method does not generate reliable values when 2.2. Cover and management (C-factor)
compared with literature values of the C-Factor. Moreover, Schmidt
et al. (2018), the state-of-the-art remote sensing products and the In this study, we used the C-factor values obtained from the
fractional green vegetation cover (FGVC) to calculate the C-factor literature (Section 2.2.1) and the C-factor derived from remote
for grasslands in combination with rainfall dynamics in sensing approaches (Section 2.2.2). We used Landsat 8 Operational
Switzerland. Tanyas, Kolat, and Lütfi Süzen (2015) developed a new Land Imager (OLI) sensor data with 30-m spatial resolution through
approach for obtaining the C-factor from remote sensing and found two equations.
suitable results, despite to highlighted concerns with some unusual
C-factor values found for pasture and forest. The authors also have 2.2.1. C-factor from literature values
estimated the sediment yield for a reservoir, but they do not have We attributed C-factors obtained from long-term monitoring of
compared their results with measured values. Jain and Das (2010) soil erosion plots under natural rainfall to the land cover and land
investigated the use of NDVI and literature values of C-factor to use (LCLU) map of the Guariroba basin, which is the most widely
estimate the sediment yield in an Indian catchment and found used approach worldwide to set the RUSLE model (Borrelli et al.,
biases of 40%. Durigon et al. (2014) developed an equation for 2018; Oliveira et al., 2011; Zare, Panagopoulos, & Loures, 2017).
obtaining the C-factor by using NDVI rescaling in tropical regions. The C-factor values were taken from the study by Oliveira et al.
However, the authors did not evaluate their results against C-factor (2015b), which provide benchmark values useful to evaluate soil
values obtained in field experiments. The application of methods erosion models in Brazil. The LCLU map was obtained by inter-
based on NDVI to estimate the C-factor is especially important in preting and classifying the 1-m spatial resolution IKONOS satellite
tropical regions, such as Brazil. The annual and interannual climatic image from 2013. We referred to this approach throughout the text
variability over Brazil are directly connected with the vegetation as CLIT and the values are 0.013 for cerrado vegetation, 0.020 for
cover over the seasons (Gurgel & Ferreira, 2010). The traditional pasture, 0.030 for Eucalyptus and 1 for bare soil.
method (using a constant C-factor found in the literature) is not
able to capture the spatial and temporal variation over a year such 2.2.2. C-factor from remote sensing approaches
as has been possible by using the NDVI approaches. To determine the C-factor from remote sensing data, we used
In this context, the aim of this study is to compare two widely two methods widely used methods based on the Normalized Dif-
used remote sensing approaches to estimate the C-factor based on ference Vegetation Index (NDVI), due their facility of application
the NDVI (Durigon et al., 2014; van der Knijff et al., 1999) with the and processing, and the high availability of the data input. The NDVI
literature and measured data. We also investigated the influence of is a widely used indicator of green vegetation vigor by the calcu-
these approaches on the RUSLE soil loss and sediment yield esti- lation of the spectral reflectance difference between red and near-
mations comparing then to measured field data in a tropical infrared bands of the satellite image (Tucker, 1979) (Equation (1))
watershed located in Central-West Brazil. and ranges from 1 to þ1, with the highest values attributed to
areas with greater vegetation cover.
2. Material and methods
 
NIR  RED
NDVI ¼ (1)
2.1. Study area and delineation NIR þ RED

This study was developed in the environmental protection area where NIR is the surface spectral reflectance in the near-infrared
A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334 327

Fig. 1. Location of the study area in the Brazilian context (a). Most of the relief is flat (b) and there are six main land uses (c) in the Guariroba basin.

Fig. 2. Flowchart summarizing the methodology adopted in this study.

band and RED is the surface spectral reflectance in the red band.
The first method used to estimate the C-factor was proposed by  
NDVI
Durigon et al. (2014) and adapted by Colman (2018) (CrA), and the CVK ¼ exp  a (3)
second by van der Knijff et al. (1999) (CVK). The CrA and CVK are ðb  NDVIÞ
calculated following Equations (2) and (3), respectively.
where Cr and CVK are the estimated C-factors; a and b are param-
  eters related to the shape of the curve that associates NDVI with the
NDVI þ 1
CrA ¼ 0:1 (2) C-factor.
2
The CrA is based on NDVI rescaling under tropical climate con-
ditions, with more intense rainfall. In relation to CVK, van der Knijff,
328 A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334

Jones, and Montanarella (2000) concluded that for European estimating the K-factor in the studied basin, with coefficient of
climate conditions, values 2 and 1 are the best representatives of correlation of 97% between estimations and measured values.
equation parameters a and b, respectively. Colman (2018) per-
formed comparisons between the C-factor computed using remote 2.5. Topographic factor (LS-factor)
sensing approaches and experimental field studies (Oliveira et al.,
2015b) and found a tenfold systematic bias for the method devel- We computed the LS-factor using a Digital Elevation Model
oped by Durigon et al. (2014) and recommended an adjustment (DEM) from IKONOS imagery (1-m rescaled to 10-m). Within the
factor of 0.1 for CrA output, in order to minimize the biases. This GISus-M system, the LS factor is calculated using the LS-TOOL
adapted approach was also used by Sone et al. (2019) that found developed by Zhang et al. (2013). The LS-factor is computed for
suitable C-factor values. Therefore, in the present study we used the each DEM's pixel, where the L factor is expressed as the Desmet and
adapted approach by Colman (2018). Govers (1996) equation:
To obtain the NDVI used in the two remote sensing approaches, 
we used Landsat 8 OLI surface reflectance imagery and performed a
mþ1  mþ1
Aij þ D2  Aijin
temporal analysis from 2013 to 2016. For each year, four images
Lijin ¼     (4)
were obtained to represent rainy and dry periods (Table 1) from the
Dmþ2  xm ij
 ð22:13Þm
Earth Explorer (https://earthexplorer.usgs.gov/).
We used Landsat 8 OLI surface reflectance orthorectified images,
which present high positional accuracy (Irons, Dwyer, & Barsi, where Lij  in is the slope length for grid cell (i,j); Aij  in is the
2012; Roy et al., 2014; Storey, Choate, & Lee, 2014). We computed contributing area at the inlet of the grid cell with coordinates (i,j), in
NDVI using these atmospheric corrected images, following rec- m2; D is the grid cell size, in m; m is the length exponent of the USLE
ommendations suggested by Song, Woodcock, Seto, Lenney, and L-factor; xij is the equation (sinai;j þ cosai;j ).
Macomber (2001) and Agapiou, Hadjimitsis, Papoutsa, Alexakis, In the RUSLE,m varies according to the ratio of the rill and inter-
and Papadavid (2011), who concluded that uncorrected atmo- rill erosion (b) according to McCool, Foster, Mutchler, and Meyer
spheric effects can impact up to 50% the results of NDVI (1989) (Equations (5) and (6)):
calculations.
m ¼ b=ð1 þ bÞ (5)

2.3. Rainfall erosivity (R-factor) where b varies according to the slope gradient.
The b value is obtained by:
We used the rainfall erosivity map developed by Oliveira,  
h
sin q i
Rodrigues, Alves Sobrinho, Carvalho, and Panachuki (2012),
b¼ 3ðsin qÞ0:8 þ 0:56 (6)
where the authors used data more than 100 rain gauges and or- 0:0896
dinary kriging to generate the rainfall erosivity values across the
The slope steepness (S) was calculated following McCool,
Mato Grosso do Sul State. The R-factor for the municipality of
Brown, Foster, Mutchler, and Meyer (1987) (Equations (7) and (8)).
Campo Grande, MS ranged from 8000 to 10,000 MJ mm ha1 h1
yr1.
S ¼ 10:8 sin q þ 0:03 if q < 9% (7)

2.4. Soil erodibility (K-factor)


S ¼ 16:8 sin q  0:50 if q  9% (8)
The K-factor was obtained from a previous study developed in where q is the slope in degrees.
the Guariroba basin by Anache, Bacchi, Panachuki, and Sobrinho
(2015), whom carried out soil samples throughout this basin to
2.6. Support practice factor (P-factor)
obtain soil characteristics, such as soil texture, permeability,
structure, and organic matter. Then the authors computed the K-
For the P-factor, we used an 1-m resolution Ikonos satellite
factor from different equations and found that the equation pro-
image and field reports from the Campo Grande City Hall to
posed by Sharpley and Williams (1990) was the most suitable for
manually delimit the practices already used in the study area,
following the values suggested by Wischmeier and Smith (1978) for
Table 1 slopes ranging from 3% to 5% as the mean slope for Guariroba basin
Information and characteristics of the satellite imagery used to obtain C-factors from is 3.75%. (Table 2).
remote sensing approaches.

Satellite Instrument Path/row Date Season 2.7. Soil loss estimation


Landsat 8 OLI 224/74 31/05/2013 Rainy
Landsat 8 OLI 224/74 18/07/2013 Dry We used the GISus-M plug-in integrated to ArcGIS Desktop 10.2
Landsat 8 OLI 224/74 03/08/2013 Dry version (Oliveira et al., 2015a). This plug-in is based on the RUSLE
Landsat 8 OLI 224/74 06/10/2013 Rainy
model (Renard et al., 1997), which calculates average annual soil
Landsat 8 OLI 224/74 11/02/2014 Rainy
Landsat 8 OLI 224/74 21/07/2014 Dry
loss according to Equation (9).
Landsat 8 OLI 224/74 22/08/2014 Dry
Landsat 8 OLI 224/74 09/10/2014 Rainy
Landsat 8 OLI 224/74 19/04/2015 Rainy Table 2
Landsat 8 OLI 224/74 22/06/2015 Dry Conservation practice factors.
Landsat 8 OLI 224/74 09/08/2015 Dry
Soil management P-factor
Landsat 8 OLI 224/74 13/11/2015 Rainy
Landsat 8 OLI 224/74 05/04/2016 Rainy Without practice 1
Landsat 8 OLI 224/74 10/07/2016 Dry Leveled tracks 0.5
Landsat 8 OLI 224/74 27/08/2016 Dry Cordons of permanent vegetation 0.25
Landsat 8 OLI 224/74 30/10/2016 Rainy Terrace 0.1
A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334 329

days. Then, streamflow was computed by a rating curve, whose


A ¼ R: K:LS:C:P (9) parameters were optimized. The suspended load samplings were
obtained for vertical integration using the Equal Width Increment
where A is the average annual soil loss, in t ha1 yr1; R is the (EWI) techniques according to Equation (10).
rainfall erosivity factor, in MJ mm ha1 h1 yr1; K is the soil
erodibility factor, in t ha h ha1 MJ1 mm1; LS is the topographic
factor (dimensionless), C is the cover management factor (dimen- Qss ¼ 0:0864  Qw  Cs (10)
sionless), and P is the conservation practice factor (dimensionless).
For the soil loss estimation, we multiplied the RUSLE's factors (R, whereQss is the suspended-sediment (SS) discharge in t d1;Qw is
K, LS, and P) by the C-factors obtained from three different ap- the water discharge in m3 s1; andCs is the mean concentration of
proaches. Thus, three combinations of soil loss estimation were SS in the cross-section in mg l1.
generated and then we investigate the influence of these values on Sediment Yield (SY) values were calculated using the discharge-
the sediment yield estimations against the measured value from weighted mean sediment concentration (Equation (11)) according
the monitored section. The average annual soil loss was computed to Ladegaard-Pedersen et al. (2017). This methodology is widely
for a spatial resolution of 30-m, because this is the coarsest spatial used to compare annual or seasonal variability of concentrations in
resolution used in our input data (Landsat images used to compute a specific river. The authors also state that this method is used to
the C-factor). estimate the annual flows of suspended sediment.

2.8. Sediment monitoring Pn


ðci ti qi Þ
SYmsd ¼ Qa Pi¼1
n (11)
The Guariroba basin has monthly hydro-sedimentological i¼1 ðti qi Þ
measurements of water and sediment discharge carried out from
September 2011 to July 2017 (see Fig. 3). We first obtained where SYmsd is the annual sediment yield, in t yr1; Qa is the mean
streamflow values by using water levels that were recorded every annual flow, in mio. m3; ci is the suspended sediment concentration
10 min using a pressure transducer sensor (Levelogger Junior Edge (SSC), in mg l1; qi is the discharge at time i, in m3 s1; and ti is the
model 3001 LT, Solinst). These data were compared using a me- time interval in which the ith sample represents the river SSC, in
chanical current meter method, which was performed every 30 days.

Fig. 3. Hydro-sedimentological monitoring in the Guariroba basin. (a) aerial view of the monitoring section from IKONOS satellite imagery; (b) general view of the field work on the
section; (c) underwater view of the sediment transport; (d) water discharge measurement for the calibration of the rating curve; and e. Sediment discharge measurement using
vertical integration method.
330 A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334

2.9. Sediment delivery ratio (SDR)


SDR ¼ 0:627  D0:403 (13)
From the soil loss values computed by the RUSLE, we estimate
the sediment yield (SY) based on the Sediment Delivery Ratio (SDR) where D is the slope of the main channel, in %.
model (Equation (12)).

SY ¼ SDR:A (12) 3. Results and discussion

where SY is the sediment yield, in t yr1; SDR is the sediment de- 3.1. C-factor estimations by traditional and remote sensing
livery ratio, dimensionless; and A is the mean annual soil loss approaches
estimated, in t ha1 yr1.
Colman et al. (2018) evaluated five methods that use morpho- Comparing C-factor values from the literature (traditional
metric characteristics of a watershed to estimate the SDR. They method) with those obtained from remote sensing derived from
concluded that the equation proposed by Williams and Berndt NDVI, CVK showed higher overestimations for the mean value of the
(1972) (Equation (13)) is the most suitable to estimate the sedi- entire Guariroba basin, and land covers of cerrado, pasture and
ment delivery ratio for the Guariroba basin, with a percent bias Eucalyptus than CrA (see Table 3). For the bare soil, both methods
ranging from 2.34% to 3.30%. Therefore, we used Equation (13) produced similar underestimations up to 97% for Cr and CVK. These
following Colman et al. (2018). results show the difficulty of methods that use NDVI as input to

Table 3
Classes of land cover and their respective mean C-factor and standard deviation.

Land cover Area (ha) CLIT (Oliveira, Wendland, & Nearing, 2013) CrA CVK

Value PBIAS Value PBIAS

Cerrado 5313 0.013 0.014 ± 0.004 8% 0.028 ± 0.048 115%


Pasture 26,887 0.020 0.025 ± 0.004 25% 0.164 ± 0.068 720%
Eucalyptus 2211 0.030 0.012 ± 0.005 60% 0.027 ± 0.036 10%
Bare soil 478 1.000 0.033 ± 0.004 97% 0.370 ± 0.135 63%
Average for the entire basin 36,000 0.032 0.023 ± 0.007 28% 0.137 ± 0.099 328%

Fig. 4. Mean NDVI (a) during the 2012e2016 period and different approaches to determine the RUSLE's C-factor as CLIT (b), CrA (c) and CVK (d).
A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334 331

estimate the C-factor in areas under bare soil. This was expected, method more accurate for a tropical basin in the Brazilian Atlantic
once these methods are strictly dependent on the biomass spectral Forest.
response captured by satellite sensors. The only case that CVK per- Concerning the development of the Durigon et al. (2014)
formed better than the CrA method was for the Eucalyptus cover. equation, 22 Landsat-5 images from 1986 to 2009 were used for
This fact can be related to the greater similarity in height and fo- calculating the cover management factor (Cr-factor) of the RUSLE
liage of Eucalyptus to the European forests, for which the method model. The mean values found by Durigon et al. (2014) and van der
was developed. Knijff et al. (1999) ranged from 0.084 to 0.228 and 0.003 to 0.065,
As the C-factor is one of the determining variables of the RUSLE, respectively. In general, studies carried out in tropical basins that,
the bias presented in the remote sensing approaches will be carried show results of the european model smaller than the tropical one,
out for any later prediction of soil loss and sediment yield esti- such as Oliveira et al. (2015a). When the values of NDVI based
mation. On the other hand, the CrA values presented more coher- methods are compared with the most used equations to estimate
ence compared with the values found in the literature, especially the C-factor, these values also indicate overestimation in relation to
for the undisturbed cerrado vegetation, which makes the soil loss the traditional methods (Vatandaşlar & Yavuz, 2017).
estimations for the Cerrado biome in Brazil more reliable. Fig. 4
shows the spatial distribution of C-factor values obtained from 3.2. Soil loss and sediment yield estimation
the literature and by remote sensing for the Guariroba basin. The
average and standard deviation values of CrA and CVK are The R-factor varied within the range from 8932 to 9112 with an
0.023 ± 0.007 and 0.137 ± 0.099, respectively. average of 9042 ± 30.29 MJ mm ha1 h1 yr1 (Fig. 5a). These
For most land use classes, CVK was not suitable to determine the values are consistent with those reported by other authors for the
C-factor, presenting greater values than those from experimental same region (Almagro, Oliveira, Nearing, & Hagemann, 2017;
fields. This can be explained by the fact that van der Knijff et al. Oliveira et al., 2013; Trindade, de Oliveira, Anache, & Wendland,
(1999) estimated the monthly C-factor based on NDVI aggregates 2016) and are classified as very high in the global context
every 10 days and then calibrated the parameters for Italy, which (Panagos et al., 2017). The estimated K-factor is of 0.039 ± 0.017 t h
has different climate and vegetation characteristics from our study MJ1 mm1 for orthic Quartzarenic Neosols (RQo) and hydromor-
area. The application proposed by the van der Knijff et al. (1999) phic Quartzarenic Neosols, and 0.028 ± 0.017 t h MJ1 mm1 for the
equation using values of 2 and 1 for a and b parameters, respec- dystrophic Red Latosol (Fig. 5b). The spatial variation of the com-
tively, were not suitable to be used in the Guariroba basin. Similarly, bined LS-factor in the basin is presented in Fig. 5c. The mean LS-
Oliveira et al. (2015a) found values of CVK smaller than Cr, and factor is 0.29 and the highest values are found near water bodies.
standard deviations greater than the mean, considering the Cr The increase in the topographic factor is related to higher rates of

Fig. 5. Spatial distribution of the (a) R-factor; (b) K-factor; (c) LS-factor; and (d) P-factor in the Guariroba basin.
332 A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334

Table 4
Mean values for the RUSLE's factors for the contributing area of the monitoring section.

RUSLE's factor Min Max Mean

Erosivity R 8932 9087 9038


Erodibility K 0.028 0.039 0.039
Topographic LS 0.01 49.18 0.95
Conservation practices P 0.1 1.0 0.7
Cover and management CLIT 0.000 0.100 0.032
CrA 0.004 0.078 0.023
CVK 0.000 1.000 0.137

runoff and erosion (Oliveira et al., 2013). Fig. 5d shows the distri- the basin reaches the monitoring section. With the results from soil
bution of existing conservation practices at the study site. The P loss and the measured annual water discharge of 5.78 m3 s1, we
factor has an average of 0.69 ± 0.39. Table 4 shows the mean value estimated the sediment yield (SY) for the different approaches. The
of each RUSLE's factor for the contributing area of the monitoring SY values were 6875 t yr1, 6468 t yr1 and 33,435 t yr1, for CLIT, CrA
section. and CVK, respectively (Table 5). According to Singh, Babu, Narain,
The mean and standard deviation values of soil loss using CrA Bhushan, and Abrol (1992), these values are classified as moder-
and CVK are 2.02 ± 5.70 and 10.07 ± 36,58 t ha1 yr1, respectively ate and severe. Comparing these results with the average annual SY
(Fig. 6b and c). The mean soil loss calculated using the C-factor measured in the Guariroba basin, which is about 5709 t yr1, the
obtained from the literature is 2.20 ± 19.85 t ha1 yr1 (Fig. 6a). closest estimation is the one associated to the C-factor obtained
These values indicate slight soil erosion for Clit and CrA and mod- using the CrA method (Table 4).
erate soil erosion for CVK. However, we observed that the three Among the values presented in Table 4, the highest percentage
methods present a spatial relationship of soil losses, which occur deviation (486%) is the estimate of SY derived from the CVK, fol-
more intensely near water bodies, making it possible to identify lowed by deviations from the Clit (20%) and CrA (13%). According to
vulnerable areas and allowing decision makers to define priority Pandey, Chowdary, and Mal (2007), a level of 20% of deviation may
areas. be considered “acceptable”, therefore, the SY associated to the C-
The SDR calculated using Williams and Berndt's (1972) equation factor obtained from the literature and CrA is considered “accept-
is 10.54%, which indicates that about 11% of the soil loss occurred in able”. The estimate of SY from the CVK resulted in a high

Fig. 6. Annual average soil loss for the Guariroba basin using different approaches to determine the C-factor: (a) Clit; (b) CrA; and (c) CVK.
A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334 333

Table 5
Annual soil erosion and sediment yield for Guariroba basin. The sediment yield was estimated using the C-factor from the literature and remote sensing approaches.

SY measured (t yr1) C-factor method Soil loss (t ha1 yr1) SY estimated (t yr1) SY bias

5709 CLIT 2.20 6875 20%


CrA 2.01 6468 13%
CVK 10.70 33,435 486%

overestimation that makes this method unsuitable for determining Personnel - Brasil (CAPES) - Finance Code 001 and CAPES PrInt.
the C-factor in our study area. As discussed before, this is relative to
the a and b parameters that were adjusted for European conditions. Appendix A. Supplementary data
Our findings indicate that Clit and CrA methods are suitable to
estimate both soil erosion and sediment yield in a tropical basin, Supplementary data to this article can be found online at
such as the Guariroba basin. However, it is important to note that https://doi.org/10.1016/j.iswcr.2019.08.005.
although the Clit has shown “acceptable” results, it generates a
static value of the C-factor, ignoring the dynamics of the land use
components (Borrelli et al., 2018). Furthermore, there is a lack of References
the literature C-factor data for many parts of the world (Oliveira
Agapiou, A., Hadjimitsis, D. G., Papoutsa, C., Alexakis, D. D., & Papadavid, G. (2011).
et al., 2015b). Therefore, methods that use remote sensing ap- The importance of Accounting for atmospheric effects in the application of
proaches appear as an alternative to minimize this problem. The NDVI and Interpretation of satellite imagery supporting Archaeological
satellite imagery used in the present investigation is freely avail- research: The case studies of Palaepaphos and Nea Paphos Sites in Cyprus.
Remote Sensing, 3, 2605e2629. https://doi.org/10.3390/rs3122605.
able, showing adequate spatial-temporal resolution for the Almagro, A., Oliveira, P. T. S., Nearing, M. A., & Hagemann, S. (2017). Projected
addressed application, making it possible to estimate the C-factor, climate change impacts in rainfall erosivity over Brazil. Scientific Reports, 7,
soil loss and sediment yield on different scales e basin, continental 1e12. https://doi.org/10.1038/s41598-017-08298-y.
Anache, J. A. A., Bacchi, C. G. V., Panachuki, E., & Sobrinho, T. A. (2015). Assessment of
and global e for ungauged areas. methods for predicting soil erodibility in soil loss modeling. Geosci^ encias, 34,
32e40.
Borrelli, P., Meusburger, K., Ballabio, C., Panagos, P., & Alewell, C. (2018). Object-
4. Conclusions oriented soil erosion modelling: A possible paradigm shift from potential to
actual risk assessments in agricultural environments. Land Degradation &
Development, 29, 1270e1281. https://doi.org/10.1002/ldr.2898.
In this study, we compared three different methods to obtain
Colman, C. B. (2018). Impacts of climate and land use changes on soil erosion in the
the cover and management factor (C-factor) from the RUSLE, as Upper Paraguay Basin. Federal University of Mato Grosso do Sul.
follows: from the literature values (Clit) and from remote sensing Colman, C. B., Garcia, K. M. P., Pereira, R. B., Shinma, E. A., Lima, F. E., Gomes, A. O.,
et al. (2018). Different approaches to estimate the sediment yield in a tropical
approaches based on NDVI according to the CrA and CVK equations.
watershed. Rev. Bras. Recur. Hídricos, 23, 2018.
We also evaluated the influence of these approaches on the RUSLE Desmet, P. J. J., & Govers, G. (1996). A GIs procedure for automatically calculating the
soil loss and sediment yield estimations comparing them to USLE LS factor on topographically complex landscape units. Journal of Soil &
measured field data in the Guariroba basin, Central-West Brazil. Water Conservation, 51, 427e433.
Didone , E. J., Minella, J. P. G., & Merten, G. H. (2015). Quantifying soil erosion and
The CrA equation presented similar (percent bias of 28%) C- sediment yield in a catchment in southern Brazil and implications for land
factor values to those obtained from the soil erosion experimental conservation. Journal of Soils and Sediments, 15, 2334e2346. https://doi.org/10.
plots, while the CVK was not capable to well estimate them (percent 1007/s11368-015-1160-0.
Dissmeyer, G. E., & Foster, G. R. (1981). Estimating the cover-management factor (C)
bias of 328%). When compared with the measured sediment yield in the universal soil loss equation for forest conditions. Journal of Soil & Water
(5709 t yr-1), the CLIT, CrA and CVK equations showed a percent bias Conservation, 36, 235e240.
of 20%, 13% and 486%, respectively. Our findings indicated that the Durigon, V. L., Carvalho, D. F., Antunes, M. a H., Oliveira, P. T. S., & Fernandes, M. M.
(2014). NDVI time series for monitoring RUSLE cover management factor in a
CVK equation was not suitable to be used in the Guariroba basin. tropical watershed. International Journal of Remote Sensing, 35, 441e453.
Therefore, this equation should be used with caution in other https://doi.org/10.1080/01431161.2013.871081.
tropical areas, and preferably after evaluating the performance Ganasri, B. P., & Ramesh, H. (2016). Assessment of soil erosion by RUSLE model using
remote sensing and GIS - a case study of Nethravathi Basin. Geoscience Frontiers,
using local/regional literature data obtained from soil erosion
7, 953e961. https://doi.org/10.1016/J.GSF.2015.10.007.
experimental plots. Gurgel, H. C., & Ferreira, N. J. (2010). Annual and interannual variability of NDVI in
We conclude that the CrA method present the most suitable Brazil and its connections with climate. International Journal of Remote Sensing,
24, 3595e3609. https://doi.org/10.1080/0143116021000053788.
alternative to compute soil loss and SY in tropical regions.
Irons, J. R., Dwyer, J. L., & Barsi, J. A. (2012). The next Landsat satellite : The Landsat
Furthermore, this approach allows large-scale evaluation and data Continuity Mission. Remote Sensing of Environment, 122, 11e21. https://doi.
temporal monitoring, therefore enhancing multi spatial and tem- org/10.1016/j.rse.2011.08.026.
poral assessment of soil erosion processes. Our findings may be also Jain, M. K., & Das, D. (2010). Estimation of sediment yield and areas of soil erosion
and Deposition for watershed Prioritization using GIS and remote sensing.
useful for soil and water conservation planning and for estimating Water Resources Management, 24, 2091e2112. https://doi.org/10.1007/s11269-
soil loss and sediment yield considering different remote sensing 009-9540-0.
approaches to determine the cover and management factor, Kinnell, P. I. A. (2010). Event soil loss, runoff and the universal soil loss equation
family of models: A review. Journal of Hydrology, 385, 384e397. https://doi.org/
enhancing the evaluation of large areas and the temporal moni- 10.1016/j.jhydrol.2010.01.024.
toring of the soil erosion processes. van der Knijff, J., Jones, R., & Montanarella, L. (1999). Soil erosion risk assessment in
Italy.
van der Knijff, J., Jones, R., & Montanarella, L. (2000). Soil erosion risk assessment in
Acknowledgements europe.
Ladegaard-Pedersen, P., Sigsgaard, C., Kroon, A., Abermann, J., Skov, K., &
Elberling, B. (2017). Suspended sediment in a high-Arctic river: An appraisal of
This study was supported by the Ministry of Science, Technol- flux estimation methods. The Science of the Total Environment, 580, 582e592.
ogy, Innovation and Communication-MCTIC and National Council https://doi.org/10.1016/j.scitotenv.2016.12.006.
for Scientific and Technological Development-CNPq [grant Lin, C.-Y., Lin, W.-T., & Chou, W.-C. (2002). Soil erosion prediction and sediment yield
estimation: The Taiwan experience. Soil and Tillage Research, 68, 143e152.
numbers 441289/2017-7 and 306830/2017-5]. We would also like https://doi.org/10.1016/S0167-1987(02)00114-9.
to thank the Coordination for the Improvement of Higher Education Li, L., Wang, Y., & Liu, C. (2014). Effects of land use changes on soil erosion in a fast
334 A. Almagro et al. / International Soil and Water Conservation Research 7 (2019) 325e334

developing area. International journal of Environmental Science and Technology, Roy, D. P., Wulder, M. A., Loveland, T. R., Woodcock, C. E., Allen, R. G.,
11, 1549e1562. https://doi.org/10.1007/s13762-013-0341-x. Anderson, M. C., et al. (2014). Landsat-8: Science and product vision for
Lunetta, R. S., Knight, J. F., Ediriwickrema, J., Lyon, J. G., & Worthy, L. D. (2006). Land- terrestrial global change research. Remote Sensing of Environment, 145, 154e172.
cover change detection using multi-temporal MODIS NDVI data. Remote Sensing https://doi.org/10.1016/j.rse.2014.02.001.
of Environment, 105, 142e154. https://doi.org/10.1016/J.RSE.2006.06.018. Schmidt, S., Alewell, C., & Meusburger, K. (2018). Mapping spatio-temporal dy-
McCool, D. K., Brown, L. C., Foster, G. R., Mutchler, C. K., & Meyer, L. D. (1987). namics of the cover and management factor (C- factor) for grasslands in
Revised slope steepness factor for the universal soil loss equation. Transactions Switzerland. Remote Sensing of Environment, 211, 89e104. https://doi.org/10.
of the American Society of Agricultural Engineers, 30, 1387e1396. 1016/j.rse.2018.04.008.
McCool, D. K., Foster, G. R., Mutchler, C. K., & Meyer, L. D. (1989). Revised slope Scho€nbrodt, S., Saumer, P., Behrens, T., Seeber, C., & Scholten, T. (2010). Assessing the
length factor for the universal soil loss equation. Transactions of the American USLE crop and management factor C for soil erosion modeling in a large
Society of Agricultural Engineers, 32, 1571e1576. mountainous watershed in Central China. Journal of Earth Sciences, 21, 835e845.
Meusburger, K., Ba €nninger, D., & Alewell, C. (2010). Estimating vegetation parameter https://doi.org/10.1007/s12583-010-0135-8.
for soil erosion assessment in an alpine catchment by means of QuickBird Sharpley, a. N., & Williams, J. R. (1990). EPIC: The erosion-productivity impact
imagery. International Journal of Applied Earth Observation and Geoinformation, calculator. U S Department of Agriculture Technical Bulletin, 235.
12, 201e207. https://doi.org/10.1016/j.jag.2010.02.009. Singh, G., Babu, R., Narain, P., Bhushan, L. S., & Abrol, I. P. (1992). Soil erosion rates in
Morgan, R. P. C. (2005). Soil erosion and conservation. Blackwell Pub. India. Journal of Soil & Water Conservation, 47, 97e99.
Nearing, M. A., Jetten, V., Baffaut, C., Cerdan, O., Couturier, A., Hernandez, M., et al. Sone, J. S., Gesualdo, G. C., Zamboni, P. A. P., Vieira, N. O. M., Mattos, T. S.,
(2005). Modeling response of soil erosion and runoff to changes in precipitation Carvalho, G. A., et al. (2019). Water provisioning improvement through pay-
and cover. Catena, 61, 131e154. https://doi.org/10.1016/j.catena.2005.03.007. ment for ecosystem services. The Science of the Total Environment, 655,
Nearing, M. A., Romkens, M. J. M., Norton, L. D., Stott, D. E., Rhoton, F. E., Laflen, J. M., 1197e1206. https://doi.org/10.1016/j.scitotenv.2018.11.319.
et al. (2000). Measurements and models of soil loss rates. Science (80-. ), 290, Song, C., Woodcock, C. E., Seto, K. C., Lenney, M. P., & Macomber, S. A. (2001).
1300e1301. https://doi.org/10.1126/science.290.5495.1300b. Classification and change detection using Landsat TM Data : When and how to
Oliveira, P. T. S., Alves Sobrinho, T., Rodrigues, D. B. B., & Panachuki, E. (2011). Correct atmospheric Effects ? Remote Sensing of Environment, 75, 230e244.
Erosion risk mapping applied to environmental zoning. Water Resources Man- https://doi.org/10.1016/S0034-4257(00)00169-3.
agement, 25, 1021e1036. https://doi.org/10.1007/s11269-010-9739-0. Storey, J., Choate, M., & Lee, K. (2014). Landsat 8 operational land imager on-orbit
Oliveira, J. A., Dominguez, J. M. L., Nearing, M. A., & Oliveira, P. T. S. (2015). A GIS- geometric calibration and performance. Remote Sensing, 6, 11127e11152.
based procedure for automatically calculating soil loss from the universal soil https://doi.org/10.3390/rs61111127.
loss equation: GISus-m. Applied Engineering in Agriculture, 31, 907e917. https:// Tanyas, H., Kolat, Ç., & Lütfi Süzen, M. (2015). A new approach to estimate cover-
doi.org/10.13031/aea.31.11093. management factor of RUSLE and validation of RUSLE model in the water-
Oliveira, P. T. S., Nearing, M. A., & Wendland, E. (2015). Orders of magnitude increase shed of Kartalkaya Dam. Journal of Hydrology, 528, 584e598. https://doi.org/10.
in soil erosion associated with land use change from native to cultivated 1016/j.jhydrol.2015.06.048.
vegetation in a Brazilian savannah environment. Earth Surface Processes and Trindade, A. L. F., de Oliveira, P. T. S., Anache, J. A. A., & Wendland, E. (2016). Vari-
Landforms, 40, 1524e1532. https://doi.org/10.1002/esp.3738. abilidade espacial da erosividade das chuvas no Brasil. Pesquisa Agropecuaria
Oliveira, P. T. S., Rodrigues, D. B. B., Alves Sobrinho, T., Carvalho, D. F. De, & Brasileira, 51, 1918e1928. https://doi.org/10.1590/S0100-204X2016001200002.
Panachuki, E. (2012). Spatial variability of the rainfall erosive potential in the Tucker, C. J. (1979). Red and Photographic infrared l , lnear combinations for
state of Mato Grosso do Sul, Brazil. Eng. Agríc. Jaboticabal, 32, 69e79. https://doi. monitoring vegetation. Remote Sensing of Environment, 8, 127e150.
org/10.1017/CBO9781107415324.004. Vatandaşlar, C., & Yavuz, M. (2017). Modeling cover management factor of RUSLE
Oliveira, P. T. S., Wendland, E., & Nearing, M. A. (2013). Rainfall erosivity in Brazil: A using very high-resolution satellite imagery in a semiarid watershed. Environ-
review. Catena, 100, 139e147. https://doi.org/10.1016/j.catena.2012.08.006. mental Earth Sciences, 76. https://doi.org/10.1007/s12665-017-6388-0.
Ostovari, Y., Ghorbani-Dashtaki, S., Bahrami, H.-A., Naderi, M., & Dematte, J. A. M. Vrieling, A. (2006). Satellite remote sensing for water erosion assessment: A review.
(2017). Soil loss estimation using RUSLE model, GIS and remote sensing tech- Catena, 65, 2e18. https://doi.org/10.1016/j.catena.2005.10.005.
niques: A case study from the Dembecha watershed, Northwestern Ethiopia. Williams, J. R., & Berndt, H. D. (1972). Sediment yield computed with universal
Geoderma Reg, 11, 28e36. https://doi.org/10.1016/J.GEODRS.2017.06.003. equation. Journal of the Hydraulics Division, 98, 2087e2098. https://doi.org/10.
Panagos, P., Borrelli, P., Meusburger, K., Alewell, C., Lugato, E., & Montanarella, L. 1038/srep11048.
(2015). Estimating the soil erosion cover-management factor at the European Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses - a guide
scale. Land Use Policy, 48, 38e50. https://doi.org/10.1016/J.LANDUSEPOL.2015. to conservation planning. U S Department of Agriculture Agriculture Handbook,
05.021. 537, 67.
Panagos, P., Borrelli, P., Meusburger, K., Yu, B., Klik, A., Lim, K. J., et al. (2017). Global Zare, M., Panagopoulos, T., & Loures, L. (2017). Land Use Policy Simulating the im-
rainfall erosivity assessment based on high-temporal resolution rainfall records. pacts of future land use change on soil erosion in the Kasilian watershed, Iran.
Scientific Report, 1e12 (In review) https://doi.org/10.1038/s41598-017-04282-8. Land Use Policy, 67, 558e572. https://doi.org/10.1016/j.landusepol.2017.06.028.
Pandey, A., Chowdary, V. M., & Mal, B. C. (2007). Identification of critical erosion Zhang, C., Xie, G., Liu, C., & Lu, C. (2011). Assessment of soil erosion under wood-
prone areas in the small agricultural watershed using USLE, GIS and remote lands using USLE in China. Frontiers of Earth Science, 5, 150e161. https://doi.org/
sensing. Water Resources Management, 21, 729e746. https://doi.org/10.1007/ 10.1007/s11707-011-0158-1.
s11269-006-9061-z. Zhang, H., Yang, Q., Li, R., Liu, Q., Moore, D., He, P., et al. (2013). Extension of a GIS
Peel, M. C., Finlayson, B. L., & McMahon, T. A. (2007). Updated world map of the procedure for calculating the RUSLE equation LS factor. Computers & Geo-
Ko€ppen-Geiger climate classification. Hydrology and Earth System Sciences, 11, sciences, 52, 177e188. https://doi.org/10.1016/j.cageo.2012.09.027.
1633e1644. https://doi.org/10.5194/hess-11-1633-2007. Zhou, W., & Wu, B. (2008). Assessment of soil erosion and sediment delivery ratio
Renard, K. G., Foster, G. R., Weesies, G. A., McCool, D. K., & Yoder, D. C. (1997). using remote sensing and GIS: A case study of upstream Chaobaihe River
Predicting soil erosion by water: A guide to conservation planning with the catchment, north China. International Journal of Sediment Research, 23, 167e173.
revised universal soil loss equation (RUSLE). U S Department of Agriculture https://doi.org/10.1016/S1001-6279(08)60016-5.
Agriculture Handbook, 404.

You might also like