Rainfall erosivity mapping for Santiago Island, Cape Verde

Geoderma 217–218 (2014) 74–82

Contents lists available at ScienceDirect

Geoderma journal homepage: www.elsevier.com/locate/geoderma

Rainfall erosivity mapping for Santiago Island, Cape Verde Juan Francisco Sanchez-Moreno a,⁎, Chris M. Mannaerts b, Victor Jetten a a b

Department of Earth System Analysis, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Netherlands Department of Water Resources, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Netherlands

a r t i c l e

i n f o

Article history: Received 21 February 2012 Received in revised form 25 October 2013 Accepted 30 October 2013 Available online 7 December 2013 Keywords: Disdrometer Rainfall erosivity Fournier index Kriging Cape Verde

a b s t r a c t Erosivity, the potential of rainfall to detach soil particles, is a parameter used in several models to link rainfall and soil losses. Erosivity is usually calculated from high temporal resolution rainfall during a long period of time, and data is not always available. For Cape Verde, off the west coast of Africa, where data is limited, researchers have calculated erosivity using 7 year precipitation data at 15 min time interval and using rainfall kinetic energy– intensity relationships developed for temperate areas. In this paper, using additional data collected with an optical disdrometer between 2008 and 2010 with a temporal resolution of 3 min, storm erosivity (EI30) was reevaluated using a new rainfall kinetic energy–intensity relationship developed for Cape Verde. A new equation for storm erosivity as a function of daily rainfall was developed. Annual erosivity R-factor resulting from adding EI30 values was correlated to annual precipitation and to the Modified Fournier Index, calculated from long term monthly data available in Cape Verde. Monthly and long term annual erosivity were mapped using the Modified Fournier Index, and the erosivity R-factor as a function of annual precipitation was mapped for a dry, a wet and an average year. Annual erosivity R-factor in Cape Verde can reach values above 1700 J mm m−2 h−1. Given the strong relationship between rainfall and elevation, high erosivity in Santiago Island occurs on higher elevations, coinciding with steep slopes and shallow soils, which makes these areas susceptible to erosion. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Rainfall erosion causes loss of fertile soil, damage to agriculture and infrastructure, and water pollution. Changes in rainfall patterns are aggravating the risk of erosion around the world (Bernstein, 2007). The effect of rainfall erosion is more evident in countries such as Cape Verde, off the west coast of Africa, where rainfall is highly variable and with strong erosive events (Mannaerts and Gabriels, 2000a,b; Sanchez-Moreno et al., in press). The main parameter to relate soil losses to rainfall erosion is erosivity, which is the ability of rainfall to detach soil particles. Mapping erosivity is of importance to detect areas susceptible to erosion for management and conservation purposes. The most commonly used index for rainfall erosivity is the EI30 developed by Wischmeier and Smith (1958) for the USLE and RUSLE equations. Data measured by Wischmeier and Smith shows that when other factors different to rainfall are constant, soil losses in cultivated fields are proportional to the storm energy E times the maximum 30-min intensity I30. EI30 values summed up and divided by the total number of

years of erosivity estimation result in the parameter R used in USLE and RUSLE. Wischmeier and Smith EI30 index relies on high temporal resolution rainfall data, observed during a long period of time, information that is not always available. When rainfall data is limited, R values can be calculated for the stations with recording rain gauges, then correlated to readily available precipitation, and extrapolated using the associated precipitation (Renard and Freimund, 1994). To cope with lack of data, other indexes, sometimes better correlated to measured erosivity than R, and using daily, monthly and annual rainfall have been created, such as the KE N 25 (Hudson, 1971), the AIm (Lal, 1976), the Fournier index (Fournier, 1960), the Modified Fournier Index (Arnoldus, 1977) and the physically-based A index (Sukhanovski et al., 2002). Different indexes have been developed because erosivity depends on geographic conditions, scale, type of measurement, among others, and there is no single parameter better than other for erosivity estimation (Renard and Freimund, 1994; Stocking, 1987). The Fournier index (Fournier, 1960), developed for the west coast of Africa, has been widely used when only monthly data is available, and is defined as: 2

FI ¼ ⁎ Corresponding author at: Hengelosestraat 99, PO Box 217, 7500 AE Enschede, The Netherlands. E-mail addresses: [email protected] (J.F. Sanchez-Moreno), [email protected] (C.M. Mannaerts), [email protected] (V. Jetten). 0016-7061/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.geoderma.2013.10.026

pmax P

ð1Þ

where pmax is the maximum monthly rainfall (mm) and P is the total annual precipitation (mm). Arnoldus (1977) correlated the Fournier index with R in Morocco, finding poor correlations, but higher when

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

using the monthly average instead of the maximum monthly rainfall, in what has been called the Modified Fournier Index: X12 MFI j ¼

1:58

Pj

:

ð2Þ

In Eq. (2), p is the average precipitation in month i (mm) in year j, and P is total annual precipitation (mm) in year j. Eq. (2) can be used for long-term annual, annual or monthly erosivity replacing p and P for the appropriate rainfall values. Erosivity for a period of N years can be calculated by (Ferro et al., 1999): MFIMA ¼

2 N X 12 X pi; j j

From these data, Mannaerts and Gabriels developed a relationship to determine erosivity (KJ m−2 mm h−1) from daily rainfall P24 (mm): EI 30 ¼ 0:0723 P 24 :

2

p i¼1 i; j

i¼1

Pj

ð3Þ

where MFIMA is basically the average of the Modified Fournier Index MFIj over a period of N years. According to Gabriels et al. (2003) and Apaydin et al. (2006), Eq. (3) captures inter annual variability and prevents underestimation of erosivity compared to using averages of ith rainfall amounts averaged again over a number of years. Renard and Freimund (1994) suggest the use of the Modified Fournier Index for areas where long term data is not available. The Fournier index has been used to map erosivity, either as a single factor or correlated with R, in several countries [e.g. Portugal (Coutinho and Tomás, 1994), Germany (Sauerborn et al., 1999), Ghana (Oduro-Afriyie, 1996), Brazil (Da Silva, 2004), Argentina (Busnelli et al., 2006), Malaysia (Shamshad et al., 2008), Spain (Angulo-Martínez and Beguería, 2009), Mauritius Islands (Nigel and Rughooputh, 2010), North Jordan (Eltaif et al., 2010)]. For Cape Verde, erosivity has been estimated by Mannaerts and Gabriels (2000b) using Wischmeier and Smith (1958) and Brown and Foster (1987) rainfall kinetic energy–intensity relationships KE–I, with precipitation data from 1980 to 1987 measured at a 15-min interval.

75

ð4Þ

Rainfall kinetic energy relationships can be expressed in terms of time, or kinetic energy expenditure (KEtime) and in terms of volume, or kinetic energy content (KEmm) (Kinnell, 1981), where KEmm is equivalent to divide KEtime (also known as the kinetic energy flux) by the rainfall intensity. The first KE–I relationship was introduced by Wischmeier and Smith (1958) using a logarithmic equation, and in terms of volume (KEmm), due to the lack of a proper system at that time, to measure rainfall at small time intervals. The inclusion of the KE–I relationship in the Universal Soil Loss Equation (USLE) (Wischmeier and Smith, 1978) made the use of KEmm the most popular way of expressing kinetic energy–intensity relationships (Salles et al., 2002). Despite their wide use, KEmm − I relationships result from an inappropriate use of statistics, as theirs are equivalent to relate KEtime/I with I, in what is known as a spurious self-correlation, where two parameters with a term in common are employed in a regression analysis (Kenney, 1982; Salles et al., 2002). One of the consequences of using KEmm − I relationships is a high dispersion of values compared to KEtime relationships, that follow a power-law relationship (e.g. Salles et al., 2002; Fornis et al., 2005; Sanchez-Moreno et al., 2012). Moreover, rainfall kinetic energy is regionally dependent (Van Dijk et al., 2002), and several authors have found that the rainfall kinetic energy–intensity relationships KE–I of Wischmeier and Smith and Brown and Foster, developed for temperate areas, produce unreliable erosivity estimates when used for tropical climates (Hudson, 1971; Kinnell, 1973; Lal, 1976). In this research, the rainfall kinetic energy–intensity relationship KE–I developed for Cape Verde by Sanchez-Moreno et al. (2012) in terms of kinetic energy expenditure KEtime, is used to modify the equation from Mannaerts and Gabriels (2000b). The new EI30 − P24 equation is employed to map daily erosivity using areal precipitation by

Fig. 1. Cape Verde and Santiago Island, with the locations of the rainfall stations used in the study, including the location of the disdrometer and of the station Achada das Vacas.

76

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

Sanchez-Moreno et al. (in press) with data obtained from the National Meteorological Institute of Cape Verde (INMG). Monthly and annual erosivity are calculated and mapped using known indexes and the rainfall data available for Santiago Island in Cape Verde. The objectives of this paper are: i) to update the daily erosivity equation from Mannaerts and Gabriels (2000b) using a rainfall kinetic energy expenditure–intensity relationship developed for Cape Verde, and including a new rainfall dataset, ii) to obtain annual erosivity R-factor as a function of readily available data in Santiago Island, and iii) to analyze the spatial distribution of erosivity for Santiago Island at different time scales.

1.1. Study area Cape Verde is an archipelago composed of 8 islands located around 500 km off the west coast of Africa (15.02 N, 23.34 W), in the North Atlantic Ocean (Fig. 1). The total area of the archipelago is approx. 4033 km2. The largest island is Santiago (991 km2), where this study is carried out. Santiago has a volcanic origin, which generated a rugged relief with a highest elevation of 1394 m. Being located in the dry tropical zone, the climate of the country is arid to semi-arid. Air temperatures range from 24 °C in winter to 29 °C during summer. Precipitation is highly influenced by relief and elevation, and varies from less than 100 mm annually at sea level to over 500 mm per year in the higher mountain ranges (Sanchez-Moreno et al., in press). Rainfall is almost entirely concentrated in the months of August to October, when the ITCZ or Inter-Tropical Convergence Zone is active over the archipelago and Western Africa (Mannaerts and Gabriels, 2000a). The main source of income for most Cape Verdians is agriculture, constrained by few land available for cultivation (Beintema et al., 1994; FAO, 2003), poor agricultural practices, rainfall scarcity, and high intensity precipitation that cause soil losses.

2. Methods and materials Two data sets were employed: rainfall data measured at station Achada das Vacas by Mannaerts and Gabriels (2000b) between August 1981 and October 1987 with a temporal resolution of 15 min (P8187), and rainfall data measured between September 1 of 2008 and September 16 of 2010 with a temporal resolution of 3 min (P0810). The latter was obtained from an OTT Parsivel disdrometer (Löffler-Mang and Jürg, 2000). The disdrometer reported 45 days with rainfall exceeding 5 mm for the measuring period (Sanchez-Moreno et al., 2012). For Santiago Island, long term monthly rainfall data for a period of 30 years (1981–2010) was provided by the National Institute of Meteorology and Geophysics of Cape Verde (INMG); however the records were not continuous for all the stations. In Cape Verde, rainfall data if available is incomplete and without information on high temporal resolution or short term evens (Sanchez-Moreno et al., in press). Rain gauges in Santiago Island are mostly semi-automatic, with the rainfall depths recorded manually and by visual inspection the day after the event occurs, which implies that very low intensities may not be recorded, and that errors are introduced while reading and recording rainfall values (Sanchez-Moreno et al., 2012). Fig. 1 shows the stations with data available and used in the study along with the location of the station Achada das Vacas and the disdrometer; Table 1 presents the stations and the number of years with data. Erosivity was calculated from precipitation using either a regression with daily rainfall or the Modified Fournier Index (MFI). The procedure to calculate erosivity was: 1. EI30 (KJ m− 2 mm h− 1) was calculated for the P8187 data, for the P0810 data, and for both datasets combined P8110, using the KE–I

Table 1 Modified Fournier Index and number of years with data available for the rainfall stations in Santiago Island. Coordinates UTM, Zone 27 N. Station

E

N

MFI

No. of years with data

Praia Aeroporto Clássica S. Francisco S. Jorge dos Orgãos Santa Cruz Chão Bom Achada Carreira Mato Mendes Montanha Banana Levada Bom Pau Sala (Renque Purga) Serrado Ribeirinha Funco Bandeira Vale de Mesa Ponte Ferro João Goto Pedra Branca Escola Agró-Pecuária S. Domingos Serra Malagueta Achada Além Ribeira da Barca S. João Baptista Assomada Portãozinho Babosa Picos Alto Casanaia Cutelo Covada Mato Limão Monte Tchota Curralinho Varzea Santana Alto Figueirinha Curral Grande S. Martinho Pequeno Macati Achada Bilim Achada Moerão Ganxemba Flamengos Ribeirão Manuel Lem Pereira Monte Contador Capela Garcia Pico Leão Achada Falcão Mancholy Bom Pau

231,715.8 232,134.9 219,936.5 224,923.3 204,000.0 208,353.6 209,423.4 222,823.8 224,334.9 228,353.2 223,142.6 222,525.8 216,132.4 222,781.0 220,666.4 215,337.3 223,434.3 219,312.7 225,885.2 210,922.1 211,504.1 203,773.7 214,062.4 213,357.1 217,495.6 217,892.0 217,817.0 218,895.3 216,952.0 219,456.5 219,045.2 222,929.2 227,570.4 224,344.7 228,070.9 207,393.1 209,361.2 209,370.0 217,245.3 209,396.7 222,769.3 208,055.9 224,359.4 215,438.8 212,326.0 211,829.5 232,661.7

1,651,588.1 1,658,029.2 1,665,790.7 1,675,949.0 1,689,711.9 1,690,995.7 1,684,985.4 1,668,881.5 1,667,516.8 1,667,950.1 1,667,779.8 1,667,421.0 1,663,341.7 1,665,984.9 1,667,116.9 1,667,328.0 1,673,753.2 1,665,355.2 1,662,024.5 1,680,413.6 1,676,478.7 1,675,415.1 1,654,007.2 1,671,125.8 1,668,621.6 1,667,279.0 1,667,789.0 1,665,191.0 1,665,634.0 1,663,366.7 1,667,000.8 1,665,118.9 1,659,886.2 1,653,661.8 1,671,462.3 1,693,923.7 1,687,024.5 1,689,090.0 1,677,411.7 1,672,020.1 1,663,706.1 1,686,524.0 1,662,684.9 1,663,085.5 1,674,549.4 1,674,503.1 1,653,816.3

11.3 20.6 50.9 26.5 28.8 21.2 28.4 17.8 9.2 25.5 5.1 29.2 12.4 33.9 32.5 24.2 28.6 55.0 33.1 105.2 39.3 18.1 17.7 76.5 66.7 63.0 25.5 33.8 32.4 67.0 47.9 28.2 16.5 21.7 24.2 17.8 39.7 16.1 25.9 69.8 50.6 24.1 18.6 106.8 105.2 85.7 29.4

12 14 30 12 10 3 14 21 12 9 14 27 14 12 27 15 15 26 11 11 8 8 6 13 14 26 16 14 10 27 9 26 11 8 7 1 9 1 2 11 2 8 5 1 1 1 9

relationship (J m−2 h− 1) obtained for Cape Verde by SanchezMoreno et al. (2012): KEtime ¼ 5:94 I

1:37

:

ð5Þ

2. EI30 was correlated to daily precipitation P24 (mm) of P8187, P0810, and for the combined dataset P8110. Data were fitted with a nonlinear regression by non-linear squares (Bates, 1988), using the nls method developed by Douglas M. Bates and Saikat DebRoy within the statistical language R (Ihaka and Gentleman, 1996). 3. The erosivity index R-factor (KJ m−2 mm h−1), defined as the sum of EI30 for a particular year RA, was calculated for P8187 and for P0810. 4. Yearly RA index values were correlated to annual precipitation P using non-linear squares. 5. The Modified Fournier Index for annual precipitation MFIj (mm) was calculated with monthly precipitation data provided by INMG using Eq. (2). 6. Multi-annual erosivity MFIMA was calculated from multi-annual

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

precipitation over the period N (1981–2010) using Eq. (3). Monthly erosivity was calculated with Eq. (3) applied only on the months of interest i over the whole period N. 7. RA was correlated by non-linear squares to annual values of MFIj for single stations (Achada das Vacas between 1981 and 1987, and Disdrometer in Saõ Jorge dos Orgaõs between 2008 and 2010), and to the average MFI j values of all the stations in the island (Table 1). Monthly and multi-annual erosivity MFIMA were mapped using Eq. (3) extrapolated to stations of Santiago Island with at least 3 years of data. To compare the inter annual variations of rainfall in Cape Verde, erosivity Rfactor as a function of annual precipitation was mapped for a relatively dry year (1998), for an average year (2004), and for a year with a high amount of rainfall compared to the series of data (2009). To obtain erosivity maps, the different erosivity equations were applied using the interpolated rainfall maps from Sanchez-Moreno et al. (in press). Several parameters influence rainfall in Santiago Island. Short duration events and daily rainfall are strongly influenced by wind direction, slope, and position of the rain-gauges in the windward or leeward sides of the island (Sanchez-Moreno et al., in press). However, monthly and long term rainfall are strongly influenced by elevation, therefore it was used as external drift for kriging interpolation, both for precipitation (as input for calculation of the R-factor), and for the maps made with the Modified Fournier Index. With kriging with external drift the expected variable is considered a linear function of a second variable that is usually continuous in space (Goovaerts, 1997; Hengl, 2003; Isaaks and Srivastava, 1990; Sanchez-Moreno et al., in press). The relevant parameters employed to fit the variogram of rainfall with elevation as external drift for multi-seasonal rainfall were a partial sill of 8500, zero nugget, and a range of 7000 m. The spatial correlation is weak

77

Table 2 Equation coefficients for EI30 as a function of daily rainfall P24. Quality of fit and predicted values of erosivity (minimum, maximum and mean) obtained from the different rainfall datasets. EQ8187M: Equation coefficients from original EI30 − P relationship by Mannaerts and Gabriels (2000b); EQ8187, EQ0810, EQ8110: Equation coefficients and calculated and predicted values for new EI30 − P relationships calculated with Eq. (7) for corresponding datasets. Erosivity R (KJ m−2 mm h−1) Coefficients

Quality of fit Calculated

Data

a

b

R2

RMSE Min

Max

Mean Min

Max

Mean

EQ8187M EQ8187N EQ0810 EQ8110 EQ8110a

0.0723 0.14 0.0005 0.26 0.24

1.58 1.39 2.84 1.31 1.30

0.90 0.85 0.84 0.63 0.76

11.4 21.2 27.2 36.8 23.1

185.2 185.2 332.2 332.2 185.2

34.0 34.0 38.1 36.0 31.1

280.7 212.6 295.8 252.2 218.8

38.8 36.3 35.8 39.7 33.6

a

0.90 0.90 0.47 0.47 0.47

Predicted

2.32 3.10 0.41 4.72 4.2

Without erosivity outlier.

even at short distances, due to the high variability of rainfall. Details on the data, the procedure employed, and the quality assessment of the interpolations used to estimate rainfall for Santiago Island can be found in Sanchez-Moreno et al. (in press). Interpolation of rainfall using elevation is described as well by Goovaerts (2000), and the use of elevation to improve the interpolation of erosivity is explained by Goovaerts (1999). 3. Results and discussion 3.1. Daily erosivity Storm erosivity EI30 calculated with the KE–I relationship for Cape Verde (Eq. (5)) resulted in lower values for most of the rainfall

Fig. 2. Daily rainfall and daily storm erosivity EI30 calculated with the KE–I equations from Wischmeier and Smith (1958) using Renard et al. (1997) coefficients (W&S), Brown and Foster (1987) (B&F) and the KE–I relationship for Cape Verde, for the events between September 1 of 2008 and September 10 of 2010 with depths of 9 mm or higher.

78

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

Fig. 3. Daily storm erosivity EI30 as a function of daily rainfall P24 and fitted equations. EQ P8187M is the equation obtained by Mannaerts and Gabriels (2000b) with data measured between 1981 and 1987; EQ 8187N is the EI30 − P24 relationship obtained with the new KE–I relationship for Cape Verde applied in dataset P8187; EQ 0810 is the relationship obtained with the dataset measured between 2008 and 2010; EQ 8110 w.o. outlier is the relationship obtained by removing the maximum erosivity value from dataset P8110; EQ 8110 is the equation obtained with the complete dataset (Eq. (7)).

intensities, compared to the equations from Wischmeier and Smith (1958) (using the coefficients from Renard et al. (1997) for the RUSLE equation) and Brown and Foster (1987) (Fig. 2). Conversely, for extremely high intensities Eq. (5) produced higher storm erosivity values (e.g. an atypical high erosivity for September 19, 2009). Besides the constraints of the equations from Wischmeier and Smith, and Brown and Foster discussed in Section 1, both equations result in a minimum kinetic energy even without rainfall, which is not physically realistic, and the latter is limited to a maximum kinetic energy value regardless of the intensity of the rain, whereas such a maximum limit is not present in the data for Cape Verde, or is reached at very high intensities.

The best fit between rainfall erosivity EI30 calculated with Eq. (5) and daily rainfall was obtained with a power-law equation:

b

EI 30 ¼ a P 24 :

ð6Þ

Values for the experimental coefficients a and b of Eq. (6) obtained with the different datasets are presented in Table 2. The table shows the EI30 − P24 equation coefficients by Mannaerts and Gabriels obtained with the dataset between 1981 and 1987 (EQ 8187M), the new equation coefficients for the same dataset applying the KE–I relationship for

Fig. 4. Monthly erosivity in Santiago Island for the rainy season (August to October) calculated by the Modified Fournier Index (MFI) (Eq. (3)) for years from 1981 to 2010.

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

79

prediction point of view, all the equations seem acceptable as erosivity values are dispersed for daily precipitation above 70 mm, with the exception of the equation obtained with the dataset P0810 that completely overestimates erosivity for daily rainfall above 100 mm. However, as the equation from dataset P8110 was derived from all the data available and was calculated with the new KE–I obtained for Cape Verde, we recommend its use for erosivity estimation from daily rainfall: 1:31

EI 30 ¼ 0:26 P 24 :

ð7Þ

Along with the temporal difference between datasets, it is important to note the location difference between the stations where the datasets were collected: the dataset P8187 was measured at the station Achada das Vacas, located about 240 m.a.s.l, while the disdrometer used to measure the dataset P0810 is located 11 km to the north and at 366 m.a.s.l. As rainfall in Santiago Island is influenced by elevation, it is expected that somewhat higher erosivity values will be estimated for the disdrometer location. Eq. (7) is an attempt to capture the erosivity variability between locations. 3.2. Monthly and multi-annual erosivity by MFI

Fig. 5. Long term annual erosivity in Santiago Island calculated by the Modified Fournier Index MFIMA (1981 and 2010, 30-year climate normal).

Cape Verde (EQ 8187N), and new equation coefficients for the dataset between 2008 and 2010 (EQ 0810) and for the complete dataset (EQ 8110). Assessment of the quality of fit for the different equations by the Root Mean Squared Error (RMSE) and the coefficient of determination R2, along with the minimum, maximum and mean erosivity calculated and predicted with the different equations are also shown in Table 2. The correlation coefficient R2 for the complete dataset P8110 was 0.63, and improved to 0.76 after removal of the atypically high erosivity of September 19, 2009 corresponding to a daily rainfall of 105 mm (Fig. 2). Despite the slight improvement in R2, the equation coefficients did not change significantly; however the RMSE was reduced. Fig. 3 presents the datasets with the different fitted equations. From the

Table 1 presents the average Modified Fournier Index for the stations in Santiago Island. Variations in recorded rainfall are caused by stations temporary out of operation, introduction of new stations, but also for data not being recorded, as rain gauges in Cape Verde are mainly semi-automatic, requiring an operator to manually register the rainfall values. The difference in the number of years with data available per station introduces a bias in the calculation of erosivity; to minimize the bias, only stations with data for at least three years were considered for mapping. Fig. 4 shows erosivity calculated by MFI with all the data available (1981–2010) for the months within the rainy season, that cover about 95% of the annual precipitation: August, September and October. July and November are outside the rainy season, but present only sporadic events with high intensity that in the long term cause very low erosivity. September is statistically the month with the highest amount of precipitation, while October is usually the driest month of the wet season, which is reflected in the long term monthly erosivity maps. Rainfall in Santiago Island is strongly affected by topography and elevation (Sanchez-Moreno et al., in press); therefore the strongest erosive effect of precipitation occurs at higher elevations, usually coinciding with

Fig. 6. Annual erosivity R-factor (KJ mm m−2 h−1) as a function of annual precipitation P (mm). Labels indicate the year.

80

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

steep slopes and shallow soils. The areas on top of the mountains are therefore more susceptible to erosion. Erosivity in Santiago Island can be classified as low for MFI below 50 mm, medium for tMFI between 50 and 100 mm, and high for MFI higher than 100 mm. For the 20 year period, the lowest erosivity values were found for October with minimums between 0 and 10 mm, while the highest were in September with a maximum of 188 mm, compared to a maximum of 147 mm for August and 123 mm for October. Fig. 5 presents the long term annual erosivity for Santiago Island calculated using the Modified Fournier Index between 1981 and 2010. For the 30 year period, representing the last climate normal, the maximum annual erosivity with MFI was 90 mm, lower than the maximum long term monthly erosivity, a result caused by several dry years during that period. 3.3. R-factor as a function of annual precipitation The best fit between the erosivity R-factor (J mm m−2 h−1) and annual precipitation P (mm) for the whole data set was obtained with a power law equation (R2 = 0.92): R ¼ 9:77 P

1:68

:

ð8Þ

The data and fitted equation (in KJ mm m−2 h−1) are presented in Fig. 6. Fig. 7 shows R-factor erosivity maps obtained by applying Eq. (8) to annual rainfall of 1998, 2004, and 2009, years considered dry, average and with high amounts of rainfall respectively, when compared to other years in the total series of data (Table 3). 3.4. R-factor as a function of MFI Little to no correlation was found between annual R-factor and MFIj for individual stations. This is due to the fact that the erosivity R-factor is calculated from precipitation higher than 9 mm, while for MFIj the total amount of precipitation reported in the station is employed. Also, very high MFI values caused by months with extreme rainfalls result in outliers for the correlation. The correlation between R-factor for individual stations and the average MFI j for the whole island (complete dataset P8110) resulted in an R2 of 0.58. The best fit was obtained with the equation:

R ¼ 487:8 MFI j

1:66

:

ð9Þ

Table 3 Summary statistics of annual erosivity R-factor (J mm m−2 h−1) as a function of annual rainfall P (mm) calculated with Eq. (8) for a dry year (1998), an average year (2004) and a wet year (2009). Annual precipitation P and seasonal precipitation PS calculated with all the stations over the island. Erosivity R (KJ m−2 mm h−1)

Precipitation (mm)

Year

Min

Max

Mean

Median

PS

P

1998 2004 2009

12.62 13.58 87.62

198.5 458.4 1728

45.07 93.49 371.2

37.26 76.52 303.5

164.3 245.8 523.9

173.9 285.3 530.0

Fig. 4 presents the data and fitted curve of the correlation between R-factor and MFI j for the complete dataset P8110. The correlation between the dataset P8187 and MFI j improved the R2 to 0.85 but causing an extreme overestimation of R-factor for values of MFI j above 100 mm. Despite underestimation of erosivity for years 1984, 1986 and 2009, representing years with extreme rainfall events, Eq. (9) overestimates erosivity for the rest of the years, and seems to represent better the fluctuations on erosivity caused by the variability of rainfall patterns in Cape Verde over a long period of time. The lower correlation coefficient for the complete dataset is caused by months with extreme rainfalls occurring between 2008 and 2010 (e.g. 545 mm reported for a single station in August 2008, and 468 mm in September 2009) (Fig. 8). 4. Conclusions In this paper, daily, monthly, multi-annual, and long term annual erosivity estimates were calculated using available rainfall data for Santiago Island. The EI30 equation from Mannaerts and Gabriels (2000b) was modified using a rainfall kinetic energy–intensity relationship in terms of time KEtime developed for Cape Verde by SanchezMoreno et al. (2012). The erosivity R-factor was evaluated from data between 1981 and 1987 and between 2008 and 2010. The Modified Fournier Index (MFI) was calculated using long term monthly data between 1981 and 2010. R-factor was correlated to annual precipitation and to the MFI. The main findings are: • The KE − I relationship for Cape Verde (Eq. (5)) underestimates erosivity for low rainfall intensities and overestimates it for extreme intensities in comparison to the equations by Wischmeier and Smith (1958) and Brown and Foster (1987). Daily erosivity calculated with the KE − I for Cape Verde can reach values between 200 and 300 KJ mm m−2 h−1 for extreme precipitation.

Fig. 7. Erosivity R-factor (J mm m−2 h−1) as a function of annual precipitation P (mm) calculated with Eq. (8) for a dry year (1998), an average year (2004) and a wet year (2009).

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

81

Fig. 8. Correlation between R-factor erosivity (KJ mm m−2 h−1) and Modified Fournier Index (mean island MFI j ).

• Storm erosivity EI30 was well correlated to daily precipitation P24 for data between 1981 to 1987 and 2008 to 2010 using a power-law equation (Eq. (7)). • Erosivity was calculated with the Modified Fournier Index (MFI). The highest monthly erosivity, for a 30 year period, occurs in September, with values up to 188 mm. The lowest erosivity values, ranging from 0 to 10 mm, are registered in October. • Long term annual erosivity for a 30 year period (1981–2010 climate normal) calculated by the Modified Fournier Index resulted in a maximum value of 90 mm for the highest elevations in Santiago Island. • The erosivity R-factor was calculated as a function of annual precipitation and as a function of the Modified Fournier Index. The correlation between R-factor and annual precipitation produced a coefficient of determination R2 of 0.92, while between R and MFI the coefficient of determination was 0.58. • Rainfall in Santiago Island is strongly influenced by elevation, therefore the highest erosivity values appear at high elevations, coinciding with steep slopes and shallow soils, making these areas susceptible to erosion. • Long term annual data analysis and evaluation of erosivity indexes such as the MFI are useful to understand the long term variations in rainfall erosivity; however, for countries such as Cape Verde with a single rainy season presenting strong inter annual variations, erosivity should be studied as well in terms of single events. • The erosivity equations presented in this study are useful approximations for Cape Verde, where high temporal resolution data is limited. The monthly and long term annual erosivity maps are useful for soil conservation planning and mitigation of erosion damages, as well as inputs for erosion risk models. Acknowledgments The authors would like to thank the personnel of the National Institute of Meteorology and Geophysics of Cape Verde (INMG) for providing the data employed in this study and the personnel of INIDA, the Instituto Nacional de Investigação e Desenvolvimento Agrário of Cape Verde for their support. The research described in this paper was conducted in the framework of the European 6th Framework Research Programme (sub-priority 1.1.6.3) — Research on Desertification — project DESIRE: Desertification Mitigation and Remediation of land — a global approach for local solutions.

References Angulo-Martínez, M., Beguería, S., 2009. Estimating rainfall erosivity from daily precipitation records: a comparison among methods using data from the Ebro Basin (NE Spain). J. Hydrol. 379, 111–121. Apaydin, H., Erpul, G., Bayramin, I., Gabriels, D., 2006. Evaluation of indices for characterizing the distribution and concentration of precipitation: a case for the region of Southeastern Anatolia Project, Turkey. J. Hydrol. 328, 726–732. Arnoldus, H., 1977. Methodology used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. FAO Soils Bull. 34, 39–51. Bates, D., 1988. Nonlinear Regression Analysis and Its Applications, Wiley Series in Probability and Mathematical StatisticsSecond edition. John Wiley & Sons, Inc. Beintema, N.M., Pardey, P.G., Roseboom, J., 1994. Statistical Brief on the National Agricultural Research System of Cape Verde. Technical Report. International Service for National Agricultural Research with Support from the Government of Italy and Special Program for African Agricultural Research (SPAAR). Bernstein, L., 2007. Climate Change 2007: Synthesis Report, Summary for Policymakers. Intergovernmental Panel on Climate Change (IPCC). Un-edited copy Prepared for COP-13 Edition. Brown, L.C., Foster, G.R., 1987. Storm erosivity using idealized intensity distributions. Trans. Am. Soc. Agric. Eng. 30, 379–386. Busnelli, J., del, V., Neder, L., Sayago, J., 2006. Temporal dynamics of soil erosion and rainfall erosivity as geoindicators of land degradation in northwestern Argentina. Quat. Int. 158, 147–161. Coutinho, M.A., Tomás, P.P., 1994. Comparison of Fournier with Wischmeier rainfall erosivity indices. Conserving soil resources: European perspectives. Selected papers from the First International Congress of the European Society for Soil Conservation, CAB International, pp. 192–200. Da Silva, A.M., 2004. Rainfall erosivity map for Brazil. Catena 57, 251–259. Eltaif, N., Gharaibeh, M., Al-Zaitawi, F., Almahad, M., 2010. Approximation of rainfall erosivity factors in North Jordan. Pedosphere 20, 711–717. FAO, 2003. Agriculture and Food in Cape Verde. Available at: http://earthtrends.wri.org/ pdf_library/country_profiles/agr_cou_132.pdf (Accessed: 2011). Ferro, V., Porto, P., Yu, B., 1999. A comparative study of rainfall erosivity estimation for southern Italy and southeastern Australia. Hydrol. Sci. J. 44, 3–24. Fornis, R.L., Vermeulen, H.R., Nieuwenhuis, J.D., 2005. Kinetic energy–rainfall intensity relationship for Central Cebu, Philippines for soil erosion studies. J. Hydrol. 300, 20–32. Fournier, F., 1960. Climat et érosion; la relation entre l'érosion du sol par l'eau et les precipitations atmospheriques, First edition. Presses Universitaires de France, Paris (In French). Gabriels, D., Vermeulen, A., Verbist, K., Van Meirvenne, M., 2003. Assessment of rain erosivity and precipitation concentration in Europe. In: Gabriels, D., Cornelis, W. (Eds.), Proceedings of the International Symposium, 25 Years of Assessment of Erosion, Ghent, Belgium, pp. 87–92. Goovaerts, P., 1997. Geostatistics for Natural Resources Evaluation. Applied Geostatistics Series.Oxford University Press, New York. Goovaerts, P., 1999. Using elevation to aid the geostatistical mapping of rainfall erosivity. Catena 34. Goovaerts, P., 2000. Geostatistical approaches for incorporating elevation into the spatial interpolation of rainfall. J. Hydrol. 228, 113–129. Hengl, T., 2003. Comparison of kriging with external drift and regression kriging. Technical Report. ITC.

82

J.F. Sanchez-Moreno et al. / Geoderma 217–218 (2014) 74–82

Hudson, N., 1971. Soil Conservation, Second edition. B.T. Batsford, London. Ihaka, R., Gentleman, R., 1996. R: a language for data analysis and graphics. J. Comput. Graph. Stat. 5, 299–314 (10618600). Isaaks, E., Srivastava, R., 1990. An Introduction to Applied Geostatistics. Oxford University Press, New York. Kenney, B.C., 1982. Beware of spurious self-correlations! Water Resour. Res. 18, 1041–1048. Kinnell, P.I.A., 1973. The problem of assessing the erosive power of rainfall from meteorological observations. Soil Sci. Soc. Am. J. 37, 617–621. Kinnell, P., 1981. Rainfall intensity–kinetic energy relationships for soil loss prediction. Soil Sci. Soc. Am. Proc. 45, 153–155. Lal, R., 1976. Soil erosion on Alfisols in Western Nigeria. III. Effects of rainfall characteristics. Geoderma 16, 389–401. Löffler-Mang, M., Jürg, J., 2000. An optical disdrometer for measuring size and velocity of hydrometeors. J. Atmos. Ocean. Technol. 17, 130–139. Mannaerts, C.M., Gabriels, D., 2000a. A probabilistic approach for predicting rainfall soil erosion losses in semiarid areas. Catena 40, 403–420. Mannaerts, C.M., Gabriels, D., 2000b. Rainfall erosivity in Cape Verde. Soil Tillage Res. 55, 207–212. Nigel, R., Rughooputh, S., 2010. Mapping of monthly soil erosion risk of mainland Mauritius and its aggregation with delineated basins. Geomorphology 114, 101–114. Oduro-Afriyie, K., 1996. Rainfall erosivity map for Ghana. Geoderma 74, 161–166. Renard, K.G., Freimund, J.R., 1994. Using monthly precipitation data to estimate the R-factor in the revised USLE. J. Hydrol. 157, 287–306. Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D., Yode, D., 1997. Predicting soil erosion by water: a guide to conservation planning with the Revised Universal Soil Loss

Equation (RUSLE). USDA Handbook, 703. United States Department of Agriculture (USDA), Washington, D.C. Salles, C., Poesen, J., Sempere-Torres, D., 2002. Kinetic energy of rain and its functional relationship with intensity. J. Hydrol. 257, 256–270. Sanchez-Moreno, J.F., Mannaerts, C.M., Jetten, V., Löffler-Mang, M., 2012. Rainfall kinetic energy–intensity and rainfall momentum–intensity relationships for Cape Verde. J. Hydrol. 454–455, 131–140. Sanchez-Moreno, J., Mannaerts, C., Jetten, V., 2013. Influence of topography on rainfall variability in Santiago Island, Cape Verde. Int. J. Climatol. http://dx.doi.org/10.1002/ joc.3747 (in press). Sauerborn, P., Klein, A., Botschek, J., Skowronek, A., 1999. Future rainfall erosivity derived from large-scale climate models — methods and scenarios for a humid region. Geoderma 93, 269–276. Shamshad, A., Azhari, M., Isa, M., Hussin, W.W., Parida, B., 2008. Development of an appropriate procedure for estimation of RUSLE EI30 index and preparation of erosivity maps for Pulau Penang in Peninsular Malaysia. Catena 72, 423–432. Stocking, M., 1987. A Methodology for Erosion Hazard Mapping of the SADCC Region. Soil and Water Conservation and Land Utilization Programme. Report no. 9.SADCC, Maseru, Lesotho; Overseas Development Group, Norwich, U.K. Sukhanovski, Y.P., Ollesch, G., Khan, K.Y., Meißner, R., 2002. A new index for rainfall erosivity on a physical basis. J. Plant Nutr. Soil Sci. 165, 51–57. Van Dijk, A.I.J.M., Bruijnzeel, L.A., Rosewell, C.J., 2002. Rainfall intensity–kinetic energy relationships: a critical literature appraisal. J. Hydrol. 261, 1–23. Wischmeier, W.H., Smith, D.D., 1958. Rainfall energy and its relation to soil loss. Trans. Am. Geophys. Union 39, 285–291. Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses: a guide to conservation planning. US Department of Agriculture Handbook No. 537, vol. 58 (58 pp.).