PreprintPDF Available

Abstract and Figures

Abstract. Rockfalls are frequent and harmful phenomena occurring in mountain ranges, coastal cliffs and slope cuts. Albeit several natural processes concur in their formation and triggering, rainfall is one of the most common causes. The prediction of rock failures is of social significance for civil protection purposes and can rely on the statistical analysis of past rainfall conditions that caused the failures. The paper describes the analysis of information on rainfall-induced rockfalls in Gran Canaria and Tenerife, Canary Islands (Spain). An analysis of the monthly rainfall versus the monthly distribution of rockfalls reveals that they are correlated for most of the year, except in summer, when other triggers act to induce collapses. National and regional catalogues with hourly and daily rainfall measurements are used to reconstruct the cumulated amount ( E ) and the duration ( D ) of the rainfall responsible for the rock failures. Adopting a consolidated statistical approach, new ED rainfall thresholds for possible rockfall occurrence and the associated uncertainties are calculated for the two test sites. As far as is known, this is the first attempt to predict this type of failure using the threshold approach. Using the rainfall information, a map of the mean annual rainfall is obtained for Gran Canaria and Tenerife, and it is used to assess the differences between the thresholds. The results of is study are expected to improve the ability to forecast rockfalls in the Canary Islands, in view of implementing an early warning system to mitigate the rockfall hazard and reduce the associated risk.
Content may be subject to copyright.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
1
RAINFALL AND ROCKFALLS IN THE CANARY ISLANDS: ASSESSING A SEASONAL
1
LINK
2
Massimo Melilloa, Stefano Luigi Garianoa, Silvia Peruccaccia, Roberto Sarrob, Rosa Marìa Mateosc,
3
Maria Teresa Brunettia
4
5
a CNR IRPI, via Madonna Alta 126, 06128, Perugia, Italia
6
massimo.melillo@irpi.cnr.it, stefano.luigi.gariano@irpi.cnr.it, silvia.peruccacci@irpi.cnr.it
7
b IGME, c/ Alenza, 1, 28003, Madrid, España
8
r.sarro@igme.es
9
c IGME, Urb. Alcázar del Genil, 4. Edificio Zulema, bajos, 18006, Granada, España
10
rm.mateos@igme.es
11
12
Correspondence to: Maria Teresa Brunetti maria.teresa.brunetti@irpi.cnr.it
13
14
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
2
Abstract
15
Rockfalls are frequent and harmful phenomena occurring in mountain ranges, coastal cliffs and slope
16
cuts. Albeit several natural processes concur in their formation and triggering, rainfall is one of the
17
most common causes. The prediction of rock failures is of social significance for civil protection
18
purposes and can rely on the statistical analysis of past rainfall conditions that caused the failures.
19
The paper describes the analysis of information on rainfall-induced rockfalls in Gran Canaria and
20
Tenerife, Canary Islands (Spain). An analysis of the monthly rainfall versus the monthly distribution
21
of rockfalls reveals that they are correlated for most of the year, except in summer, when other triggers
22
act to induce collapses. National and regional catalogues with hourly and daily rainfall measurements
23
are used to reconstruct the cumulated amount (E) and the duration (D) of the rainfall responsible for
24
the rock failures. Adopting a consolidated statistical approach, new ED rainfall thresholds for possible
25
rockfall occurrence and the associated uncertainties are calculated for the two test sites. As far as is
26
known, this is the first attempt to predict this type of failure using the threshold approach. Using the
27
rainfall information, a map of the mean annual rainfall is obtained for Gran Canaria and Tenerife, and
28
it is used to assess the differences between the thresholds. The results of is study are expected to
29
improve the ability to forecast rockfalls in the Canary Islands, in view of implementing an early
30
warning system to mitigate the rockfall hazard and reduce the associated risk.
31
32
Keywords: Rockfall, rainfall threshold, Canary Islands.
33
34
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
3
1 Introduction
35
Rockfalls are instability processes affecting mountainous regions, coastal cliffs and slope cuts. Being
36
very rapid, they are extremely dangerous and life-threatening, especially when they occur in
37
populated areas, along roads and railways. The most frequent triggering factors of rockfalls are
38
rainfall, cycling thermal stress, and seismic activity (Wieckzorek and Jaeger, 1996; Keefer, 2002;
39
Mateos, 2016; Ansari et al., 2015; Collins and Stock, 2016; Sarro et al., 2018; Saroglou, 2019). At
40
regional and global scales, empirical approaches to forecast the occurrence of rockfalls may
41
contribute reducing risk. Generally, for rainfall-induced slope failures the forecast can rely upon the
42
definition of rainfall thresholds, i.e. the rainfall conditions that when reached or exceeded are likely
43
to trigger the failure. Rainfall thresholds are calculated through the statistical analysis of historical
44
rainfall conditions that have resulted in landslides (e.g., Guzzetti et al., 2007, 2008; Cepeda et al.
45
2010; Sengupta et al., 2010; Ruiz-Villanueva et al. 2011; Berti et al., 2012; Staley et al., 2013; Zêzere
46
et al., 2015; Palenzuela et al., 2016; Rosi et al., 2016; Peruccacci et al., 2017; Segoni et al., 2018;
47
Valenzuela et al., 2018, 2019). The definition of reliable empirical rainfall thresholds relies on the
48
use of objective procedures for (i) the reconstruction of the rainfall events responsible for the failures
49
and (ii) the calculation of the thresholds. For the purpose, Melillo et al. (2018) have proposed an
50
algorithm that reconstructs rainfall events, identifies the rainfall conditions that have resulted in slope
51
failures, and calculates probabilistic cumulated event rainfall-rainfall duration (ED) thresholds at
52
different non-exceeding probabilities and their associated uncertainties (Peruccacci et al., 2012). The
53
obtained thresholds are a set of parallel power-law curves in a log-log (D,E) plane, which are
54
characterized by a slope and an intercept, the last being a function of the non-exceeding probability
55
value (Brunetti et al., 2010).
56
In this work, a relationship between the amount of rainfall and the occurrence of rockfalls is assessed
57
and empirical rainfall thresholds are defined for two test sites in Gran Canaria and Tenerife, Canary
58
Islands (Spain). The possible prediction of rainfall-induced rock failures is of fundamental importance
59
primarily for the safety of the inhabitants and for preserving infrastructures such as roads and
60
buildings. An increasing level of safety against this type of hazard is also important for the local
61
economy, one third of which is based on tourism. As far as is known, this is the first attempt to predict
62
rock failures triggered by rain using the threshold approach. Recently, in Italy it has been observed
63
that the slope of the power-law curve is dependent on the mean annual rainfall (MAR). In particular,
64
the higher is the MAR the steeper is the threshold (Peruccacci et al., 2017). This relationship is
65
explained assuming that where the landscape has been shaped over long time periods by landslides
66
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
4
triggered by a given minimum amount of rainfall, it is likely necessary at least as much rainfall to
67
trigger the next landslides (Chen, 2015). For improving the discussion of the results, it has been
68
considered worthwhile producing a map of the MAR for the islands of Gran Canaria and Tenerife
69
using the available rainfall data sets.
70
The manuscript is organized as follows. After a description in Section 2 of the general settings of the
71
two test sites, Section 3 describes the rainfall and rockfall datasets, and the methods used to determine
72
ED rainfall thresholds and the map of the MAR. Section 4 illustrates in detail the relationship between
73
the rainfall regime and the occurrence of rock failures, and presents the rainfall thresholds for the
74
possible rockfall occurrence in the two test sites. Finally, in Section 5, the main findings of the work
75
are summarised and discussed.
76
2 Test site description
77
The Canary Islands (Spain) are one of the major volcanic chain in the oceans. The archipelago
78
consists of eight islands in the Atlantic Ocean, aligned along a W-SW to E-NE direction: El Hierro,
79
La Palma, La Gomera, Tenerife, Gran Canaria, Fuerteventura, La Graciosa and Lanzarote. The
80
geological origin of the Canary archipelago (800 km in length) is still under debate, but it has been
81
traditionally interpreted as a hotspot track (Fullea et al., 2015).
82
The steep topography and the geological complexity of the archipelago influence the activation of an
83
intense slope failures activity. Rockfalls are the most frequent landslide type in the Canary Islands,
84
causing damage on built-up areas and communication networks.
85
Two test sites are selected for assessing the relationship between the rainfall and the occurrence of
86
rockfalls. The first site (GC) is located in the north-western part of Gran Canaria island, and the
87
second site (TEN) is the entire Tenerife island.
88
2.1 Gran Canaria Island (GC-200 road)
89
Gran Canaria is the third island in size of the Canarian archipelago. With an area of 1560 km2 and a
90
maximum altitude of 1956 m a.s.l., the island is approximately circular in shape (Fig. 1). The origin
91
of Gran Canaria can be dated about 15 million years ago (Miocene) with the first submarine building
92
stages of the Gran Canaria Volcano. From the geological point of view, the island presents the greatest
93
variability of igneous rocks of the entire archipelago. Besides the distinctive lavas of the basanite
94
basalt to trachyte phonolite series, Gran Canaria presents also other types of magma, such as tholeiitic
95
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
5
basalts and rhyolites (Troll and Carracedo, 2016). Massive flank failures and erosion give place to
96
chaotic deposits that cover large areas.
97
The test site is the GC-200 road located in the north-western extreme of Gran Canaria, and specifically
98
between the localities of Agaete and Aldea. The road constitutes the main transportation corridor
99
between the two localities. With a length of 34 km, the road path is very tortuous following the
100
contour of the coast, a very step coastline with some of the highest cliffs in Europe. The road has
101
heavy traffic estimated on average at 1500 vehicles per day. The geology of the test site area is within
102
the domain of the basaltic shield stage, Middle Miocene in age. Along the road, an alternance of
103
alkaline basaltic deposits and piroclastic flows can be observed. In some parts, gravitational deposits
104
(mainly colluvial) also outcrop covering wide areas.
105
Regarding climatological conditions, Gran Canaria is located in a transitional zone between temperate
106
and tropical conditions. The conical morphology of Gran Canaria retains the humidity of the
107
predominant N-NE trade winds of the subtropical Azores anticyclone on the north side of the island.
108
As a result, the northern flanks are humid and vegetation is vigorous, while the south part of the island
109
is very dry and the conditions are very arid and desert-like. Annual rainfall ranges between 100 and
110
1000 mm on average, increasing with altitude. In the test site the climate is very dry, with low average
111
annual rainfall (< 100 mm) and high average annual temperature (~ 20°C).
112
2.2 Tenerife island
113
Tenerife (Fig. 1) is the largest (2057 km2) and the most populated (950,000 inhabitants and 13.2
114
million visitors in 2019) island of the archipelago. It is home to the third largest volcano in the world
115
(Pico del Teide, 3718 m a.s.l.).
116
From a geological point of view, Tenerife was constructed via MiocenePliocene shields that now
117
form the vertices of the island. The shields were unified into a single edifice by later volcanism that
118
continued in central Tenerife from approximately 12 to 8 million years ago and was followed by a
119
period of dormancy. Rejuvenation at approximately 3.5 Ma is recorded by the central Las Cañadas
120
Volcano, and long residence times of magmas during this period favoured magmatic differentiation
121
processes to produce an episode of felsic and highly explosive felsic volcanism (Troll and Carracedo,
122
2016).
123
The steep orography of the island and the climate variety have resulted in a diversity of landscapes
124
and geographical formations. Very impressive coastal cliffs (till 500 m in height) are present in the
125
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
6
northern corner of Tenerife. This area is also characterized by narrow and deep ravines which
126
determine an intense slope activity.
127
The climate of Tenerife is subtropical oceanic; the minimum and maximum annual average
128
temperatures are about 15ºC in winter and 24ºC in summer. Tenerife offers a large variety of micro-
129
climate zones controlled by the altitude and the winds.
130
3 Data and methods
131
The availability of rainfall measurements and landslide information is fundamental to define reliable
132
rainfall thresholds. For the selection of the rain gauges, the data quality and the location of the rain
133
gauges are assessed, given that these features are crucial to characterize the spatial-temporal variation
134
of the precipitations. Similarly, the calculation of the MAR relies on the availability of sufficiently
135
long rainfall series (at least 30 years). This is difficult to achieve for a dense network of rain gauges,
136
where sensors may exhibit different operating time periods. The World Meteorological Organization
137
(WMO) guidelines on the calculation of the annual standard normal, specifically the MAR,
138
recommend at least 10 years to define at least provisional MAR maps (WMO, 1989). This is the case
139
in the test sites, where a lot of rainfall information is limited to short time periods (the average is 15.6
140
years), thus hampering the calculation of the MAR with a detailed space resolution.
141
3.1 Rainfall data
142
In the GC test site, hourly rainfall data (purple triangles in Fig. 1) from the Spanish National
143
Meteorological Service (AEMET) network (in total 25 stations among which 4 are close to the study
144
area) are used for the calculation of rainfall thresholds. Moreover, daily rainfall data (orange triangles
145
in Fig.1) are provided from the Consejo Insular de Aguas de Gran Canaria (CIAGC) regional rain
146
gauge network (13 stations) and from AEMET (92 stations, among which 7 are close to the study
147
area). Some of the sensors of the AEMET network provide both hourly and daily rainfall, in different
148
time periods. Details of the rainfall series are reported in Table 1.
149
For the TEN test site, rainfall measurements are provided by AEMET with the contribution of
150
regional networks. As for the GC test site, the rainfall analysis is performed using both hourly and
151
daily data. The two networks in the TEN test site are composed by 34 rain gauges recording hourly
152
data (purple triangles in Fig. 1) and 66 rain gauges recording daily data (orange triangles in Fig. 1).
153
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
7
To calculate the MAR for the two test sites, yearly and monthly rainfall data provided by AEMET
154
and by Sistema de Información Agroclimático y de Regadíos (SIAR), respectively, are used (Table
155
1). In particular, in order to obtain homogeneous maps, data recorded in the 20-year period from
156
January 2000 to December 2019 in both test sites are selected. Following WMO guidelines (WMO,
157
1989), only stations with at least 10 years of data are included in the analysis. Overall, 72 (one every
158
22 km2) and 67 (one every 31 km2) rain gauges are used to calculate MAR in Gran Canaria and
159
Tenerife, respectively. The average number of sensors operating per year in the considered period is
160
56 (84%; one every 28 km2) in Grand Canaria and 47 (65%; one every 43 km2) in Tenerife. The used
161
rain gauges are homogeneously distributed over the test site areas.
162
Using the monthly and annual rainfall data recorded by the 103 rain gauges in the two islands, the
163
MAR for the period 2000-2019 was calculated for each station. Moreover, the coefficient of variation
164
of the MAR is calculated by dividing its standard deviation by the MAR. This coefficient represents
165
the variability of the MAR in the considered time interval. The map of the MAR and of its coefficient
166
of variation are calculated using the tension spline tool in ESRI ArcMAP 10.7.1.
167
3.2 Rockfall data
168
The information on the rockfalls was collected by the Canarian Civil Protection Authorities in the
169
TEN test site and by the Road Maintenance Service in the GC test site. In particular, for the GC test
170
site a total of 8174 rockfall events occurred from January 2010 to March 2016 was documented. A
171
catalogue was prepared defining accurately the location of each impact along the road using
172
orthophotos available for the region and technical reports. The information for each event includes
173
kilometre point, number of events, date, and boulder size. In GC only 535 rockfalls characterized by
174
medium to large size are included in the analysis for the thresholds, whereas small and very small
175
rockfalls (< 10-3 m3) are discarded. Analogously, a catalogue of 1898 rockfalls that impacted along
176
Tenerife roads from January 2010 to November 2017 was prepared. For each event, the information
177
includes rockfall localization, geographic accuracy, occurrence day, month, year, and time (if
178
available), and temporal accuracy.
179
The influence of the rainfall on the occurrence of rockfalls is assessed analysing the distribution of
180
monthly rainfall (Figs. 2a,b) and monthly number of rockfalls (Figs. 2c,d) on the two test sites. As
181
expected, an increase of the rainfall in the autumn-winter period, between October and March, is
182
observed in both islands, with a maximum in November.
183
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
8
The monthly distribution of rockfalls in Gran Canaria (Fig. 2c) is coherent with the rainfall values in
184
the period January-April, with a maximum in February (~ 130; Fig. 2a). For the remaining dry (May
185
to September) and wet (October to December) months the number of rock failures decreases and
186
becomes almost flat (below 50). This behaviour suggests the presence of triggering mechanisms other
187
than the rainfall. For the TEN test site, the number of rockfalls per month (Fig. 2b) is similar to the
188
rainfall distribution, confirming anyway the presence of one or more additional triggers as evidenced
189
by the abundance of failures between May and September (Fig. 2d) when the rainfall is irrelevant.
190
3.3 Empirical rainfall thresholds
191
Empirical ED thresholds are represented by the following power law curve:
192
E = (α±ΔαD(γ±Δγ) (1)
193
where E is the cumulated event rainfall (in mm), D is the duration of the rainfall event (in hours or in
194
days), α and γ are the intercept and the slope of the curve, respectively, and Δα and Δγ are the
195
uncertainties associated with them. Thresholds at different non-exceedance probabilities are
196
calculated adopting the frequentist approach and the bootstrap nonparametric statistical technique
197
(Brunetti et al., 2010; Peruccacci et al., 2012), using 5000 randomly selected synthetic series of DE
198
pairs. A threshold at 5% non-exceedance probability should leave 5% of the empirical DE pairs below
199
itself. The parameter uncertainties depend mostly on the number and the distribution of the rainfall
200
conditions. The minimum number of DE pairs needed for having stable mean values of the parameters
201
α and γ (i.e. reliable thresholds) depends on the distribution and dispersion of the empirical data points
202
in the DE domain.
203
3.4 The CTRL-T algorithm for threshold calculation
204
The quantitative identification of the rainfall responsible for slope failures and the definition of
205
reliable thresholds are fundamental steps towards a well-founded event prediction (Peruccacci et al.,
206
2017; Melillo et al., 2018). The use of standardized procedures for the reconstruction of the rainfall
207
conditions able to trigger past failures and for the definition of thresholds is necessary for enhancing
208
the objectivity and reproducibility of the curves. The tool named CTRL-T (Calculation of Thresholds
209
for Rainfall-induced Landslides - Tool) proposed by Melillo et al. (2018) is exploited to calculate ED
210
thresholds for the two test sites. CTRL-T reconstructs rainfall events starting from continuous rainfall
211
series. For each rockfall, the algorithm: 1) identifies automatically the representative rain gauge; 2)
212
identifies multiple (D,E) rainfall conditions responsible for the failure; 3) selects among them the
213
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
9
maximum probability rainfall conditions (MPRCs). Then, analysing the distribution of the MPRCs it
214
calculates probabilistic rainfall thresholds at different non-exceeding probabilities and their
215
associated uncertainties. In order to avoid using wrong temporal information (i.e., incorrect dates for
216
the occurrence of rockfalls) in the definition of the thresholds, the rainfall conditions having a delay
217
longer than 48 hours between the rainfall ending time and the rockfall occurrence are discarded.
218
Using CTRL-T, 82 rockfalls occurred between 2012 and 2016 in GC test site and 626 rockfalls
219
occurred between 2010 and 2016 in the TEN test site are selected (light green dots in Fig. 2). The
220
remaining records are discarded due to the: 1) absence of rainfall data in the period including the
221
collapse occurrence time; 2) absence of rain gauges within a buffer of 15 km radius centred on the
222
rockfall; 3) lack of an evident correlation with the rainfall. The definition of rainfall thresholds relies
223
only upon rainfall conditions that triggered the first failure in each event. As a consequence, numerous
224
rockfalls (106, 39% in GC and 271, 30% in TEN) which occurred at the same date and in the same
225
location, and which are associated with the same rainfall event are discarded. In GC among the
226
remaining rockfalls 53 are analysed with daily and 29 with hourly rainfall data, respectively. The low
227
number of rock failures associated to hourly-based rain gauges is to be ascribed to the low density of
228
the sensors in the area. In TEN 245 rockfalls are reconstructed with hourly data and 381 with daily
229
rainfall data. Note that for 83 failures it was possible reconstructing the rainfall conditions using
230
sensors from both the two rain gauge networks. As a consequence, the reconstructed (D,E) rainfall
231
conditions have different temporal resolutions and are used to define both hourly-based and daily-
232
based rainfall thresholds.
233
4 Results
234
A correlation between the rainfall and the observed failures is confirmed by the comparison between
235
the monthly rainfall and the corresponding number of rockfalls both in GC and in TEN (Fig. 3).
236
Figures 3a,b,c show the boxplots of cumulated monthly rainfall based on the data recorded in rain
237
gauges used to reconstruct the rainfall responsible for rockfalls for GC and TEN test sites. Inspection
238
of these figures reveals that the rainfall pattern in the two test sites is typically Mediterranean, with a
239
maximum in winter (but also in October and November) and a minimum in summer, with practically
240
no rain in the warmest months. Analysing data from seven daily-based rain gauges in GC (GC-d), it
241
turns out that the rainiest months are February and November with an average rainfall of 52.2 mm
242
and 55.7 mm, a highest rainfall of 98.6 mm and 133.9 mm, and a median rainfall of 42.3 mm and
243
39.8 mm, respectively (Fig. 3a). A similar trend is found for Tenerife using both daily and hourly
244
data. Data from 40 daily-based rain gauges in TEN (TEN-d) are analysed finding an average rainfall
245
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
10
of 64.6 mm and 86.4 mm, a highest rainfall of 183.5 mm and 183.6 mm, and a median rainfall of 56.1
246
mm and 93.2 mm for February and November, respectively (Fig. 3b). Data from 21 hourly-based rain
247
gauges in TEN (TEN-h) are analysed finding an average rainfall of 88.8 mm and 82.0 mm, a highest
248
rainfall of 169.8 mm and 190.8 mm, and a median rainfall of 97.5 mm and 66.4 mm for February and
249
November, respectively (Fig. 3c).
250
Figures 3d,e,f portray the monthly number of rockfalls associated with rainfall events for GC-d, TEN-
251
d and TEN-h. The GC catalogue lists 53 collapses occurred in the period from November 2012 to
252
October 2016, with the majority of the failures in 2015 (22). The month with the largest number of
253
rockfalls (14) is February, followed by January (8) and November (7). The least number of failures
254
is reported in September (1) and no rainfall-induced rockfalls are reported in May and July (Fig. 3d).
255
The 245 rock failures in the TEN-d catalogue cover the period from September 2010 to February
256
2016, with the majority of records in 2014 (66). The month with the largest number of rockfalls (80)
257
is November, followed by October (37) and December (36). The least number of failures is reported
258
in May (1) and no rainfall-induced rockfalls are reported in June and July (Fig. 3e). The TEN-h
259
catalogue lists 381 rockfalls occurred in the period from September 2010 to November 2016, with
260
the majority of the failures in 2014 (90). The month with the largest number of rockfalls (115) is
261
November, followed by December (72) and October (64). The least number of failures is reported in
262
May (1) and no collapses are reported in July (Fig. 3f).
263
The rainfall that triggered the rockfalls is classified according to the method proposed by Alpert et al.
264
(2002), based on six daily rainfall (Ed) categories from “light” to “torrential” over the Mediterranean
265
(Table 2). Using the procedure adopted by Melillo et al. (2016), each rainfall condition responsible
266
for rock failures (MPRC) is attributed to a specific category. In particular, for events lasting less than
267
24 hours, a category based on the total cumulated rainfall of the event is assigned. For events lasting
268
more than 24 hours, the maximum value of the cumulated rainfall in 24 hours in a moving window is
269
used. In GC, over 40% of the MPRCs responsible for the collapses are classified as moderate-high
270
(MH); in TEN, approximately 30% as high (H) and high-torrential (HT). No MPRCs are found in the
271
lowest Alpert’s category (light, Table 2). Figures 3g,h,i show the cumulated percentage of rainfall
272
events per month grouped according to Alpert’s classification. In GC-d, in February (Fig. 3g) 6
273
rockfalls (43%) are triggered by a rainfall classified as H, 3 (21%) as torrential (H), 3 (22%) as MH,
274
and 1 as light-moderate (LM) and HT each (14%). In TEN-d, in November, 29 rockfalls (36%) are
275
triggered by a rainfall classified as HT, 26 (33%) as MH, 22 (28%) as H, 2 (2%) as LM, and 1 (1%)
276
as T (Fig. 3h). In TEN-h, in November, 5 (4%), 26 (23%), 26 (23%), 31 (27%) and 27 (23%) rockfall
277
are triggered by a rainfall classified as LM, MH, H, HT and T, respectively (Fig. 3i).
278
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
11
Using the catalogues of rainfall events with rockfalls described above and the CTRL-T tool, ED
279
thresholds, and their associated uncertainties are calculated for GC and TEN test sites. Table 3 lists
280
the number of MPRC used to define the thresholds, the equations of the power law curves, and the
281
range of validity for the thresholds, expressed in hours or days. Note that D must be expressed in days
282
in the equations for the thresholds calculated with daily data, and in hours in the equations for the
283
thresholds calculated with hourly data (Gariano et al., 2020).
284
Figure 4a shows, in logarithmic coordinates, the distribution of the (D,E) rainfall conditions,
285
reconstructed with daily data, that have caused rockfalls in GC (53 blue dots) and in TEN (245 green
286
dots). In particular, the 53 daily rainfall conditions responsible for the rockfalls in GC have durations
287
in the range 1 ≤ D ≤ 11 days (with an average value of 2 days) and cumulated rainfall in the range
288
16.5 E ≤ 219.9 mm (average value 51.6 mm). All the conditions were recorded in rain gauges
289
located at a maximum distance of 5.7 km from the failures, with a mean value of 2.8 km. The 245
290
daily-based rainfall conditions associated with the collapses in TEN have durations ranging from one
291
to 15 days, with a mean value of two days. The cumulated rainfall ranges from 15.4 to 235.0 mm,
292
with an average of 71.5 mm. The average distance between the rockfalls and their representative rain
293
gauges is 2.2 km, with a maximum distance of 5 km. Figure 4a portrays also the 5% ED thresholds
294
for GC (T5,GC-d, blue curve) and TEN (T5,TEN-d, green curve). The shaded areas around the threshold
295
lines show the uncertainty regions associated to the thresholds (Table 3). Figure 5b portrays the same
296
T5,GC-d and T5,TEN-d, in linear coordinates, in the range 1 ≤ D ≤ 7 days.
297
Figure 4c shows, in logarithmic coordinates, the distribution of the (D,E) rainfall conditions,
298
reconstructed with hourly data, that have triggered rock failures in TEN (381 purple dots). The hourly
299
rainfall conditions associated to rockfalls have durations ranging from 2 to 712 hours and mean value
300
of 111 hours. The cumulated rainfall ranges from 10.6 to 433.9 mm, with an average of 105.6 mm.
301
The average distance between the rockfalls and the representative rain gauges is 6.7 km, with a
302
maximum distance of 14.9 km. In the log-log plot the purple curve is the 5% threshold for TEN
303
(T5,TEN-h) obtained with hourly data. Figure 5d portrays the same T5,TEN-h, in linear coordinates, in the
304
range 1 ≤ D ≤ 120 hours. The uncertainty associated with the threshold (purple shaded area in Figs.
305
4c,d) is also shown.
306
The difference between the T5,GC-d and T5,TEN-d thresholds can be ascribed to the different MAR in the
307
two test sites. Figure 5 portrays the maps of the MAR and of its coefficient of variation, which is the
308
percentual variability (standard deviation) of the MAR in the considered time interval. The
309
geographical distribution of the MAR values exhibits the highest values in the northern parts of both
310
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
12
islands, where it overcomes 800 mm (Fig. 5a). On the contrary, the highest values of the coefficient
311
of variation (i.e. an index of the MAR variability) are localized in the southern part of the islands,
312
where the rain gauge density is lower (Fig. 5b).
313
5 Discussion and conclusions
314
In Canary Island rainfall is the most important triggering factor for rockfalls (Fig. 2). Nevertheless,
315
there are other factors that predispose directly or indirectly the trigger of the failure (Temiño et al.,
316
2013a). Factors that greatly accentuate this hazard in the two test sites are wind, geomorphological
317
characteristics (e.g., slope, aspect), type of soil and seismic activity. Regarding the wind many
318
collapses are caused by strong gusts of wind that affect the northern side of Tenerife Island and the
319
road GC-200 from Agaete to Aldea in Gran Canaria. (Temiño et al., 2013b). Regarding the
320
geomorphology, the existence of many sections of road running through the old basaltic massifs with
321
significant sub-vertical jointing, makes the area very susceptible to rock failures. In addition, the
322
action of the trade winds on the higher altitude areas, produces an increase in the relative humidity,
323
as large masses of water vapor are retained by steep slopes resulting in an intense weathering (and
324
weakening) of the rock masses. Finally, the large flank instability of the two test sites (especially in
325
the northwest sector of the Gran Canaria island) could be related to structural control and to seismic
326
activity connected to the dynamic geologic condition that characterizes them. (Masson et al., 2002;
327
Temiño et al. 2013b; Urgeles et al., 2001).
328
By selecting the subset of rockfalls triggered by rainfall it can be observed that their monthly
329
frequency is linked to the monthly distribution of the rainfall measured in nearby rain gauges (Figs.
330
3a-f). For GC-d (Figs. 3a,d) the correlation is apparently weaker in fall than in winter, but this could
331
be ascribed to a statistical fluctuation and should be confirmed by increasing the number of events.
332
Conversely, for TEN-d (Figs. 3b,e) the monthly number of rock failures well reflects the monthly
333
rainfall amount, suggesting that rainfall is the only triggering cause. Hourly rainfall data in TEN-h
334
(Figs. 3c,f) confirm partially this outcome, since even with a median lower amount of rainfall, a
335
higher number of rock failures is expected to occur from October to December than in February.
336
The number of rockfalls for which it has been possible to reconstruct the rainfall conditions (MPRCs)
337
using daily and hourly data in the TEN test site (Figs. 3e,f) is different. This is mostly due to the worst
338
temporal resolution of the TEN-d dataset.
339
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
13
In the two test sites, the majority of the rainfall responsible for rockfalls belongs to the Alpert’s MH
340
category (Figs. 3g,i). In TEN-h, 31 events belong to the most severe category T, whereas in TEN-d
341
only one event is found in the T category. This result could be ascribed to the time step of the moving
342
window used to assign the Alpert’s category. For a rainfall event lasting more than one day, the
343
Alpert’s category varies depending on the data temporal resolution, since the time step is one hour or
344
one day for the hourly and daily data, respectively. In TEN-d, the total amount of rainfall responsible
345
for the failure is shared in two or more consecutive days, causing a lowering of the Alpert’s category,
346
as confirmed by the paucity of T events in TEN-d.
347
Figure 4 shows that T5,TEN-d is higher and steeper than T5,GC-d. This means that, at increasing values
348
of D, a smaller amount of rainfall (E) is necessary to trigger the collapses in GC than in TEN.
349
Comparing Figures 1 and 5, the recorded rockfalls in the TEN test site are localized in areas including
350
several classes of MAR (ranging from 100 to 800 mm), while in GC test site they fall in the area
351
characterized by the lowest class of MAR (≤ 100 mm). The different ranges of MAR values in the
352
two test sites are able to explain the observed differences in the two daily ED thresholds (Fig. 4a).
353
This finding confirms that where the MAR is higher, the minimum rainfall conditions able to trigger
354
a failure, specifically a rockfall, are also higher.
355
Moreover, the threshold defined for the TEN test site has an uncertainty smaller than the threshold
356
for GC test site. Peruccacci et al. (2012) observed that the parameter uncertainty reduces as the
357
number of MPRC used to calculate the threshold increases. In particular, as derived from Table 3, the
358
relative uncertainty of the intercept, Δα/α is 9.8% for T5,GC-d and 4.9% for T5,TEN-d. Regarding the
359
slope of the curves, Δγ/γ is 16.1% for T5,GC-d and 6.7% for T5,TEN-d. Given the lower uncertainty range
360
and relative uncertainties of both parameters, T5,TEN-d has a reliability higher than that of T5,GC-d. The
361
same analysis for the T5,TEN-h threshold gives Δα/α = 9.3% and Δγ/γ = 4.2. Thresholds with an hourly
362
temporal resolution and having relative uncertainties of the parameters α and γ lower than 10% could
363
be implemented in an operative system for the prediction of rainfall-induced failures (Peruccacci et
364
al., 2012; 2017). The thresholds for different non-exceedance probabilities obtained for TEN test site
365
using hourly rainfall data are suited for the design of probabilistic schemes for the operative prediction
366
of rainfall-induced rockfalls. An improvement in the number of rain gauges providing hourly
367
measurements, as well as in the number of recorded rock failures, would be necessary in GC test site
368
in order to reduce the uncertainty of the threshold.
369
Currently, neither prototype nor operative early warning systems for rainfall-induced failures are
370
present in the Canary Islands (Guzzetti et al. 2020). The findings of this work can contribute to the
371
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
14
understanding of the rainfall conditions that can trigger rainfall-induced rockfalls in Tenerife and in
372
the western part of Gran Canaria, and their relationship with mean annual rainfall regime. These
373
findings have scientific and social implications given that in both test sites also spring and autumn
374
are characterized by a moderately occurrence of rock failures, with relevant impacts on the
375
population, tourism activities, and local economy. As long as a sufficient amount of empirical data
376
will be available in both test sites (and also in other islands of the archipelago), the method adopted
377
in this work for the definition of reliable rainfall thresholds can be replicated, and the results can be
378
implemented in a prototype early warning system.
379
6 Acknowledgements
380
Research conducted within the framework of the U-Geohaz project (Geohazard Impact Assessment
381
for Urban Areas) funded by the European Commission, Directorate-General Humanitarian Aid and
382
Civil Protection (ECHO), under the call UCPM-2017-PP-AG. This work was also funded by the
383
Salvador de Madariaga Mobility Program from the Spanish Ministry of Science, Project:
384
PRX18/00020.
385
7 References
386
Alpert, P., Ben-Gai, T., Baharan, A., Benjamini, Y., Yekutieli, D., Colacino, M., Diodato, L., Ramis,
387
C., Homar, V., Romero, R., Michaelides, S., and Manes, A.: The paradoxical increase of
388
Mediterranean extreme daily rainfall in spite of decrease in total values, Geophy. Res. Lett., 29(11),
389
31-131-4, https://doi.org/10.1029/2001GL013554, 2012.
390
Ansari, M.K., Ahmed, M., Rajesh Singh, T.N., and Ghalayani, I.: Rainfall, a major cause for rockfall
391
hazard along the roadways, highways and railways on hilly terrains in India, in: Lollino, G., Manconi,
392
A., Clague, J., Shan, W., and Chiarle, M. (Eds.) Engineering Geology for Society and Territory
393
Vol. 1, pp. 457460, Springer, Cham, https://doi.org/10.1007/978-3-319-09300-0_87, 2015.
394
Berti, M., Martina, M.L.V., Franceschini, S., Pignone, S., Simoni, A., and Pizziolo, M.: Probabilistic
395
rainfall thresholds for landslide occurrence using a Bayesian approach, J. Geophys. Res, 117, F04006,
396
https://doi.org/10.1029/2012JF002367, 2012.
397
Brunetti, M.T., Peruccacci, S., Rossi, M., Luciani, S., Valigi, D., and Guzzetti, F.: Rainfall thresholds
398
for the possible occurrence of landslides in Italy, Nat. Hazards Earth Syst. Sci., 10, 447458,
399
https://doi.org/10.5194/nhess-10-447-2010, 2010.
400
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
15
Cepeda, J., Höeg, K., and Nadim, F.: Landslide-triggering rainfall thresholds: a conceptual
401
framework, Q. J. Eng. Geol. Hydroge., 43, 6984, https://doi.org/10.1144/1470-9236/08-066, 2010.
402
Chen, C., Saito, H., and Oguchi, T.: Rainfall intensityduration conditions for mass movements in
403
Taiwan, Prog. Earth Planet. Sci., 2, 14, https://doi.org/10.1186/s40645-015-0049-2, 2015.
404
Collins, B.D., and Stock, G.M.: Rockfall triggering by cyclic thermal stressing of exfoliation
405
fractures. Nat. Geosci., 9, 395400, https://doi.org/10.1038/ngeo2686, 2016.
406
Fullea, J., Camacho, A.G., Negredo, A.M., and Fernández, J.: The Canary Islands hot spot: new
407
insights from 3D coupled geophysicalpetrological modelling of the lithosphere and uppermost
408
mantle, Earth Planet. Sci. Lett., 409, 7188, https://doi.org/10.1016/j.epsl.2014.10.038, 2015.
409
Gariano, S.L., Melillo, M., Peruccacci, S., and Brunetti, M.T.: How much does the rainfall temporal
410
resolution affect rainfall thresholds for landslide triggering?, Nat. Hazards, 100, 655670,
411
https://doi.org/10.1007/s11069-019-03830-x, 2020.
412
Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C.P.: Rainfall thresholds for the initiation of
413
landslides in central and southern Europe, Meteorol. Atmos. Phys., 98(3), 239267,
414
https://doi.org/10.1007/s00703-007-0262-7, 2007.
415
Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C.P.: The rainfall intensityduration control of
416
shallow landslides and debris flows: an update, Landslides 5(1), 317,
417
https://doi.org/10.1007/s10346-007-0112-1, 2008.
418
Guzzetti, F., Gariano, S.L., Peruccacci, S., Brunetti, M.T., Marchesini, I., Rossi, M., and Melillo, M.:
419
Geographical landslide early warning systems, Earth-Sci. Rev., 200, 102973,
420
https://doi.org/10.1016/j.earscirev.2019.102973, 2020.
421
Keefer, D.K.: Investigating landslides caused by earthquakes A historical review, Surv. Geophys.,
422
23, 473510, https://doi.org/10.1023/A:1021274710840, 2002.
423
Masson, D.G., Watts, A.B., Gee, M.J.R., Urgeles, R., Mitchell, N.C., Le Bas, T.P., and Canals, M.:
424
Slope failures on the flanks of the western Canary Islands, Earth-Sci. Rev., 57, 135,
425
https://doi.org/10.1016/S0012-8252(01)00069-1, 2002.
426
Mateos, R.M., García-Moreno, I., Reichenbach, P., Herrera, G., Sarro, R., Rius, J., and Aguiló, R.:
427
Calibration and validation of rockfall modelling at regional scale: application along a roadway in
428
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
16
Mallorca (Spain) and organization of its management, Landslides, 13, 751-763,
429
https://doi.org/10.1007/s10346-015-0602-5, 2016.
430
Melillo, M., Brunetti, M.T., Peruccacci, S., Gariano, S.L., and Guzzetti, F.: Rainfall thresholds for
431
the possible landslide occurrence in Sicily (southern Italy) based on the automatic reconstruction of
432
rainfall events, Landslides, 13(1), 165172, https://doi.org/10.1007/s10346-015-0630-1, 2016.
433
Melillo, M., Brunetti, M.T., Peruccacci, S., Gariano, S.L., Roccati, A., and Guzzetti, F.: A tool for
434
the automatic calculation of rainfall thresholds for landslide occurrence, Environ. Model. Softw., 105,
435
230243, https://doi.org/10.1016/j.envsoft.2018.03.024, 2018.
436
Palenzuela, J.A., Jiménez-Perálvarez, J.D., and Chacón, J.: Assessing critical rainfall thresholds for
437
landslide triggering by generating additional information from a reduced database: an approach with
438
examples from the Betic Cordillera (Spain), Nat. Hazards, 84, 185212,
439
https://doi.org/10.1007/s11069-016-2416-8, 2016.
440
Peruccacci, S., Brunetti, M.T., Luciani, S., Vennari, C., and Guzzetti, F.: Lithological and seasonal
441
control of rainfall thresholds for the possible initiation of landslides in central Italy, Geomorphology,
442
139140, 7990, https://doi.org/10.1016/j.geomorph.2011.10.005, 2012.
443
Peruccacci, S., Brunetti, M.T., Gariano, S.L., Melillo, M., Rossi, M., and Guzzetti, F.: Rainfall
444
thresholds for possible landslide occurrence in Italy, Geomorphology, 290, 3957,
445
https://doi.org/10.1016/j.geomorph.2017.03.031, 2017.
446
Rosi, A., Peternel, T., Jemec-Auflič, M., Komac, M., Segoni, S., and Casagli, N.: Rainfall thresholds
447
for rainfall-induced landslides in Slovenia, Landslides, 13, 15711577,
448
https://doi.org/10.1007/s10346-016-0733-3, 2016.
449
Ruiz-Villanueva, V., Bodoque, J.M., Díez-Herrero, A., and Calvo, C.: Triggering threshold
450
precipitation and soil hydrological characteristics of shallow landslides in granitic landscapes,
451
Geomorphology, 133(3), 178189, https://doi.org/10.1016/j.geomorph.2011.05.018, 2011.
452
Saroglou, C.: GIS-based rockfall susceptibility zoning in Greece, Geosciences, 9(4), 163,
453
https://doi.org/10.3390/geosciences9040163, 2019.
454
Sarro, R., Riquelme, A., García-Davalillo, J., Mateos, R., Tomás, R., and Pastor, J.: Rockfall
455
Simulation Based on UAV Photogrammetry Data Obtained during an Emergency Declaration:
456
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
17
Application at a Cultural Heritage Site, Remote Sens., 10(12), 1923,
457
https://doi.org/10.3390/rs10121923, 2108.
458
Segoni, S., Piciullo, L., and Gariano, S.L.: A review of the recent literature on rainfall thresholds for
459
landslide occurrence, Landslides, 15, 14831501, https://doi.org/10.1007/s1034 6-018-0966-4, 2018.
460
Sengupta, A., Gupta, S., and Anbarasu, K.: Rainfall thresholds for the initiation of landslide at Lanta
461
Khola in north Sikkim, India, Nat. Hazards, 52, 3142, https://doi.org/10.1007/s11069-009-9352-9,
462
2010.
463
Staley, D.M., Kean, J.W., Cannon, S.H., Schmidt, K.M., and Laber, J.L.: Objective definition of
464
rainfall intensityduration thresholds for the initiation of post-fire debris flows in southern California,
465
Landslides, 10, 547562, https://doi.org/10.1007/s10346-012-0341-9, 2013.
466
Temiño, J.Y., Rodrìguez-Peces, M.J., Marchesini, S., Leyva, S., and Dìaz-Hernàndez, J.L.:
467
Amplification of the destructive effects of rock falls by sliding on volcanic soils: examples from the
468
Anaga Massif (Tenerife Island, Spain), in: Margottini, C., Canuti, P., and Sassa, K. (Eds.) Landslide
469
Science and Practice, Springer-Verlag Berlin Heidelberg, Vol. 1, pp. 191195,
470
https://doi.org/10.1007/978-3-642-31325-7_25, 2013a.
471
Temiño, J.Y., Rodrìguez-Peces, M.J., Sànchez, N., Galindo, I., and del Potro, R.: Geomorphologic
472
evidences of flank instabilities in the eastern sector of the Tejeda volcano (Canary Islands, Spain)
473
during the Quaternary, in: Margottini, C., Canuti, P., and Sassa, K. (Eds.) Landslide Science and
474
Practice, Springer-Verlag Berlin Heidelberg, Vol. 7, pp. 6572, https://doi.org/10.1007/978-3-642-
475
31313-4_9, 2013b.
476
Troll, V.R., and Carracedo, J.C.: The Geology of the Canary Islands, Elsevier, pp. 636. ISBN 978-0-
477
12-809663-5, https://doi.org/10.1016/C2015-0-04268-X, 2016.
478
Urgeles, R., Canals, M., and Masson, D.G.: 2001 Flank stability and processes off the western Canary
479
Islands: a review from El Hierro and La Palma, Sci. Mar., 65(1), 2131,
480
https://doi.org/10.3989/scimar.2001.65s121, 2001.
481
Valenzuela, P., Domínguez-Cuesta, M.J., Mora García, M.A., and Jiménez-Sánchez, M.: Rainfall
482
thresholds for the triggering of landslides considering previous soil moisture conditions (Asturias,
483
NW Spain), Landslides, 15, 273282, https://doi.org/10.1007/s10346-017-0878-8, 2018.
484
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
18
Valenzuela, P., Zêzere, J.L., Domínguez-Cuesta, M.J., and Mora García, M.A.: Empirical rainfall
485
thresholds for the triggering of landslides in Asturias (NW Spain), Landslides, 16, 12851300,
486
https://doi.org/10.1007/s10346-019-01170-2, 2019.
487
Wieczorek, G.F., and Jaeger, S.: Triggering mechanisms and depositional rates of postglacial slope
488
movement processes in the Yosemite Valley, California, Geomorphology, 15, 1731,
489
https://doi.org/10.1016/0169-555X(95)00112-I, 1996.
490
WMO World Meteorological Organization: Calculation of Monthly and Annual 30-Year Standard
491
Normals, WMO/TD No. 341, WCDP-No. 10, Geneva, 1989.
492
Zêzere, J.L., Vaz, T., Pereira, S., Oliveira, S.C., Marques, R., and Garcia, R.A.C.: Rainfall thresholds
493
for landslide activity in Portugal: a state of the art, Environ. Earth Sci., 73(6), 29172936,
494
https://doi.org/10.1007/s12665-014-3672-0, 2015.
495
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
19
496
Figure 1. GC and TEN test sites. Location of the rain gauges providing hourly (purple triangles) and daily
497
(orange triangles) rainfall measurements, and of rockfalls used for threshold calculations (light green dots).
498
Hillshade derived from MDT05 2009 CC-BY 4.0 scne.es.
499
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
20
500
Figure 2. Comparison between monthly rainfall and rockfall occurrence. (a, b) Annual variation of monthly
501
rainfall measures in GC (cyan) and TEN (grey). The whiskers show 1.5 times the interquartile range. (c, d)
502
Number of rockfalls per month in the two test sites.
503
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
21
504
Figure 3. Comparison between monthly rainfall and rainfall-induced rockfalls and Alpert classification. (a, b,
505
c) Annual variation of monthly rainfall measures in the test sites. Legend: GC-d, daily rainfall data in GC test
506
site; TEN-d, daily rainfall data in the TEN test site; TEN-h, hourly rainfall data in TEN test site. (d, e, f)
507
Number of rainfall-induced rockfalls per month. (g, h, i) Cumulated percentage of rainfall events per month
508
classified according to Alpert et al. (2002). Legend: LM, light-moderate (4 < Ed ≤ 16 mm); MH, moderate-
509
heavy (16 < Ed 32 mm); H, heavy (32 < Ed ≤ 64 mm); HT, heavy-torrential (64 < Ed ≤ 128 mm); T, torrential
510
(Ed > 128 mm).
511
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
22
512
Figure 4. Rainfall thresholds for the possible rockfall occurrence in the two test sites. (a) Rainfall duration D
513
(x-axis, in days) and cumulated event rainfall E (y-axis, in mm) conditions that have produced rockfalls in GC
514
(53 blue dots) and TEN (245 green dots) test sites, respectively. Green and blue curves are the 5% power law
515
thresholds (T5,TEN-d, T5,GC-d). (b) 5% daily ED thresholds for GC and TEN in linear coordinates, in the range of
516
durations 1 ≤ D ≤ 7 days. (c) Rainfall duration D (x-axis, in hours) and cumulated event rainfall E (y-axis, in
517
mm) conditions that have produced rockfalls in TEN (381 purple dots) test site. Purple curve is the 5% power
518
law threshold (T5,TEN-h). (d) 5% hourly ED thresholds for GC and TEN in linear coordinates, in the range of
519
durations 1 ≤ D ≤ 120 hours.
520
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
23
521
Figure 5. Maps of (a) mean annual rainfall and (b) of its coefficient of variation. The rain gauges used for these
522
analysis (cf. Table 1) are also shown.
523
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
24
Table 1. Summary of the three available rain gauge networks (CIAGC, AEMET, SIAR) in the two test sites
524
(GC and TEN) i.e., network name, network operating time period, temporal resolution, test site, number of
525
used rain gauges, their average operating time, and the use of data.
526
Network
Period
Temporal
resolution
Test site
Average
operating time
(year)
Data
application
CIAGC
Jan 2010 - Dec 2017
daily
GC
13
8.0
Thresholds
AEMET
Jan 1951 - May 2019
92
41.8
Oct 1997 - May 2019
hourly
GC
25
16.5
Jan 2010 - Mar 2018
TEN
34
5.5
Jan 2010 - May 2018
daily
TEN
66
8.2
Jan 2000 - Dec 2019
yearly
GC
67
15.2
MAR
TEN
58
13.8
SIAR
Jan 1999 - Dec 2019
monthly
GC
5
18.2
TEN
9
15.1
527
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
25
Table 2. Summary of the number (#) and percentage (%) of MPRC in the categories proposed by Alpert et al.
528
(2002), in the two test sites.
529
Category
Ed (mm)
GC-d
TEN-d
TEN-h
#
%
#
%
#
%
Light (L)
Ed ≤ 4
0
0
0
0
0
0
Light-moderate (LM)
4 < Ed ≤ 16
11
20.7
11
4.5
28
7.3
Moderate-heavy (MH)
16 < Ed ≤ 32
23
43.4
92
37.5
86
22.6
Heavy (H)
32 < Ed ≤ 64
14
26.4
80
32.7
117
30.7
Heavy-torrential (HT)
64 < Ed ≤ 128
2
3.8
58
23.7
116
30.5
Torrential (T)
Ed >128
3
5.7
4
1.6
34
8.9
530
531
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
Rainfall and rockfalls in the Canary Islands: assessing a seasonal link
26
Table 3. ED rainfall thresholds at different non-exceedance probabilities (1%, 5%, 10%, 20%, 35% and 50%)
532
for the GC and TEN test sites. The number of MPRC and the duration range of each threshold are also reported.
533
Threshold
name
Number of
MPRC
Threshold equation
Duration
range
T1,GC-d
53
E = (8.3±1.0)×D(0.62±0.10)
1-11 days
T5,GC-d
E = (12.3±1.2)×D(0.62±0.10)
T10,GC-d
E = (15.1±1.4)×D(0.62±0.10)
T20,GC-d
E = (19.5±1.8)×D(0.62±0.10)
T35,GC-d
E = (25.5±2.5)×D(0.62±0.10)
T50,GC-d
E = (31.9±3.6)×D(0.62±0.10)
T1,TEN-d
245
E = (11.6±0.6)×D(0.75±0.05)
1-15 days
T5,TEN-d
E = (16.3±0.8)×D(0.75±0.05)
T10,TEN-d
E = (19.6±0.8)×D(0.75±0.05)
T20,TEN-d
E = (24.4±1.0)×D(0.75±0.05)
T35,TEN-d
E = (30.6±1.4)×D(0.75±0.05)
T50,TEN-d
E = (37.1±1.8)×D(0.75±0.05)
T1,TEN-h
381
E = (2.8±0.3)×D(0.48±0.02)
2-712 hours
T5,TEN-h
E = (4.3±0.4)×D(0.48±0.02)
T10,TEN-h
E = (5.3±0.5)×D(0.48±0.02)
T20,TEN-h
E = (6.9±0.6)×D(0.48±0.02)
T35,TEN-h
E = (9.1±0.7)×D(0.48±0.02)
T50,TEN-h
E = (11.4±1.0)×D(0.48±0.02)
534
https://doi.org/10.5194/nhess-2020-111
Preprint. Discussion started: 21 April 2020
c
Author(s) 2020. CC BY 4.0 License.
... The maximum precipitation takes place during the autumn and winter months, being December the rainiest month. Heavy storms are frequent, associated with intense rainfall and strong winds, with events measuring of up to 75 mm in 24 h (Melillo et al., 2020). Regarding vegetation, the island is the most deforested of the Archipelago with a predominance of scrubs well adapted to arid conditions. ...
Article
Full-text available
The island of Gran Canaria (Canary Islands, Spain) is characterized by a large variability of volcanic rocks, reflecting its volcanic evolution resulting from the built-up process of an intraplate oceanic island. The geological map provided by Geological Survey of Spain at 1:25.000 scale shows more than 109 different lithologies and it is too complex for environmental and engineering purposes. This work presents a simplified geotechnical map with a small number of classes grouping up units with similar geotechnical behaviours. The original lithologies were grouped using about 350 representative rock samples, collected in the seven major islands of the Archipelago. The samples were characterized by laboratory tests and in situ analysis. The geotechnical map was used to model and evaluate rockfall hazard in the entire island of Gran Canaria, where rockfalls are an important threat with a high social, economic impact. The rockfall map was validated with 128 rockfall events, occurred during the period 2010-2016, along the GC-200 road (34 Km), located in the NW sector of Gran Canaria. About 96% of the events occurred along sections of the road where the number of expected trajectories is high or moderate.
Article
Full-text available
Given its geological and climatic conditions and its rugged orography, Asturias is one of the most landslide prone areas in the North of Spain. Most of the landslides occur during intense rainfall episodes. Thus, precipitation is considered the main triggering factor in the study area, reaching average annual values of 960 mm. Two main precipitation patterns are frequent: (i) long-lasting periods of moderate rainfall during autumn and winter and (ii) heavy short rainfall episodes during spring and early summer. In the present work, soil moisture conditions in the locations of 84 landslides are analysed during two rainfall episodes, which represent the most common precipitation patterns: October–November 2008 and June 2010. Empirical data allowed the definition of available water capacity percentages of 99–100% as critical soil moisture conditions for the landslide triggering. Intensity-duration rainfall thresholds were calculated for each episode, considering the periods with sustained high soil moisture levels before the occurrence of each analysed landslide event. For this purpose, data from daily water balance models and weather stations were used. An inverse relationship between the duration of the precipitation and its intensity, consistent with published intensity-duration thresholds, was observed, showing relevant seasonal differences.
Article
Full-text available
Abstract Rainfall is the most important physical process responsible for the landslide triggering in Portugal. Results obtained worldwide have shown that control of rainfall on landslides differs substantially depending upon landslide depth and kinematics and the affected material. Therefore, the critical rainfall conditions for failure are not the same for different types of landslides, and may be strongly influenced by regional geologic and geomorphologic conditions. Rapid debris flows are typically triggered by very intense showers concentrated in just a few hours, and shallow translational soil slips are usually triggered by intense precipitation falling within the 1–15 days long range. On the contrary, activity of deep-seated landslides of rotational, translational and complex types is related to periods of nearly constant rainfall, lasting from several weeks to several months. The different rainfall intensity– duration conditions are associated with different hydrologic mechanisms for slope failure. The generation of surface run-off and high peak discharges in first-order mountain catchments is a critical triggering mechanism for debris flows. The intense rainfall allows the rapid growth of pore water pressure and the drop of capillarity forces that sustain the apparent cohesion of thin soils. As a consequence, shallow soil slips occur within the soil material or at the contact with the underlying less permeable bedrock. Long lasting rainfall episodes enable the steady rise of the groundwater table and the development of positive pore water pressures into the soil. Consequently, deep-seated failures occur in relation to the reduction of shear strength of affected materials. In this work, we present the state of the art concerning the proposition of empirical rainfall thresholds in Portugal for different types of landslides observed in different zones of the country: the Lisbon region, the Douro Valley and the NW Mountains, and the Povoação Municipality in São Miguel Island (Azores). The empirical thresholds applied in Portugal are based on the identification of past landslide events and include (i) the computation of antecedent rainfall threshold defined by linear regression, (ii) the normalization of rainfall by the mean annual precipitation, (iii) the definition of lower limit and upper limit rainfall thresholds and (iv) the definition of combined rainfall thresholds, which integrates the rainfall event and the antecedent rainfall for different time periods.
Article
We examined information collected from 395 reports of slope-movement events during about the past 150 years in Yosemite National Park, central Sierra Nevada, California, to identify the most prevalent types of slope movements and their triggering mechanisms. Rock slides and rock falls have been more numerous than debris slides, debris flows, and miscellaneous slumps. Rock falls have produced the largest cumulative volume of deposits. About half of slope movements had unreported or unrecognized triggering events. Earthquakes and rain storms individually accounted for the greatest cumulative volumes of deposits from recognized triggers of all types of historical slope movements; snowmelt, human activities and freeze-thaw conditions accounted for only a small proportion of the volumes from reported triggers. A comparison of the historical and postglacial average annual rates of deposition from slope-movement processes in a portion of the Yosemite Valley indicates that, during the period 1851–1992, slope-movement processes have been producing about half the average rate of deposits than during the past 15,000 years.
Landslide-triggering rainfall thresholds
  • J Cepeda
  • K Höeg
  • F Nadim
Cepeda, J., Höeg, K., and Nadim, F.: Landslide-triggering rainfall thresholds: a conceptual https://doi.org/10.3989/scimar.2001.65s121, 2001.
Meteorological Organization: Calculation of Monthly and Annual 30-Year Standard
  • Wmo -World
WMO -World Meteorological Organization: Calculation of Monthly and Annual 30-Year Standard