Keywords

1 Introduction

Worldwide, landslide hazard scientists and mitigation workers seek to determine landslide type and distribution at multiple scales to build susceptibility zonation maps. To that end, they may compile landslide inventories and model landslide susceptibility by using a GIS (Washington State Department of Natural Resources (WSDNR), Forest Practices Division 2006; Hervás and Bobrowsky 2009; Slaughter et al. 2017; Shao et al. 2022). To estimate landslide hazard, the frequency, density, and geometry of landslides are recorded using historical or multi-temporal landslide inventory maps and their related geodatabases (Hervás and Bobrowsky 2009; Guzzetti et al. 2012; Slaughter et al. 2017; Du et al. 2020; Shao et al. 2022). This landslide mapping is the foundation for modeling landslide susceptibility and hazard.

In Mexico, local landslide mapping has been carried out by using GIS and remote sensing (Capra and Lugo-Hubp 2006; Legorreta-Paulín et al. 2013; Gaidzik et al. 2017; Murillo-García and Alcántara-Ayala 2017; Oliva-González and Gallardo-Amaya 2018; Montgomery et al. 2020). Despite these efforts, there are few landslide inventory maps—landslide hazard maps that systematically record the type, distribution, abundance, and hazard within any region of Mexico. In addition, one main issue in modeling landslide hazard is the lack of systematic comparison of methods to outline the advantages and limitations of mapping the spatial distribution of landslides and the landslide hazard. There is no standardized procedure to prepare inventory or hazard maps.

This is the case for Iztaccíhuatl volcano, the third highest mountain in Mexico (5215.128 m a.s.l.), which has great potential to produce landslides because of its large area of weakened deposits affected by steep slopes, high seasonal rainfall, and tectonic activity. The Río Xopanac on the western flank of Iztaccíhuatl volcano has been selected as a case study area. Landslides form major natural hazards in hilly terrain in the area and cause extensive damage to roads, human settlements, and agricultural land. Therefore, it is important to prepare a landslide inventory map and a hazard map of the region.

The main goal of this work is to develop a multi-temporal landslide inventory map and a landslide hazard zonation map per landform to provide a standard methodology for mapping that supports governmental authorities in planning and mitigation in Mexico. During this study, 345 landslides covering 0.203 km2 were mapped from Google Earth imagery, aerial photographs, and GIS thematic layers (such as elevation, slope, hill shade, aspect, and geology), and verified in the field. A map of landslide hazard per landform unit was prepared by using the Landslide Hazard Zonation (LHZ) Protocol of the WSDNR, Forest Practices Division (WSDNR, Forest Practices Division 2006), supported by GIS. Derivation of landform units used classification of morphometric parameters, expert knowledge, and field verification. This analysis divided the watershed into 13 mass wasting landforms that were assigned slope stability hazard ratings from low to very high. For each landform the landslide area rate (LAR) and the landslide frequency rate (LFR) were calculated, as well as the overall hazard rating for the watershed. The overall hazard rating for the study area was found to be very high.

2 Study Area

The Río Xopanac is at 19°08′05″-19°09′06″ N latitude and 98°32′02″-98°26′03″ W longitude, on the western flank of Iztaccíhuatl volcano, within Puebla state, Mexico (Fig. 1). The volcano is in the eastern part of the Trans-Mexican Volcanic Belt (TMVB) physiographic province. The TMVB is an active volcanic chain that extends 1000 km in an approximately west-east direction, from the Pacific Ocean to the Gulf of Mexico. The dormant Iztaccíhuatl volcano is the third highest peak (5215.128 m a.s.l.) in Mexico. The volcano presents a great potential threat for the formation of landslides triggered by rain, earthquakes, and anthropic activity. The study area covers 6.6 km2 with an elevation range from 2304 to 2880 m a.s.l. and hillslopes between <5° (inner valleys of relatively flat plains) and 66° (mountainous terrain). Climate is classified as Subtropical semi-cold (C(E)(w2)) at 2300–2750 m a.s.l., and Subtropical temperate, subhumid (C(w2)) at >2750 m a.s.l. (INEGI 2008). The Río Xopanac erodes through Tertiary and Quaternary volcanic avalanche deposits, pyroclastic flows, lahars, and fall deposits (García Tenorio 2008). The study area is prone to landslides due to its large area of weathered and/or disaggregated material, under high seasonal rainfall and earthquakes. Also, the area is prone to landslides due to deforestation and agricultural activity. In the study area, episodic evacuation of debris by shallow mass movement takes place along the watershed. The steep hills capped with ash and pyroclastic deposits are affected by active and dormant deep-seated landslides, and where the stream erodes compact pyroclastic deposits, rock falls have occurred.

Fig. 1
A location map of the study area and a zoomed in map of the same. 1. It is located on the southeastern state of Puebla in Mexico. 2. It lies between 19 degrees 8 minutes and 9 minutes latitude and negative 98 degrees 32 minutes and 26 minutes longitude.

Study area, Rio Xopanac, State of Puebla, Mexico

Full size image

3 Method

Landslide data were collected from background information, Google Earth imagery, two sets of aerial photographs, and fieldwork, to create a multi-temporal landslide inventory map. Background information included roads, topographic, and geologic hardcopy maps at a scale of 1:50,000, and hardcopy maps of land use, climate, and hydrology at a scale of 1:250,000. All hardcopy maps were converted to a 10 m raster format, georeferenced, and incorporated as GIS layers into ArcMap. Google Earth images were from 2001, and two sets of aerial photographs and orthophotos were from 1993 and 1995 at a scale of 1:20, 000. A 3-m LiDAR digital elevation model (DEM), and its derived slope angle, slope curvature, and vertical erosion maps were used.

Information was collected from these sources to establish a generalized characterization of landslide processes and to aid analysis, interpretation, and mapping of mass wasting potential per landform.

Fieldwork was conducted twice in 2002 along the main river and in some tributaries. Field verification supported the existence of 15% of all mapped landslides, which enhances confidence in the assessment from images. Landslides were mapped and classified into shallow undifferentiated landslides, debris flows, debris slides, deep-seated landslides, earthflows and rock falls, according to the landslide hazard zonation protocol (2006) of WSDNR, Forest Practices Division 2006. Shallow undifferentiated landslides are characterized by very shallow failures composed mostly of soil. They are called undifferentiated because satellite imagery and/or aerial photograph interpretation alone may not be able to differentiate between various types (debris slide, debris flow) and processes, including slump, translational, or flow (Sarikhan et al. 2008).

All landslides were mapped onto photo transparencies and digitized directly onto the screen into GIS at the same scale as the photographs. In parallel with the digitizing, a spatial geo-database of landslides was constructed. Pertinent attributes of mapped landslides were recorded in the GIS geo-database, including: 1) a slide-specific identification tag, 2) type of mass wasting process, 3) degree of certainty of the observation (definite: the (human) originator of the landslide information is certain that this is a landslide; probable: the originator of the landslide information is almost certain that this is a landslide; questionable: the originator of the landslide information is not certain that this is a landslide, but is including it for completeness of the inventory), 4) landslide size (length, width, and depth), 5) landslide area, 6) landslide activity (active, inactive or relict), 7) slope shape (divergent, convergent, or planar), 8) field slope gradient, 9) map gradient measured from the 3 m DEM, 10) land use, 11) elevation at point of origin of the landslide, 12) landform, 13) date of landslide identification in the field or on aerial photograph, 14) field or aerial photograph identification number, and 15) researcher’s comments (WSDNR, Forest Practices Division 2006).

In parallel with the landslide mapping, areas of similar landslide potential were grouped into distinct landforms. Specific landforms that exist across the study area were defined by rules adopted by Washington Forest Practices Protocol to address landslide hazards (WSDNR, Forest Practices Division 2006). These landforms are called rule-identified landforms (inner gorges, bedrock hollows, convergent headwalls, outer edges of meanders, and active scarps of deep-seated landslides). Their differentiation is based on slope gradient and shape, lithology, landslide density, and sensitivity to forest practices.

GIS layers and fieldwork were also used to identify other areas that do not meet the DNR’s rule-identified slope landform definitions. These areas are called non-rule-identified landforms (such as non-rule-identified inner gorges, non-rule-identified bedrock hollow, and convergent headwalls). For instance, both rule- and non-rule-identified bedrock hollow landforms are colluvial, bedrock or simple hollows with concave spoon-shaped topography. Mapping them as one or as another landform depends only on the slope. Rule-identified bedrock hollows have slopes steeper than 70%, whereas non-rule-identified bedrock hollows involve slopes of 30–70% (WSDNR, Forest Practices Division 2006; Sarikhan et al. 2008). Both rule- and non-rule-identified landforms were entered into GIS as part of a landform polygon feature.

A landform hazard map per landform, as well as the overall hazard ratings for the entire study area were produced based on a semiquantitative approach (WSDNR, Forest Practices Division 2006). By default, all rule-identified landforms received a high hazard rating. They may later be “upgraded” to be flagged as having “very high” hazard on the basis of a semi-quantitative assessment. For both rule- and non-rule-identified landforms, the semi-quantitative overall hazard rating is derived from values that correspond to the landslide area rate (LAR) and landslide frequency rate (LFR) (Fig. 2a). The LAR is the area of landslides within each landform, normalized by the period of time spanned by the aerial photographs used, and by the area of each landform. These values are then heuristically multiplied by one million and rounded for easier interpretation (Powell 2007). The LFR is calculated similarly, however, the number of landslides is used instead of the area of landslides. This procedure is restricted to those landslides that deliver sediment to public resources. After the LAR and the LFR were determined, each landform was assigned a Low, Moderate, High, or Very High hazard rating (Fig. 2b). The DNR established the hazard rating class boundaries for LAR and LFR through limited applications in some watersheds of Washington State (WSDNR, Forest Practices Division 2006; Powell 2007; Sarikhan et al. 2008). The LAR and LFR values were then entered into a matrix (Fig. 2c) to determine the Overall Hazard Rating to be assigned to the landforms and the watershed. For instance, the rule-identified bedrock hollows landform has a LFR value of 32,649.8 (32 landslides/32.67 acres/30 years × 1000000) and a LAR value of 3805.7 (3.73 acres/32.67 acres/30 years × 1000000). Each of these values is classified as very high (Fig. 2b). Based on the plotting of these two qualitative values from Fig. 2b into Fig. 2c, this landform has a Very high landslide susceptibility.

Fig. 2
A process flow diagram of landslide inventory and landform hazard map in 3 parts. a to c include obtaining L A R and L F R, a table giving the respective entries for L A R and L F R for low to very high qualitative ratings, and a matrix of L F R by L A R to get the overall landform hazard ratings.

General procedure for producing landslide inventory and landform hazard map: (a) obtain LAR and LFR. (b) qualitative ratings matrix for numerical LAR and LFR. (c) assign overall hazard rating for each landform and for the watershed

Full size image

4 Results

During assessment of the Río Xopanac, a representative sample of 362 mass-wasting features was inventoried (Fig. 3). Five types of mass wasting processes were identified: 34.1% were shallow undifferentiated failures, 45.9% were debris flows, 6.9% were debris slides, 13% were deep-seated landslides, and 0.3% were earthflows. Of the 362 inventoried landslide features, 94.5% were classified as definite, and 5.5% as probable (Table 1).

Fig. 3
A landslide inventory map of Rio Xopanca. Debris flow, debris slide, and earth flow is from west to east.

Landslide inventory map

Full size image
Table 1 Landslide features mapped in the Río Xopanac watershed
Full size table

Thirteen landforms were recognized from physical attributes of the landscape (Fig. 4 and Table 2). They may be used to predict areas within the watershed that pose hazards for mass wasting, and are as follows:

  1. 1.

    Inner gorges cover 1.89% of the watershed area and represent 3.3% of the mapped landslides. Their potential for mass wasting and delivery is very high. The inner gorges can be either asymmetrical or symmetrical and may be intermittent lateral extent. Slopes are generally >70%, although the failure area of these gorge landforms may include convergent to planar slopes of <65%.

  2. 2.

    Bedrock hollows cover 1.99% of the area and represent 9.1% of the mapped landslides. Their potential for mass wasting and delivery is very high. Bedrock hollows are also called colluvium-filled bedrock hollows, swales, bedrock depressions, or simply hollows. They are long, pointed ellipses or round in planform, and on a hillslope occur as inverted spoon-shaped features. These features are found primarily on convergent slopes but can also occur on planar slopes. They are often found upslope from inner gorges and on steep slopes (>70%). Bedrock hollow failures can also occur on less steep terrain.

  3. 3.

    Convergent headwalls cover 0.81% of the area and represent 3.6% of the mapped landslides. Their potential for mass wasting and delivery is very high. Convergent headwalls are funnel-shaped landforms, broad at the ridge top and terminating where headwaters converge into a single channel. A series of convergent bedrock hollows may form the upper part of a convergent headwall. They are often found upslope from inner gorges and on steep slopes (>70%). This landform unit is generally uniformly scattered in the hilly terrain throughout the basin. The region mapped into this landform unit likely contains un-mapped inner gorges and bedrock hollows.

  4. 4.

    Active scarps of deep-seated landslides cover 2.92% of the area and represent 10.5% of the mapped landslides. Their potential for mass wasting and delivery is very high. This landform encompasses active head and lateral scarps of deep-seated landslides. There is a high potential for additional secondary landsliding within the entire area (headscarp, body and toe) of active deep-seated landslides due to post-landslide steepening by streams that cut through the features.

  5. 5.

    Meander Bends/Overbank deposits/Surge Plains cover 0.13% of the area and represent 4.7% of the mapped landslides. Their potential for mass wasting and delivery is very high. In the watershed, slope failures occur through stream undercutting of the outer banks of meanders along valley walls or high terraces of an unconfined meandering stream.

  6. 6.

    Flats cover 0.24% of the area, but no landslides have been mapped in this landform. Their potential for mass wasting and delivery is low. They include slope forms that are mostly gentler than 10%. Although the landform includes gentle slopes, there are occasional cuts with slopes >62%. This category consists predominantly of valley bottoms, floodplains, and flat terrace surfaces.

  7. 7.

    Low-gradient hillslopes cover 18.67% of the area, represent 5% of the mapped slope failures, and include all slope forms (convergent, divergent, and planar) and gradients between 11% and 40%. Their potential for mass wasting and delivery is high.

  8. 8.

    Moderate-gradient hillslopes cover 31.21% of the area, represent 32.6% of the mapped slope failures, and include all slope forms (convergent, divergent, and planar) with gradients that range between 41% and 70%. Their potential for mass wasting and delivery is very high.

  9. 9.

    Steep-gradient hillslopes cover 11.45% of the area, represent 22.9% of the mapped slope failures, and include all slope forms (convergent, divergent, and planar) with gradients of >70%. This category contains other very-high-hazard landforms (e.g. inner gorges and bedrock hollows) that are the initiation points for many debris slides and debris flows. The potential for mass wasting and delivery is very high.

  10. 10.

    Ridges tops and noses cover 24.99% of the area, represent 0.3% of the mapped slope failures, and include all ridge tops and ridge noses with gradients between 0% and 10%. Landslides have occurred below and adjacent to some of these low-gradient ridge tops. Their potential for mass wasting and delivery is low.

  11. 11.

    Non-rule-identified inner gorges cover 5.34% of the area and represent 5.5% of all slope failures. Their potential for mass wasting and delivery is very high. This category has characteristics similar to those of rule-identified inner gorges but involves convergent slopes of 30–70% gradient. They appear as gentle-slope-walled canyons or gullies on DEM slope maps, and seem to have been eroded by stream action, but with evidence of mass wasting along their sidewalls. They may be either symmetrical or asymmetrical in profile and are commonly intermittent in lateral extent.

  12. 12.

    Non-rule-identified bedrock hollows cover 0.23% of the area and represent 0.8% of the mapped landslides. Their potential for mass wasting and delivery is very high. This category has characteristics similar to those of rule-identified bedrock hollows, but involves convergent slopes of <70% gradient. This landform is the initiation point for many debris slides and debris flows.

  13. 13.

    Non-rule-identified convergent headwalls cover 0.13% of the area and represent 1.7% of the mapped landslides. The potential for mass wasting and delivery is very high. This category (slope <70%) includes bedrock hollows and inner gorges that are usually separated one from another by steep ridges. The slope geometry within this landform concentrates both surface water and groundwater, which produces areas of positive pore pressures during intense or extended precipitation events.

Fig. 4
A landform distribution map of the Rio Xopanac watershed highlights 12 features. It includes deep-seated landslides lodged between moderate to steel-gradient hillslopes with a central stretch of inner gorges, west-east.

Landform distribution in the Rio Xopanac watershed

Full size image
Table 2 Landslide hazard rating of landforms
Full size table

5 Discussion and Conclusions

The implementation and development of a GIS-based landslide inventory map and its related geo-database, in conjunction with a landslide hazard zonation methodology is fundamental in prognostic studies of slope instability. In this paper we briefly introduce the implementation of a method for mapping and modeling landslide hazard by adapting the WSDNR, Forest Practices Protocol. The landslide inventory shows that, in the study area, shallow landslides (including debris slides and debris flows) are the predominant type (86.7%), followed by deep-seated landslides (including earthflows) (13.3%). Along this stretch of the Rio Xoponac on the western flank of Iztaccíhuatl, we recognized 13 landform units. Two landforms (Moderate- and Steep-gradient hillslopes) encompassed more than a half of the landslides (55.5%), followed by active scarps of deep-seated landslides (10.5%), bedrock hollow (9.1%), and inner gorge (8.8%; including Rule- and Non-rule-identified) landforms.

Field observations and satellite and aerial imagery suggest that a predisposition to trigger landslides may stem from the combined effect of land-use changes, geological conditions, and the occurrence of very highly susceptible landform units. For example, the Moderate-gradient and Steep-gradient hillslopes units had very high landslide susceptibility. These landforms, originally covered by forest, are on fertile weathered and unconsolidated soils that have favored a conversion of large areas to agriculture; however, unsuitable treatment of the remaining forest has triggered the high incidence of gravitational processes and erosion. The landforms are more prone to landslides in the middle and upper portions of the watershed, where there are steep slopes, loose volcanic ash, pyroclastic deposits and a buffer width of only 50 ̶100 m of forest along the hillsides.

This finding is important in understanding the long-term evolution of the stream system on the western flank of Iztaccíhuatl volcano, and may prove useful in the quantification, assessment, and modeling of landslides that occur continually in volcanic terrains. The coalescence up-stream of landslides in the watershed has increased the destructive power of mass wasting processes that threatens towns such as Santa Maria Nepopualco, Santa Maria Tianguistenco, and Cholula, with a total population of >141,000 people.

We acknowledge that there is a more quantitative geomorphological method to define and characterize landform units. However, the production of morphometric maps is time consuming, and requires expert knowledge in geomorphology. The Landslide Hazard Zonation Protocol of the WSDNR, Forest Practices Division used here was preferable; it focuses on a more generic and rapid approach to allow the definition and classification of volcanic landform units of the study area. Nevertheless, the resulting landslide hazard classes allow comparison between landforms because their semi-quantitative value is normalized by the landform size. The landform units and the hazard classification could also be correlated to the volume delivered, and may therefore be useful in the quantification, assessment, and modeling of debris flows in neighboring areas.

Future research should involve the assessment and modeling of landslide volume per individual landform, and the calibration of the WSDNR, Forest Practices Division, method to other Mexican watersheds. Knowledge of the landslide inventory and hazard mapping issues is essential in informing policies determined by local authorities in Mexico.