1. Introduction
Landslide (LS) is a general term that includes: (1) downhill movement of pebbles, stones of small to bigger sizes from hillslopes, (2) rockslide, (3) rock-fall, and (4) debris flow. Landslides (LSs) occur at many places in the Western Ghat (WG) mountain regions of Maharashtra state in India (
Figure 1) in the summer monsoon (June–September) season. The word “Ghat” is in local Indian languages and has three meanings. These are (1) approaching the road to the river with stony steps, (2) roads in mountain areas, and (3) mountains. When it is used in the sense of a road in the mountain area, the name of the city is prefixed with it, indicating the road connecting to that city. The word “Ghat” is used as an adjective for mountains on the west coast of India, referred to as “Western Ghat (WG)”.
In the past, information about LS events remained confined to the local areas and hence has missed the attention of the people at large. On 30 July 2014, an LS occurred in the village of Malin (19.16° N, 73.68° E) in the Ambegaon taluka of the Pune district in Maharashtra, India [
1,
2,
3]. The author of [
4] reported in his blog that it was the second-worst LS of the year till that date. Malin is about 150 km from Pune (18.5° N, 73.8° E) city. The Malin LS wiped out the whole village except a primary school with a cement concrete structure. The LS at Malin occurred early in the morning while residents were in a deep sleep. A total of 151 people were killed. In addition to the reported deaths, a total of 160 people were believed to be buried under the debris. Accumulated rainfall of 215 mm was recorded over the area in the 7 days prior to the incident. The incident was widely reported in the press and electronic media.
In the WG mountain region, LS events are found to occur even with one-day very extreme rainfall activity. As per the India Meteorological Department’s (IMD) definition, the term “extreme” is used if the rainfall in a 24-h period is more than 200 mm. There is no category in the IMD classification for one-day rainfall of more than 200 mm. Here, we define the “high rainfall” category if the 1-day rainfall is 300 mm and more or accumulated in a 7-day period. A large number of deaths and losses occurred in the past in high rainfall events. The two events worth quoting in this category are (1) the Ghatkopar (19.08° N, 72.91° E) LS event on 12 July 2000. Ghatkopar is a suburb in the thickly populated city of Mumbai. Mumbai is the capital city of the state of Maharashtra. The 24-h rainfall was 350 mm. A total of 78 people died. (2) The Jui (18.0° N, 73.4° E) LS event on 25 June 2005. Jui is a village in Mahad district on the west coast of Maharashtra. The 16-h rainfall was 476.02 mm. In this LS event, three villages in the surrounding area were completely buried. The number of reported human deaths was 48. These are two representative examples of disastrous LS events which occurred due to high rainfalls before the Malin 30 July 2014 LS event. However, such LS events were overlooked by the citizens, media, and government authorities and did not attract serious attention from the geological and meteorological community in India. Therefore, studies regarding LSs over the WG region are almost non-existent before the Malin event.
The Malin LS event drew people’s attention from different sections of society, including print and electronic media, geologists, and meteorologists. The reasons are improvement in communication systems and increased awareness about the severe losses of human lives and materials. After the Malin incident, the Geological Survey of India (GSI) initiated an action to identify LS-prone sites in the WG [
5]. The Malin LS event was an eye-opener for taking a serious note about the impacts of LS events. It raised concerns about the safety of the people residing in such LS-prone areas. In recent times, in the monsoon season, the frequency of extreme events has increased due to climate change impact [
6]. Further, vulnerable LS-prone areas have increased many times in the WG region due to increased anthropogenic activities, such as deforestation, improper land use planning, and road and building constructions on hillslopes. These have exposed people residing in such vulnerable areas to more losses and deaths in recent times. As a consequence, demand arose in all the sections of society for identifying LS-prone areas and, if possible, an early warning system of LS occurrences over the WG region. The early warning system is considered to be a proactive measure in reducing the death toll by evacuating the people staying in vulnerable areas.
The factors which drive LSs are grouped into two categories. These are (1) preparatory variables and (2) dynamic triggering. The preparatory variables include topography, tectonics, the geological history of the region, land use, anthropogenic activities such as deforestation, terracing of the hills for agriculture, and construction activity. The construction activity is a continuous process required to develop the roads, railway tracks, and dams in the mountainous areas in the WG region. The dynamic triggering variables include heavy rainfall, earthquake, and coastal erosion [
7]. The scale of damages by an earthquake is very large, and an LS is a small part of the total damages by the earthquake. The earthquake phenomenon shows no regularity and is therefore impossible to predict well in advance. The LS events occur due to coastal erosion during the landfall of a cyclonic storm on the coast. The cyclonic systems form over the Indian region in pre-monsoon (April–May), sometimes in the early part of June, and post-monsoon (October–November) seasons. Therefore, in non-seismic and non-cyclonic conditions, rainfall activity in the monsoon season is the most common trigger responsible for LS occurrences in a large number and at many locations in the WG region.
The author of [
8] speculated the relationship between rainfall and the triggering of a shallow LS. Interestingly, in the present study region, apart from the term high rainfall, we define the term “medium rainfall” as the accumulated rainfall of the order of 50–100 mm in a week’s period. LSs occur in high as well as in medium rainfall conditions. The mechanisms that drive an LS and operate in these two rainfall regimes are: (a) in high rainfall cases, the rainwater percolates in the soil layer covering rock. After soil saturation by rainwater, some rainwater drains out through the channels in the rocks and soil cover. In some cases, if the channels on the surface are open, then rainwater drains out easily, forming waterfalls. Percolated rainwater to the bottom of the soil layer forms a thin layer of mud between the rock’s surface and soil resting on the rock. This decreases the frictional force between the two. The frictional force keeps the layer of deep soil in stable conditions on the surface of the rock. The shear strength at a point on the slope depends upon (1) the cohesive force of the material and (2) the total stress on the hillslope, which is proportional to the weight of pore-water [
7]. When the total weight of a volume of saturated soil exceeds the shear strength of rock, a sliding force is generated, overcoming the geological resisting force to trigger an LS. Another factor contributing the slope stability is the vegetation cover. High rainfall areas are favourable for the thick and widespread vegetation cover. The deep soil layer supports thick and widespread vegetation cover comprising tall trees whose roots go deep inside the soil layer. The thick vegetation and roots of big trees hold the soil together in stable conditions against the sliding force. At Malin, the soil thickness was 10 m. An important factor that was responsible for the LS in the Malin case was the terracing of the hillslope for agriculture activity. The soil and rocky material after terracing were dumped on the slopes. This choked the natural drainage channels and caused obstructions to the free flow of the rainwater. All the rainwater percolated in the soil 10 m deep. This destabilized the slope, resulting in the LS. The (b) LS occurrence in medium rainfall regime: As per the IMD classification, India is divided into 36 meteorological subdivisions (
Figure 2). The meteorological subdivisions of Madhya Maharashtra (MM) and Marathwada (MDA) lie on the leeward side of the WG mountains (
Figure 1). Monsoon seasonal rainfalls in these subdivisions are low of the order of 500–600 mm. These areas have low vegetation cover and have vegetation of shallow shrub types. The soil layer is thin, and therefore the roots of shallow shrubs extend only in the upper layers of the soil. The rainfalls of the order of 50–100 mm occur in the weekly period under favorable synoptic conditions in these subdivisions. As the soil layer is thin, very little rainwater is absorbed into the soil. Most of the rainwater flows downward on the surface of the mountain slope. The material on the slopes is loose and unconsolidated. The downward moving rainwater on the slope carries these materials with it, resulting in shallow LSs. However, these types of shallow LSs are not disastrous. If the rocky and soil materials fall on roads, they cause temporary road blockages.
The author of [
9] introduced the term “rainfall-intensity-duration (RID)” for triggering an LS. Since then, RID has been considered a useful parameter for predicting LS occurrence. The author of [
10] identified RID for debris flow in the central Santa Cruz Mountains, California. The authors of [
7] developed a real-time warning system for issuing warnings of LSs for the San Francisco Bay region using the RID method based on empirical and analytical relationships between rainfall and LS occurrence. They reported that the threshold of antecedent rainfall was 250 to 400 mm for the occurrence of an LS in the San Francisco Bay region.
The author of [
11] studied 73 global LS events to determine the relationship between the RID and LSs. The authors of [
12] developed RID thresholds for the Puerto Rico region. The authors of [
13] estimated the RID threshold for the North Island of New Zealand. The authors of [
14] used the RID method for predicting an LS in areas of the Himalayas of Nepal. The author of [
15] used RID thresholds for LS occurrences in the Garhwal Himalayas region of India. The authors of [
16] used the RID method for LS predictions over central and southern Europe.
The authors of [
17] prepared a global LS susceptibility map based on indices derived from six parameters, viz. (1) slope, (2) soil type, (3) soil texture, (4) elevation, (5) land cover, and (6) drainage density. Their map shows that the WG region has a high susceptibility index. The authors of [
18] prepared the global LS atlas (GLC) using three years (2003, 2007, and 2008) of LS data. They classified all the LS events on scales ranging from 1 to 5. Scale 1 indicates the types of LS that occur at single locations and on small slopes, and scale 5 indicates the types of LS that occur at steep slopes and multiple sites causing numerous fatalities. As per [
18]’s study, India, among other Asian countries, has a large number of LS occurrences and fatalities for all the years considered. India lies in the top-ranked countries based on a number of LS-related fatalities. Further, they reported that the catalogue could be further improved by integrating LS inventory studies at the regional levels.
However, studies regarding LSs over the WG region are few, except for [
1,
2,
4,
19,
20,
21]. The lack of studies may be attributed to (1) LS events occurring mostly at isolated remote locations that go unreported and occur only in monsoon season. This has developed a lack of seriousness toward such incidents in Indian society. (2) LS events are mostly scaled 1, meaning occurring over small slopes and at isolated sites. (3) Media reporting tends to be biased toward LSs with human casualties [
22]. (4) In India, studies related to LS events do not come under the roof of one department of government agencies as they occur due to combined meteorological and geological forcings. Therefore, neither the IMD nor India’s geological department has a dedicated group or unit to systematically gather and archive LS data and predict LS activity. In India, a good number of LS studies are carried out in the geography and geological departments of universities, however, these are scattered. The studies regarding predictions of LS events are lacking in India.
From the past data, it is noticed that LSs in the WG area occur in monsoon season during persistent high rainfall episodes for a week’s period. The thresholds of rainfalls are different for different locations. With the establishment of monsoon circulation over India, persistent rainfall for the period of a week and more occur due to conducive synoptic systems.
After the launch of the Tropical Rainfall Measurement Mission (TRMM) satellite in the year 1997, the rainfall data over the tropical region became available to researchers. TRMM Multi-satellite Precipitation Analysis (TMPA) rainfall data at 0.250 × 0.250 3-h intervals have shown promise in predicting LS events using satellite rainfall data. The authors of [
23,
24,
25,
26] developed an algorithm for LS predictions using LS susceptibility and satellite-derived rainfalls. The authors of [
27] created an online decision support system for forecasting potential LS activity in near real time, termed “Landslide Hazard Assessment for Situational Awareness”.
The fatalities and damages in the Malin landslide incident prompted the necessity of identifying the hotspots of LSs over the WG region and developing a real-time forecasting system to help the disaster management authority of the state government. LS forecasts on specific LS locations are much more useful to the disaster mitigating authority for taking precautionary actions. The preventive actions include pre-warning, waking residents living in vulnerable areas regarding the forthcoming disaster, and mass evacuation from the danger spots.
Internationally, interest in quantifying LS risk has been developed since the activity initiated by the International Association of Engineering Geology (IAEG) Commission on LSs to compile a list of worldwide LS events for the UNESCO annual summary of information on natural disasters in 1971 [
22]. Asia records the highest number of LSs, and a substantial number in the Himalayan region [
22], and the next region is the WG in India. Considering the societal requirement, and as per [
18]’s recommendation that LS studies on the regional scale are required, an inventory of LSs over the WG region is prepared using past data for the period of 2000–2016. A prediction system for LS occurrences in the WG region of Maharashtra was developed using the data of LSs and rainfalls for the period from 2000 to 2016. The system has been in operation since June 2017 under the name “Satark”. Satark is a word in the local language “Marathi” meaning “vigilant”. The LS forecasts are uploaded in real-time on the internet site
www.satarkindia.wordpress.com, accessed on 1 April 2022, for societal and governmental use. The initial results were presented at the National Space Science Symposium held in 2015 [
28].
The first objective of the paper is to present LS inventory over the WG region. This data may find a place in [
18] Global Landslide Catalog (GLC). The second objective of the paper is to present the real-time LS forecasting system “Satark” developed by CCS. To the best of our knowledge, this is the first effort of its kind in the Indian region.
The paper is divided into five sections. The geological conditions and rainfall distributions are important features of LS events.
Section 2 describes these features of the WG region.
Section 3 describes the materials and methods used in the study.
Section 4 provides an inventory of the LS events that occurred over the region. It also presents the “Satark” LS prediction system. Conclusions are given in
Section 5.
3. Materials and Methods
For the model development, archived TMPA rainfall data with spatial resolution 0.250 × 0.250 degrees (250 × 250 m) have been used to compute cumulative rainfalls for each location at the LS-prone sites. The IMD publishes the Indian Daily Weather Report (IDWR), which provides descriptions of synoptic systems present on every day. The IDWR data for the period 2000–2016 have been used to get large scale synoptic systems prevailing prior to LS events. The IMD issues medium-range (10-day period) and short-range (3-day period) rainfall forecasts over the Indian region using numerical models (
www.imd.gov.in, accessed on 1 April 2022). These have been used in real time to anticipate the occurrences of persistent high rainfall activity over the study area. An “S” band doppler radar is in operation by the IMD at Mumbai (19.1° N, 72.8° E
Figure 1). There is another radar at Panjim (15.5° N, 73.8° E) in the state of Goa, on the Indian west coast region, south of Mumbai, at a distance of approximately 600 km. Both these radars cover most of the study area. The radar products’, viz. spatial and vertical reflectivities, surface rainfall intensities (SRI) are used to get spatial and temporal distributions of convection and rainfalls at a high temporal resolution (every 11 min) over the study area and to supplement the rainfalls estimated from the satellite and inferred empirically from synoptic-LPS systems.
Persistent high rainfalls on the ILS region are due to convection generated by the HJ phenomenon [
38]. The HJ is estimated using the Froude (Fr) number. Fr = U/(NH). It is the ratio of zonal wind speed (U) to the product of Brunt Vaisala frequency (N) and the height of the mountain (H) [
38]. N is estimated using standard formula N =
. Daily values of potential temperature θ and gradient of potential temperature dθ/dz were computed using Mumbai radiosonde data available on the Wyoming site. The mean height of the Western Ghat mountains have been taken as 1000 m to estimate the potential temperature at the mean height of the Western Ghat mountains. If Fr ≥1, then wind flows over the mountain and experiences HJ. If Fr < 1, the winds go around the mountain. The height of mountain H at each LS site on the ILS area varies between 800–1200 m above mean sea level. The value of N is of the order of 0.01 s
−1. The wind speed required to obtain Fr > 1 for these representative values of N and H is 20 ms
−1. Daily Fr values were estimated using daily radiosonde data for the period 2000–2016 of stations Mumbai and Goa from the Wyoming site (
http://weather.uwyo.edu/upperair/sounding.html, accessed on 1 April 2022) (from 1 June 2000 to 30 September 2016) for the model development and (from 1 June 2017 to 2021) for the period of 2017–2021 for real-time LS predictions.
The Center for Citizen Science (CCS) is a non-governmental organization established in Pune, India, involving citizens in scientific data collection to foster scientific research to address societal problems. The CCS installed 10 ordinary rain gauges in villages in the areas near to LS-prone sites in Raigad–Ratnagiri subareas in the WG region in 2021 in the monsoon season. Daily rainfall data in the monsoon season of 2021 from these rain gauges are ingested into the LS prediction system “Satark” developed by the CCS in the 2021 monsoon season. The LS predictions are made available to the public and media through the website
www.satarkindia.wordpress.com, accessed on 1 April 2022.
LS Prediction Model
It is seen that the maximum frequency of LS events is on the ILS of the WG mountains. High rainfalls over this region occur due to the hydraulic jump (HJ) phenomenon. Occurrence of HJ depends on the Froude number (Fr). Values of Fr are considered for estimating the probabilities of LS occurrences along with the IMD medium-, short-range weather forecasts, radar reflectivity data, and synoptic weather conditions.
The cumulative rainfalls curves are taken as reference curves for initiating predictions of LSs for those locations. The forecast system operates in three stages.
Stage 1 Watch: If the IMD forecast charts show presence of transient systems, southerly position of MT, an increase in rainfall over the LS-prone areas for the ensuing 7 days, and if Fr > 1, then the “watch” stage is invoked. The people in the area are sensitized to keep watch on the rainfall intensity and duration. The rainfall at LS locations from TRMM is monitored continuously.
Stage 2 Alert: If the IMD synoptic charts show presence and westward movements of transient systems and arrival of the systems near to the north of the Maharashtra state, southward displacement of MT, Fr > 1, and cumulative rainfalls for 3 days are higher than the reference cumulative rainfalls, then the “Alert” stage is invoked. The people in area are informed to be ready for moving out from their houses in LS areas.
Stage 3 Warning: If the high rainfalls persist for the next 3 days and cumulative rainfalls at LS locations are more than the reference values, then the prediction of LS is continued. Then, LS forecasts are given for the next day. The people in the areas are suggested to move to safe shelter locations with essential material to survive for a few days till the debris/stony materials fallen due to an LS are removed.
The real-time LS forecasts have been in operation in monsoon season for 5 years since 2017 and the forecasts are uploaded on the site (
www.satarkindia.wordpress.com, accessed on 1 April 2022) for societal and state government disaster management call. The disaster management department of the government of Maharashtra has started taking note of LS predictions since 2018.