1. Introduction
The sub-basin of La Sabana River, located in Guerrero, a state in southern Mexico, occupies an inhabited area characterized by strong economic and social contrasts. The floodplain of this river has been almost completely covered by urban settlements of a medium–low socio-economic level in the northernmost regions, and of a higher level approaching the beach, where Acapulco, the famous tourist Mexican destination, is located. The population increase in the last 20 years has determined the uncontrolled growth of settlements that stick to the riverbed, which has led to an increase in vulnerability of this area [
1]. Linked to this increase in urbanization is the fact that the Mexican territory loses hundreds of hectares of forest and jungle every year due to an intense deforestation [
2,
3]. In particular, from 1970 to 2010 the state of Guerrero lost 110 km
2 of natural vegetation to produce building land, and it is estimated that the deforestation of the mountainous region of the state has doubled the susceptibility to flooding of the alluvial plains with particular consequences precisely on the Acapulco area [
1]. This, along with other problems such as the increase in non-formal economic activities and hydro-meteorological effects caused by climate change, has caused an increase in the susceptibility of the area to flood events, especially in correspondence with the occurrence of extreme weather phenomena.
As a matter of fact, due to its particular geographical location in the Pacific Ocean, the coast of the Guerrero state has been subject, over the years, to the impact of several hurricanes, some of which formed in front of the coast and others that landed at the port of Acapulco [
4]. In particular, for the period between 1970 and 2011, this state has suffered the direct impact of at least 24 tropical cyclones, highlighting the years 1974 and 1996 when three cyclones occurred in each season [
5]. During the 2015 hurricane season, there were 18 hurricanes in the Pacific Ocean, some of which had a direct impact on the coast of Guerrero [
6]. Among the events that occurred in the mentioned periods, the cases of hurricanes Pauline and Manuel are sadly well known. The first one occurred in October 1997. The global analysis of rainfall during this event indicates that in the La Sabana River Basin a maximum precipitation of 297 mm was recorded during a single day, equivalent to 1.9 times the average precipitation of the month of October [
7]. This was an event that caused a high environmental, economic and social impact, due to violent floods, water erosion and high mass transport of sediments, which resulted in the loss of human life, as well as an economic loss of the order of
$447.8 million of dollars [
7].
In September 2013, two simultaneous hurricanes affected Mexico: Manuel on the Pacific Ocean side and Ingrid on the Gulf of Mexico. This meteorological phenomenon produced torrential rains over a large part of the country that led to the overflow of various rivers. According to information from the World Meteorological Organization [
8], the antecedent of the presence of two tropical cyclones affecting both coasts of the country was in 1958, when Hurricane Alma, through the Gulf of Mexico, and Tropical Storm Number 2 in the Pacific ocean, struck at the same time in mid-June.
During the passage of Hurricane Manuel, the coast of Guerrero registered a considerable flood in various urban settlements and hotels located in the Acapulco tourist area. The accumulated rain sheet from 12 September to 17 September was 619.33 mm. The heavy rains forced the overflowing of La Sabana river towards the settlements located on the right bank, 6 km from the river mouth in the Tres Palos lagoon. The overflow verified after the first heavy rains, which suggests that the channel does not have sufficient hydraulic capacity to carry a volume like the one that accumulated during this hurricane [
9]. The two cited extreme events caused enormous damage to the entire floodplain, affecting both urban settlements in the northern portion and tourist areas towards the coast.
Extreme weather phenomena capable of producing heavy rainfall are far from being unlikely or infrequent at present. Emanuel [
10] finds an increase in the global average frequency of tropical cyclones in the range of 10–40% during the first three quarters of the 21st century. Along with the frequency, the average intensity of tropical cyclones is also increasing by ~15% [
11]. Shamir et al. [
12], in their review of current knowledge on climate change trends of precipitation, found that heavy rainfall events are expected to intensify as well as the frequency and intensity of large eastern pacific tropical cyclones. The increase of the storm and precipitation intensity are connected with an increase in the heat content of the oceans [
13], so the combination of warmer air and cold water can lead to an increase in rainfall [
14]. All these considerations are also in agreement with Knutson et al. [
15], who find, through the global simulation of a tropical cyclone using a High-Resolution Atmospheric Model, that while the global number of tropical cyclones tends to decrease in the warmer climate of the late 21st century, their intensity increases as does the precipitation rate.
On the basis of these evidences, it is therefore clear that there is a need to develop and apply methodologies that make it possible to identify and propose possible solutions capable of minimizing the risk of flooding in areas like La Sabana River floodplain, which are highly exposed to extreme weather phenomena, and lack both meteorological and maximum flood level monitoring systems. Note that this need is even more significant in areas of developing countries, such as Mexico, susceptible to severe flooding, where there is a lack of an extensive know-how, resources are limited and the scarcity of data make the situation even more complicated. Moreover, in Mexico there is not a policy particularly dedicated to the prevention of damage caused by extreme weather phenomena and the decentralization policy of the organizations dedicated to people protection [
16] makes the need to develop methodologies for the analysis of high-risk areas and for the determination of solutions aimed at the risk prevention and minimization more urgent. Areas at risk of flooding and without monitoring or extensive studies on the hydrological and hydraulic situation in Mexico are many, however, research focused at analyzing the individual and specific situations that cause flood risks is still few or fragmented. At the government level, the National Water Commission (CONAGUA), in 2014, established guidelines for the preparation of flood risk maps [
17], which led to the creation of a national risk atlas, available online at the Water Information Systems Portal [
18]. This portal provides hydro-meteorological information, vulnerability indices and flood susceptibility maps of some regions of the country. However, probabilistic studies aimed at characterizing the occurrence of the natural phenomenon, as well as at calculating the damage it can cause in the short and long range, are scarce, although this type of study is essential to provide the authorities with the necessary elements to define possible scenarios of risk and for adequate territorial planning. This derives from the hydrological complexity of the country, which makes an integral study of the whole territory very complex, but also from the lack, in many risk zones, of rivers monitoring.
Methodological approaches aimed to prevent damage caused by floods or to minimize them are varied: starting with the construction of hazard maps, which are a fundamental tool for prevention plans given their strong space-time component [
19], then passing through the use of data driven models, which are useful for prediction purposes with the great advantage of requiring minimal inputs [
20,
21], arriving at hydrological, and hydrological and hydraulic combined models that allow, among other things, to determine rainfall thresholds and critical flood levels. Deriving rainfall thresholds is the most commonly used method for developing flash flood warning systems, because it provides information on the amount of rainfall that, for a given duration and a given basin, can cause a flood [
22]. The methodologies developed in recent decades to calculate rainfall thresholds responsible for triggering floods are based on the use of distributed and semi-distributed hydrological models [
23,
24]. The combination of a hydrological study and a hydraulic model not only allows to accurately predict the design hydrographs of the basin but, very important, consent to simulate flooding in the alluvial plain in order to analyze the specific terrain response to flooding [
25,
26]. This combined methodological approach is particularly useful in the case of ungauged rivers where in-situ measurements of river critical levels as well as flood susceptibility maps are lacking.
The main objective of this study is therefore to demonstrate the usefulness of applying a hydrodynamic numerical model to characterize the terrain response to a river flood event in areas where there are no meteorological or flood levels monitoring systems. Hydraulic simulations can provide detailed information on the potential flood points of a river, even in correspondence with non-extreme weather events, and can also indicate which critical tie rods can lead to the flooding and the respective flood levels in the floodplain. The application of such methodology, in countries like Mexico where there is no free access or the availability of observations deriving, for example, from radar or meteorological satellites covering the whole territory, can certainly facilitate the characterization of flood zones, to improve the preparedness to flood scenarios. The application of a hydraulic model, therefore, can facilitate, both at the local and at the basin scale, the analysis of the area with respect to the risk of flooding, because it can identify the causes that can facilitate the runoff and the consequences that derive from it. This preliminary analysis of the territory can be carried out without the need for on-site inspections or without resorting to expensive on-site monitoring tools; a prospect that is advantageous in flood-prone areas of developing countries.The importance of demonstrating the reliability and practicality of the application of hydraulic models lies, finally, in the trend, for the future, to implement high-performance hydrodynamic models that predict full-scale river flood processes starting from the source, that is, using the rainfall data as input. In recent years, scientific advances in high-throughput computation, management and file format for data transfer have made the application of hydrodynamic numerical models more precise and widespread. Currently, there is a great variety of tools for simulating free-sheet water flows in two dimensions, among which the most consolidated are Mike-21 [
27], and the various calculation modules of SMS which is equipped with 1D and 2D models for the calculation of flood levels, water quality and analysis of hydraulic structures [
28,
29]. Moreover, models that use finite volumes for solving the St. Venant equations in two dimensions [
30] are, among others, the latest version of Mike-21 and the Iber software [
31], which will be used for the hydraulic analysis of this work. The choice of this model is based on its extensive validation which has shown it to be adequate for solving various engineering problems and makes it particularly effective for the calculation of discontinuous flows and for areas with complex geometries.
Therefore, in this work a study of the alluvial plain of La Sabana River was carried out, applying a combined methodological approach (hydrological and hydraulic), with the aim of studying the terrain response of the floodplain in terms of flood levels and determining the main critical points of flooding and associate them with certain rainfall sheets and their duration. Specifically, the methodology of this work develops as follows: first of all, a hydrological study of La Sabana River basin was carried out, with the objective of calculating project hydrographs in correspondence with different return periods and different precipitation duration; these hydrographs were then be used as input for the hydraulic analysis of the study area using the Iber hydrodynamic model. Results not only provided the flood levels, but also allowed to identify the potential points of the river first flooding.
The reminder of this paper is organized as follows.
Section 2 presents a description of the study area, the hydrological study applied for the estimation of flood discharges within La Sabana catchment, and the description of the methodology used for numerical simulations of different inundation scenarios. In this section, in order to validate the model in the study area, the numerical simulation of the hurricane Manuel hydrograph is also presented, and results of the calculated flood levels are compared with values measured on site. Results of this work are shown in
Section 3, where inundation levels, first flood point and occurrence times in correspondence with two precipitation duration (10 and 30 min) and different return periods are described. Results discussion is presented in
Section 4.
3. Results
Figure 6 shows the maximum flood levels calculated for a precipitation lasting 10 min, for the return periods 10 and 1000 years. Results obtained for the other return periods can be seen in the
Appendix A. From a qualitative and general look at the two maps of
Figure 6 it can be observed that for the lowest return period (10 years), the flooding is minimal, with depths mainly less than 50 cm. At the greatest return period (1000 years) the flooded area increases covering a predominantly rural zone, with flood levels ranging from 50 cm to more than 1 m.
As it would be expected, the situation changes if we look at the effects of a longer lasting precipitation (30 min). In
Figure 7, it can be seen that, already in correspondence with the lowest return period (10 years), the flooded area is more extended, touching the inhabited areas in the southern part of the alluvial plain, with flood levels greater than 60 cm, while, still in the south, the rural area can be covered by flood levels that exceed 1 m in depth. Both the flooded area and the water depths increase in the highest return period (1000 years), with flood levels reaching 3 m in the inhabited area south of the river.
Simulations results also allowed the identification of the main potential river flood points and to obtain the time in which the first flood occurs based on the input hydrographs.
Figure 8 is a snapshot of the moment when the first river flood occurs and refers to the simulation of a rain with a 10 min duration for the 10 years return period. The moment of the simulation corresponds to the time-step 28,800 s (8 h), shortly after the peak flow in the hydrograph has been reached.
The overflow point is located slightly south of the residential settlement “Las Gaviotas”, the area where the river first flooded during Hurricane Manuel. As the return period increases, the overflow time is anticipated, occurring up to 2 h before the maximum flow is reached in the corresponding hydrograph. By increasing the duration of precipitation (30 min), other two flood points can be identified (
Figure 9), which correspond to the first area from which the river flooded during the hurricane of 2013.
With precipitation of longer duration, the overflow occurs about 2 h earlier than the time in which the peak flow is reached in the hydrograph, already with the lowest return period.
4. Discussion
Simulation results provided water depth values deriving from hydrographs constructed with two rainfall durations (10 and 30 min) and in correspondence with five return periods (10, 50, 100, 500 and 1000 years). The flood level maps shown in
Figure 6 and
Figure 7 and in
Appendix A show that critical flood scenarios occur at high return periods, confirming the tendency of the area to suffer the effects of important floods only when extreme meteorological phenomena occur, which involve intense and long-lasting rainfall. During less frequent events (return periods from 100 years upwards), the urban settlements located in the southern part of the alluvial plain, in particular “Las Gaviotas” and “La Marquesa” neighborhoods, as well as part of the “Llano Largo”, can flood with levels ranging from 1 m to more than 2 m. The flood area increases when scenarios originated by precipitation of longer duration are simulated, reaching part of the tourist area south of the river. These are the areas that were most damaged during the hurricanes that hit the zone. However, even for the lowest return periods, with the increase in rainfall duration it is possible to notice that the river floods occur at very specific points, as it can be seen in
Figure 9. The points identified by the hydrodynamic model actually correspond with the flooded areas during Hurricane Manuel, as reported in Mejía [
9]. This result once again confirms the reliability of the Iber model for the hydraulic terrain response analysis to flood phenomena. Furthermore, results of the numerical simulations also provide the time of delays with respect to the input hydrograph in which the first flood occurs. This element is important to have information on the relationship between the moment in which the maximum peak discharge is reached and the moment when the overflow occurs. For short-term rainfall, during the most frequent events (return periods from 10 to 100 years), the flooding occurs when the peak flow of the hydrograph is reached, while for infrequent events (return periods from 100 to 1000 years) flooding can occur up to one hour before the river reaches its maximum discharge. For a longer precipitation duration, overflow may occur 2 h before the time of the maximum peak, even at the lowest return periods. This result is significant for the study area, as it indicates that the river does not have the capacity to contain the maximum flow when a precipitation occurs with a rain sheet greater than 20 mm falling for 30 min. These considerations can have an important implication for both long- and short-range prevention. For the purpose of long-term prevention, this result tells us that if the river is no longer able to contain the maximum design flow, this is probably due to the change in land use of the basin which has determined a decrease in infiltration, and therefore it serves to plan actions that can reduce superficial runoffs. The overflow of the river in specific points of the alluvial plain can also be a consequence of the expansion of the urban area which has reached the river margins in a few years. The study area has been heavily modified by anthropization over the last few decades, which has not only led to an increase in urban settlements, but also to the diversion of the river to obtain irrigation canals. The application of accurate numerical models and advanced methodologies aimed at combining the prediction of land use change with hydraulic models [
41,
42], in addition to methods that exploit analytical hierarchy processes for the construction of risk maps in urban areas [
43], has greatly improved in recent years, with the aim of accurately determining peak flow rates according to the forecasts of urbanization. For the specific results of this work, the application of the hydrological model has allowed the identification of specific points of overflow of the river in the urban area, providing information on the attitude of the river in case of flood phenomena, information that is more precise than the one arising from observations made after an inundation event. This represents a useful result for an unmonitored river and further applications of a hydrodynamic model could be devoted to calculating how much flood levels would increase if urban sprawl increases in an uncontrolled way, as already done by [
44], in another region of Mexico at high risk of flooding due to high urban expansion. For the purpose of short-range prevention, knowing the rain sheet that determines the first river overflow, represents another aspect that the application of a hydraulic model can have, as it could lay the basis for a study aimed at determining rainwater risk thresholds for the area.