2 Application of a Machine Learning Technique for Developing Short-Term Flood …
13
and destruction of infrastructure (bridges). In addition, hydrological droughts cause
water shortage problems, which obligate ETAPA EP to restrict the water supply of
the city and surrounding rural areas. In response to this necessity, the official flood
early warning system of the city was launched on 2014. It merely consists on monitoring (real time) the main currents at specific locations with the purpose to inspect
the hydrograph transit (Fernández de Córdova Webster and Javier Rodríguez López
2016). The limitation is the dependence on instrumentation which could be damaged
during extreme events. Additionally, the time in advance in which an alert can be
emitted is in the order of hours (maximum concentration time of less than 6 h for
the Yanuncay catchment). To date, there is no implementation of extreme runoff
forecasting models for operational and warning purposes.
The main flash-flood driving forces are precipitation, soil humidity (humid areas)
and topography (Braud et al. 2016; Ruin et al. 2008). However, in mountainous areas,
flood and drought forecasting is even more challenging considering that information
other than precipitation and discharge is not commonly available due to budget
constraints, remoteness of the study areas and more importantly due to extreme
spatio-temporal variability of the aforementioned driving forces. This is especially
true for the Tropical Andes in South America, the longest and widest cool region in
the tropics. Consequently, a simple approach, although useful, is the development of
precipitation-runoff forecasting models.
In 2015, the Sendai Framework for Disaster Risk Reduction (SFDRR) 2015–2030
(UNISDR 2015) was adopted at the Third UN World Conference in Sendai, Japan.
Seven global targets were proposed by 2030, in short, they are aimed to reduce disaster
mortality, affected people, disaster economic loss, disaster damage, and to increase
risk reduction strategies, international cooperation to developing countries, and early
warning systems to the people. Similarly, the International Council for Science
(ICSU) proposed a program entitled Integrated Research on Disaster Risk (IRDR). It
was launched to address the challenge of natural and human-induced environmental
hazards. Its first objective is devoted to hazard characterization, vulnerability, and
risk through development of forecasting capacities.
Moreover, the Science Plan of the IRDR program is related to geophysical,
oceanographic and hydrometeorological trigger events, flooding, storms, droughts,
climate change, etc. Both the SFDRR and the Science Plan of the IRDR claim the
development of studies aimed to forecast hazards (e.g., floods and droughts) and
to mitigate their associated impacts. Therefore, runoff forecasting-related studies fit
well with the scope of the initiatives previously mentioned.
As countermeasure against floods and droughts, forecasting activities have globally become an emerging field of research for water management and risk analysis
(Chang and Hwang 1999). The ultimate goal is to perform an integrated social,
economic, and environmental impact assessment to support the decision-making
regarding flood control policy (Brouwer and Van Ek 2004). Several different types
of models can be used for hydrological forecasting. Traditionally, fully distributed
models are enhanced to describe the physical processes of a catchment in a detailed
way. Nevertheless, severe data scarcity and the assumptions of the process-based
structure applied to specific catchments often limit the model operational value
13
and destruction of infrastructure (bridges). In addition, hydrological droughts cause
water shortage problems, which obligate ETAPA EP to restrict the water supply of
the city and surrounding rural areas. In response to this necessity, the official flood
early warning system of the city was launched on 2014. It merely consists on monitoring (real time) the main currents at specific locations with the purpose to inspect
the hydrograph transit (Fernández de Córdova Webster and Javier Rodríguez López
2016). The limitation is the dependence on instrumentation which could be damaged
during extreme events. Additionally, the time in advance in which an alert can be
emitted is in the order of hours (maximum concentration time of less than 6 h for
the Yanuncay catchment). To date, there is no implementation of extreme runoff
forecasting models for operational and warning purposes.
The main flash-flood driving forces are precipitation, soil humidity (humid areas)
and topography (Braud et al. 2016; Ruin et al. 2008). However, in mountainous areas,
flood and drought forecasting is even more challenging considering that information
other than precipitation and discharge is not commonly available due to budget
constraints, remoteness of the study areas and more importantly due to extreme
spatio-temporal variability of the aforementioned driving forces. This is especially
true for the Tropical Andes in South America, the longest and widest cool region in
the tropics. Consequently, a simple approach, although useful, is the development of
precipitation-runoff forecasting models.
In 2015, the Sendai Framework for Disaster Risk Reduction (SFDRR) 2015–2030
(UNISDR 2015) was adopted at the Third UN World Conference in Sendai, Japan.
Seven global targets were proposed by 2030, in short, they are aimed to reduce disaster
mortality, affected people, disaster economic loss, disaster damage, and to increase
risk reduction strategies, international cooperation to developing countries, and early
warning systems to the people. Similarly, the International Council for Science
(ICSU) proposed a program entitled Integrated Research on Disaster Risk (IRDR). It
was launched to address the challenge of natural and human-induced environmental
hazards. Its first objective is devoted to hazard characterization, vulnerability, and
risk through development of forecasting capacities.
Moreover, the Science Plan of the IRDR program is related to geophysical,
oceanographic and hydrometeorological trigger events, flooding, storms, droughts,
climate change, etc. Both the SFDRR and the Science Plan of the IRDR claim the
development of studies aimed to forecast hazards (e.g., floods and droughts) and
to mitigate their associated impacts. Therefore, runoff forecasting-related studies fit
well with the scope of the initiatives previously mentioned.
As countermeasure against floods and droughts, forecasting activities have globally become an emerging field of research for water management and risk analysis
(Chang and Hwang 1999). The ultimate goal is to perform an integrated social,
economic, and environmental impact assessment to support the decision-making
regarding flood control policy (Brouwer and Van Ek 2004). Several different types
of models can be used for hydrological forecasting. Traditionally, fully distributed
models are enhanced to describe the physical processes of a catchment in a detailed
way. Nevertheless, severe data scarcity and the assumptions of the process-based
structure applied to specific catchments often limit the model operational value