12
P. Muñoz et al.
for optimal inputs composed only by features accounting for 80% of the model’s
outcome variance. Moreover, we found that a set of RF hyper-parameters can be
transferred to a comparable catchment with a maximum model performance reduction of 0.10 (NSE). Overall, the forecasting of hydrological extremes (especially
floods) is challenging mainly due to lack of relevant data (driving forces) and sufficient extreme events from which RF models learn. The applicability of this study is
to assist authorities in flood and drought management to evaluate hazard risks and
to found the basis for developing integrated action plans from a local and regional
perspective.
Keywords Natural hazards · Early warning system · Flood · Drought · Timeseries
forecasting · Hydroinformatics · Machine learning · Random forest
2.1 Introduction
Floods and droughts are among the most common natural disasters worldwide.
Both hydrological extremes have major impacts on society (e.g., human losses,
increased health risks, interruption of water and sewer services), economy (e.g.,
losses of agricultural production, damage to infrastructure), and ecosystems (e.g.,
hydro-geomorphic conditions, alteration of river and floodplain habitats and biodiversity) (Vos et al. 1999). Even worst, recent studies concur that the frequency and
severity of floods are expected to increase with climate change and land use changes
(Min et al. 2011; Sofia et al. 2017). For instance, in the Andes of Ecuador, flood and
drought events cause human losses and perturb the everyday life of people by interrupting the water supply service, damaging transportation networks, among others
(Chang and Hwang 1999).
A report of the Andean community
1 for the period 1970–2007 revealed that in
the Andes of Ecuador 263 floods, 30 droughts, and 357 landslides (as a side effect,
mostly in the city of Cuenca) caused 429 human deaths as well as destruction of 2149
houses. Cuenca, which is the third largest city of Ecuador (almost 0.6 million inhabitants), is crossed by four rivers, the Tomebamba, the Yanuncay, the Tarqui, and the
Machángara. Together, the Tomebamba and the Yanuncay upper catchments provide
nearly the 60% of the water demand of the city. Additionally, around 85% of the total
area of the upper catchments is covered by páramo and native forests ecosystems.
However, the water regulation function of these ecosystems is increasingly amended
by natural and human processes (Buytaert et al. 2011, 2006). The vulnerability of
the catchments led to declare them legally as protected zones
2 in 1985.
According to the local water company, the Empresa Pública Municipal de Telecomunicaciones, Agua potable, Alcantarillado y Saneamiento de Cuenca (ETAPA
EP), on an annual basis, the city of Cuenca is affected by flood and hydrological
drought events. Local media has reported inundation events causing human losses
1 https://www.comunidadandina.org
2 https://www.etapa.net.ec
P. Muñoz et al.
for optimal inputs composed only by features accounting for 80% of the model’s
outcome variance. Moreover, we found that a set of RF hyper-parameters can be
transferred to a comparable catchment with a maximum model performance reduction of 0.10 (NSE). Overall, the forecasting of hydrological extremes (especially
floods) is challenging mainly due to lack of relevant data (driving forces) and sufficient extreme events from which RF models learn. The applicability of this study is
to assist authorities in flood and drought management to evaluate hazard risks and
to found the basis for developing integrated action plans from a local and regional
perspective.
Keywords Natural hazards · Early warning system · Flood · Drought · Timeseries
forecasting · Hydroinformatics · Machine learning · Random forest
2.1 Introduction
Floods and droughts are among the most common natural disasters worldwide.
Both hydrological extremes have major impacts on society (e.g., human losses,
increased health risks, interruption of water and sewer services), economy (e.g.,
losses of agricultural production, damage to infrastructure), and ecosystems (e.g.,
hydro-geomorphic conditions, alteration of river and floodplain habitats and biodiversity) (Vos et al. 1999). Even worst, recent studies concur that the frequency and
severity of floods are expected to increase with climate change and land use changes
(Min et al. 2011; Sofia et al. 2017). For instance, in the Andes of Ecuador, flood and
drought events cause human losses and perturb the everyday life of people by interrupting the water supply service, damaging transportation networks, among others
(Chang and Hwang 1999).
A report of the Andean community
1 for the period 1970–2007 revealed that in
the Andes of Ecuador 263 floods, 30 droughts, and 357 landslides (as a side effect,
mostly in the city of Cuenca) caused 429 human deaths as well as destruction of 2149
houses. Cuenca, which is the third largest city of Ecuador (almost 0.6 million inhabitants), is crossed by four rivers, the Tomebamba, the Yanuncay, the Tarqui, and the
Machángara. Together, the Tomebamba and the Yanuncay upper catchments provide
nearly the 60% of the water demand of the city. Additionally, around 85% of the total
area of the upper catchments is covered by páramo and native forests ecosystems.
However, the water regulation function of these ecosystems is increasingly amended
by natural and human processes (Buytaert et al. 2011, 2006). The vulnerability of
the catchments led to declare them legally as protected zones
2 in 1985.
According to the local water company, the Empresa Pública Municipal de Telecomunicaciones, Agua potable, Alcantarillado y Saneamiento de Cuenca (ETAPA
EP), on an annual basis, the city of Cuenca is affected by flood and hydrological
drought events. Local media has reported inundation events causing human losses
1 https://www.comunidadandina.org
2 https://www.etapa.net.ec