1
TIFAC-IDRiM Conference
28th –30th October 2015
New Delhi, India
HOW OFTEN WILL A 1 IN A 1000 YEAR FLOOD OCCUR? THE 2014
KARNALI FLOOD IN WEST NEPAL
S. DUGAR1, S. BROWN2
1
Practical Action Consulting, Kathmandu, Nepal, <Sumit.Dugar@practicalaction.org.np>
2
Practical Action Consulting, Rugby, UK, <Sarah.Brown@practicalaction.org.uk>
ABSTRACT
Nepal experiences regular flooding causing immense socio-economic losses (NCVST 2009).
Current flood risk management approaches in Nepal are underpinned by assessments of
historic levels of flood risk (DHM 2009). However, it remains to be seen whether data on
historic flood risk is an effective guide for assessing current (and future) flood risk. Gaining a
better understanding of the intensity of the 2014 flood in West Nepal, and comparing this to
past flooding levels will give initial insight into a (potentially) changing risk environment.
INTRODUCTION
Nepal experiences regular flooding, predominantly in the monsoon, resulting in significant
loss of life, property and livelihoods1 (NCVST 2009, DHM 2009, DesInventar n.d). In recent
years the incidence of flooding has increased dramatically in both magnitude and frequency
due to changes in the precipitation patterns (MoHA 2009). This paper provides an overview
and analysis of flood data from the 2014 Karnali floods, comparing actual rainfall and flood
levels against historical data.
The Karnali River is a perennial trans-boundary river, originating in the Tibetan Plateau and
flowing through the steep mountainous terrain of West Nepal. It is the longest river in Nepal
and the total catchment area of the Karnali basin is approximately 45,000 km2 (Zurich 2015).
Nepal is categorized as the 4th most climate vulnerable country in the world (Maplecroft 2011
cited in DHM 2015). With a changing climate, the magnitude and frequency of floods are
expected to increase in the near future. 80% of rainfall occurs during the monsoon months of
June to September (Kansakar et al. 2004, MoHA 2009), though the monsoon precipitation
pattern is changing, with fewer days of rain and more high intensity rainfall events (DHM
2015).
1
Floods 1971-2011 caused 3,329 deaths, affected 3.9 million people and caused $5.8 billion dollars of
damage (DesInventar n.d.)
2
An analysis of precipitation data in Nepal (1961-2006) has shown an escalating trend of total
and heavy precipitation events (Baidya et al. 2008). This is expected to contribute to greater
likelihood of floods, heavier erosion and flood runoff and an increase in damages and losses
resulting from floods (Nepal DHM 2015). In the fifty years prior to 2014, significant flood
events occurred in 1983, 2008, 2009 and 2013 (based on DHM dataset from Chisapani
station).
Figure 1: Flood History of the Karnali River based upon maximum instantaneous discharge
datasets for the Chisapani station (also quoted in Zurich, 2015).
THE 2014 KARNALI FLOOD
Extreme rainfall on 14th-15th August 2014 culminated in a devastating flood event in Karnali
basin, Western Nepal, with 222 people killed and 6,859 households displaced (Zurich 2015).
Meteorological and Hydrological Conditions at Chisapani station
The Chisapani hydrological and meteorological station is at the center of an Early Warning
System for the Karnali basin. The following analysis of the 2014 Karnali floods is based upon
both automatic and manual readings from the hydro-met station at Chisapani.
Figure 2A and 2B: 2A - Hourly Rainfall at Chisapani for August 14-16 along with
cumulative rainfall (note more than 400 mm rainfall between Hr 18 - 30); 2B - Snapshot of
the rainfall event that led to flooding, accumulated precipitation of more than 120 mm in a
span of 3 hours as captured by the NASA-TRMM satellite during the 2014 Karnali Floods
3
Chisapani station recorded 493mm of extreme rainfall in 9 hours (the highest figures since
records began). The previous 24 hour maximum was 367.2mm in 1981. 200mm of this fell in
the last three hours (3 am - 6 am). Accumulated 3 hour precipitation depths in excess of 200
mm have only been exceeded 13 times in the past 50 years.
On the night of August 14 into the early hours of August 15, torrential rain in steep terrain
caused the Karnali River to rise rapidly. At 1 a.m. the river rose to over 10 meters, triggering
an ‘amber’ alert. The flood conditions worsened and within two hours the river had risen to
above 11 meters, triggering a ‘red’ alert2, with water levels still rising. The automatic gauge
station at Chisapani stopped working at 2am on 15th August and manual readings were not
possible between 3am and 6am, as the trail to the gauge station became blocked by landslides
and torrents (Zurich 2015). When manual readings resumed at 6am the water level had
already risen to 16.10 meters, significantly above the 15 meter maximum measure for which
the gauge had been designed.
Figure 3: Flood Hydrograph at Chisapani station for August 14-16 [Note data not available
at Hr 28-29 (between these two hours water level rose from 11m to 16m) and Hr 39, 50-55]
ANALYSIS OF THE 2014 KARNALI FLOOD
The 2014 Karnali flood event (reaching 16.1 meters) was the highest recorded flood event
since hydrological records at Chisapani began in 1962. The only time levels previously
exceeded 15 meters was in 1983 (15.2 meters). This event is certainly an extreme in terms of
gauge heights recorded. Record-breaking rainfall upstream in the foothills Karnali catchment
led to rise of water level in such a short span of time. Since the 2014 flood event exceeded the
maximum design height of the river gauges, it highlights the importance of flood monitoring
and management approaches taking extreme events into consideration.
Rating Curves and Uncertainties in Predicting Discharge
Rating curves depict the relationship between water discharge (Q) and the flood stage (W) at a
particular point in time for a given gauge, information which can be used to map predicted
2
See Zurich (2015) for details of the response of the Community Based Early Warning System in Karnali
4
flooding downstream (Subramanya 1994) and to calculate flood return periods (see below). A
rating curve for the Chisapani station has been drawn (FIG 3) based upon past datasets of
instantaneous discharge corresponding to gauge heights (up to the maximum limit of 15.2
meters observed during the 1983 floods. Such rating curve relationships are determined and
updated by DHM on an annual basis.
Figure 4: Rating curve for the Chisapani station that indicates the relationship between
discharge and water levels. [Data obtained from Department of Hydrology and Meteorology]
It is important to note that during extreme flood events, the relationship expressed by the
rating curve becomes unreliable due to the dynamic nature of changing water levels, sediment
and river-bed topography (Zurich 2013). The stage-discharge relationship at any given river
cross section also varies with changes in bed load, river velocity, channel geometry,
longitudinal slope and bed roughness (Baldassarre and Montanari 2009, Ghimire and Reddy
2010). Rivers are influenced by the above dynamic factors, making it extremely difficult to
estimate flood discharge and river stage during extreme events.
For the Chisapani station at Karnali, if the rating curve was extrapolated to 16.10 meters, the
curve would provide an indicative discharge value, however the upper band of the rating
curve is usually marred by uncertainties (Baldassarre and Montanari 2009), therefore, the
peak discharge can only be speculated for the 2014 flood event.
Return periods associating Extreme Flood Events
Floods are complex natural events, resultant of numerous complex parameters making it
difficult to model analytically (Subramanya 1994); especially estimation of flood peaks based
upon the assumption that the flood regime is stationary (Petrow and Merz 2009).
Flood return periods are calculated to determine the probability of a flood of a particular
magnitude occurring, a key tool used for flood resilience planning (Zurich 2013). A flood
with a 100 year return period (misleadingly sometimes referred to as a 1 in a 100 year flood)
has a 1% probability of occurring in a given year. Flood return periods are calculated through
statistical method frequency analysis such as Gumbel’s Method is used (Payrastre et al.
2005), which is also applicable to other hydrologic processes where the input data is supposed
to be homogeneous, independent and stationary (Merz and Thieken 2009).
5
In order to calculate return periods for Karnali at Chisapani, instantaneous peak flood
discharge data made available by the Department of Hydrology and Meteorology (DHM) has
been sorted using Gumbel’s Method (Figure 5) with return periods calculated.
Figure 5: Return Period Graph using Gumbel’s Method, extrapolated to 1000 years to
estimate discharge during the 2014 floods
The analysis suggests that the 2014 Karnali floods might have been a flood with a 1000 year
return period – a flood with a 0.1% probability of occurring each year (Zurich 2015).
However, it is critical that the numerous uncertainties and errors associated with such return
periods are noted. It is generally acknowledged that flood peak measurements are associated
with considerable errors that lead to uncertainty in calculation of return periods (Ward et al.
2011). The datasets used to calculate such return periods are limited, with few rivers having
records for 50-100 years (Zurich 2013). For the Karnali River at Chisapani, instantaneous
flow values are only available for 1962-2013. Peak flow data for extreme flood events may be
particularly unreliable, as shown in the case of the 2014 Karnali floods when the automatic
gauge stopped working and the manual readings were not possible for several hours (Zurich
2015). Likewise for the majority of the dataset period (1962-2010), only manual readings are
available, which are vulnerable to measurement errors (Baldassarre and Montanari 2009).
Importantly, flood return periods are extrapolated from data on past floods. In a location like
Nepal, where flood risk is changing (NCVST 2009, MoHA 2009), return periods will not be
representative of the realm of possible future flood events. For these reasons, it is critical that
flood resilience efforts, whether flood risk assessments, hazard maps, or associated disaster
preparedness, disaster risk reduction and wider development decisions, take into account
extreme events.
CONCLUSION
The 2014 Karnali flood can be described as a 1 in a 1000 year flood (Zurich 2015). However,
having experienced a ‘1 in a 1000 year flood’, individuals are tempted to assume that such a
flood is unlikely to affect them again in the near future (Pielke 1999). It is therefore important
to focus on the statistical probability of an event occurring each year, defining it as a flood
with a 0.1% probability of occurring each year.
6
It is also important to note the numerous uncertainties in such flood risk calculations,
particularly in contexts such as Nepal where flood risk is changing (MoHA 2009). The
intensity of the 2014 Karnali flood underscores the danger of extreme events and may bring
into question any assumption that past flood risks can effectively guide future risk. For this
reason, it is critical that flood risk assessments and plans consider a range of flood scenarios
including extremes such as 1000 year return periods.
REFERENCES
Baidya, S. Shrestha, M. Madan, L. and Sheikh Muhammad, M. (2008) Trends in daily
climatic extremes of temperature and precipitation in Nepal, Journal of Hydrology and
Meteorology – Society of Hydrologists and Meteorologists Nepal 5 (10) 44-48
Baldassarre, G. D., & Montanari, A. (2009). Uncertainty in river discharge observations: a
quantitative analysis. Hydrology and Earth System Sciences, 13(6), 913-921.
Department of Hydrology and Meteorology (DHM) (2009). Determination of Flood Danger
Level in Flood Forecasting Stations. Kathmandu, Nepal.
Department of Hydrology and Meteorology (DHM) (2014). Preliminary Report on Monsoon
2014. Kathmandu, Nepal.
Department of Hydrology and Meteorology (DHM) (2015) Terms of Reference for
Establishment of End to End Flood Early Warning System in Koshi and West Rapti River
Basins to Support and Strengthen Disaster Risk Management. Government of Nepal
DesInventar (n.d.). http://www.desinventar.net/DesInventar/profiletab.jsp?countrycode=npl
(Accessed at: 07/08/15)
Dhakal, S. (2013). Flood Hazard in Nepal and New Approach of Risk Reduction,
International Journal of Landslide and Environment, 1(1), 13-14.
Ghimire, B. N., & Reddy, M. J. (2010). Development of stage-discharge rating curve in river
using genetic algorithms and model tree in International Workshop Advances in Statistical
Hydrology, Taormina, Italy (pp. 23-25).
Kansakar, S. L., Hannah, D. M., Gerrard J., Rees, G. (2004). Spatial pattern in the
precipitation regime of Nepal. International Journal of Climatology. 24, 1645–1659.
Merz, B., & Thieken, A. H. (2009). Flood risk curves and uncertainty bounds, Natural
Hazards, 51(3), 437-458.
Ministry of Home Affairs (MoHA) (2009) Nepal Disaster Report 2009. Kathmandu, Nepal.
NCVST (2009). Vulnerability through the Eyes of Vulnerable: Climate Change Induced
Uncertainties and Nepal’s Development Predicaments. Kathmandu: ISET-N.
Payrastre, O., Gaume, E., & Andrieu, H. (2005). Use of historical data to assess the
occurrence of floods in small watersheds in the French Mediterranean area. Advances in
Geosciences (2) 320.
Pielke Jr, R. A. (1999). Nine fallacies of floods. Climatic Change, 42(2), 413-438.
Petrow, T., & Merz, B. (2009). Trends in flood magnitude, frequency and seasonality in
Germany in the period 1951–2002. Journal of Hydrology, 371(1), 129-141.
Subramanya, K. (1994). Engineering Hydrology. Tata McGraw-Hill Education.
Ward, P. J., Moel, H. D., & Aerts, J. C. (2011). How are flood risk estimates affected by the
choice of return-periods? Natural Hazards & Earth System Science, 11(12), 3181-3195.
Zuich Insurance (2013). Risk Nexus: Central European floods 2013: A Retrospective. Zurich.
Zurich Insurance (2015). Risk Nexus: Urgent case for recovery: what we can learn from the
August 2014 Karnali River floods in Nepal.