Hydrological Modeling To Assess Water Resources Resilience Towards Climate Change Impacts in Tacloban, Philippines
Hydrological Modeling To Assess Water Resources Resilience Towards Climate Change Impacts in Tacloban, Philippines
Hydrological Modeling To Assess Water Resources Resilience Towards Climate Change Impacts in Tacloban, Philippines
Rita Marteleira1 · Maria José Roxo1 · Rodel Lasco2,3 · Pedro Santos Coelho4
Interdisciplinary Center of Social Sciences, Faculty of Social Sciences, New University of Lisbon
1
2
Oscar M. Lopez Center for Climate Change Adaptation and Disaster Risk Management Foundation Inc
3
World Agroforestry Centre
4
Department of Environmental Sciences, Faculty of Sciences and Technology, New University of Lisbon
Received: 15 June 2017 / Accepted: 15 January 2018 / Published online: 13 February 2018
Abstract
Tacloban, Philippines, the city most devastated by typhoon Haiyan in 2013, faces severe water supply
constraints, particularly on the relocation sites that are expected to accommodate 40% of Tacloban’s
population by 2018. This region’s hydrological system was modeled using QSWAT to evaluate the
impacts of climate change, namely, rising temperatures and variation of rainfall patterns, on three rivers
(Binahaan, Pongso, and Cabayugan). These rivers are the potential sources of water for the two supply
solutions currently being considered. Thus, it is essential to assess their resilience as sources of water.
Two climate scenarios were generated for 2020 and 2050 using the Change Factor (CF) method, and the
system response was compared to the baseline data (1979–2010).
Results show that, regardless of the chosen supply solution, for certain time periods (most severely for
the quarter of March to May), the percentage of days in which the river flows were insufficient to meet
the proposed water catchment flows was significant (reaching up to 25%). Outputs also suggest that
Binahaan River, where the current water catchment is located, may fail in providing sufficient water
to the region for future scenarios. However, it should be noted that Pongso and Cabayugan Rivers in
northern Leyte can represent a resilient water source to supply the new resettlement areas.
Keywords: Climate change adaptation ∙ hydrological modeling ∙ QSWAT ∙ resilience ∙ typhoon haiyan ∙ water
resources management
DOI: https://doi.org/10.18783/cddj.v003.i01.a02
Corresponding Author:
Rita Marteleira
[email protected]
Volume 3 Issue 1 February 2018
Climate change (CC) has recently become a matter of the utmost The Republic of the Philippines is one of the most vulnerable
importance in water studies, as it is capable of altering water territories to CC, given its geographic location in the Tropical
availability worldwide due to variation of rainfall patterns. The Pacific, archipelagic formation, and population distribution. In
potential impacts include the changes in timing and magnitude recent years, this country was ranked highest in the world in
of surface runoff, stream discharge, evaporation, and flood terms of vulnerability to tropical cyclone (TC) occurrence and
events (Neupane & Kumar, 2015). Several studies have third in terms of people exposed to seasonal events (Climate
demonstrated that even small perturbations in the frequency Change Commission, 2012). The Philippines has already
and/or quantity of precipitation can have significant impacts in experienced noticeable adverse effects in recent years, with the
terms of watershed runoff (Minville, et al., 2008). observed changes in its climate in most recent times having never
been seen in the past 140 years. Climate trends evaluated for
However, coupling hydrological and climate studies is a the 1951–2009 period (using reference data from 1971 to 2000)
relatively young discipline (Teutsschbein, 2013). CC impact revealed an increase in the annual mean temperature by 0.65ºC,
studies on hydrological systems faced the obstacle of the Global with an increase of 0.36ºC and 0.1ºC in terms of maximum
Circulation Models (GCMs), which were used to simulate and minimum temperatures, respectively. The trends of rainfall
future climate scenarios but did not provide sufficiently detailed (extreme rainfall intensity and frequency) are not so clear,
outputs. This obstacle was surpassed only with the use of with very little spatial coherence, varying between islands and
dynamic downscaling tools, namely, Regional Climate Models regions (increasing in some, while decreasing in others). These
(RCMs), even though some bias correction is still recommended climate simulations were obtained by the Philippine Atmospheric,
when using directly RCM outputs on hydrological studies Geophysical and Astronomical Services Administration (PAGASA)
(Fowler & Kislby, 2007). using the project Providing Regional Climates for Impact Studies
(PRECIS), an RCM developed by the UK Met Office Hadley Center
The concept of resilience, initially coined for an ecological for Climate Prediction and Research, considering boundary data
context, has conquered prominence in CC discussions, from the ECHAM4 and HadCM3Q0 Global Climate Models. For
particularly those regarding adaptation. In brief, the resilience this downscaling, three emission scenarios were assumed: A2 (high
of a system can be defined, for a certain system, as the “ability range), A1B (medium range), and B2 (low range) (PAGASA, 2011).
to reduce both the magnitude and the duration of the deviation
from the equilibrium stadium” (Proag, 2014). Likewise, in the Although there is still no sound evidence of an overall nationwide
context of disaster resilience, Klein, Nicholls, and Thomalla increase in the frequency of TCs, for the Visayas region, TCs
(2003) claim that the concept of resilience sits under the broader seem to have increased from 1971 to 2000, as compared with
concept of “adaptive capacity” as one of the numerous factors previous periods. Moreover, a slightly superior number of
capable of influencing it. Hence, resilience appears as an idea extreme cyclones have been observed (from the typhoon
with direct application potential, especially when moving from category, with winds above 150 km/h (PAGASA, 2011). These
disaster recovery (as for this particular case study) towards observations are partially sustained by the work of Cinco et al.
hazards prediction, or disaster prevention and preparedness (2016), stating that, for the period of 1951–2013, there have
(Klein et al., 2003). actually been fewer typhoons reaching the country, but there is
an increase in the frequency of the most extreme ones.
Resilience indicators for water resources focus mostly on water
scarcity or quality degradation issues. Possible indicators for The city of Tacloban is located at the northeastern coast of Leyte
water scarcity are average precipitation, average river discharge, island, Visayas region, Philippines, between the San Juanico
and average soil moisture or groundwater level—these are Strait and the island’s highlands—Sta. Helena and Naga-Naga
examples of components that can be accurately studied through mountain ranges—that reach above 450 m of altitude (Wernstedt
hydrological modeling, facilitating the benchmarking of & Spencer, 1967). To most authors, the climate in this region
various scenarios and the development of a sensitivity analysis resembles a Type II climate, with no dry season and a period
to better understand the weight of each variable. The sensitivity of relatively more intense rainfall from November to February.
of a system to water scarcity is therefore determined by the From the analysis of local climatological patterns for the period
influence of the stress factor (e.g., CC): for example, in some of 1987–2006, Tacloban’s annual average rainfall totaled 2600
basins, a reduction of rainfall can result in a reduction in water mm. The Philippines coincides with the typhoon belt, with
availability, whilst in other basins, it can have only a negligible circa 20 storms crossing the country per year and the strongest
impact on stream flow. Other possible indicators are the ones being registered from July to October (Gultiano, et al.,
changes in water demand for future scenarios, comparatively 2003). However, on November 8, 2013, super typhoon Haiyan,
to a reference period (and the possible disaggregation by the strongest registered so far to ever make landfall, roared ashore
water users and sectors); the loss of industrial and agricultural over Tacloban, causing catastrophic destruction and over 6000
production (or jobs), income, and livelihoods; desertification; casualties (NDRRMC, 2014), with its outrageous proportions
and land degradation (Elis, 2014). being often linked to CC causes. Being one of the major cities hit
by the 6-m tsunami-like storm surge, Tacloban soon became the
focus of world media attention and international relief operations.
14
Climate, Disaster and Development Journal
To the present day, the city is undergoing major relocation efforts QSWAT interface presents a three-step configuration to be
towards north that will create new residential, institutional, and followed by the user in order to prepare the SWAT simulation
commercial areas on a supposedly storm-surge-safer zone. (Dile, et al., 2016), followed by a fourth step for output
Leyte Metro Water District (LMWD), the local water utility, analysis. Step 1: Delineate Watershed, where the user selects
is responsible for supplying Tacloban City and the seven the elevation map grid to be used and can optionally burn an
surrounding municipalities of Dagami, Palo, Pastrana, Sta. Fe, existing stream network and a predefined watershed. In this
Tabontabon, Tanauan, and Tolosa. Despite only 25% of the step also, the threshold for creating subbasins is set, either by
city’s households being actually connected to this utility, “due area or number of cells. Step 2: Create Hydrological Response
to political disputes, LMWD does not intend to increase their Units (HRUs), where the user sets the conditions to delineate
service area within the City of Tacloban” (United States Agency HRUs. The HRUs are smaller partitions of the subbasins,
for International Development [USAID], 2015). However, a which share a unique land use, soil, and slope; therefore, in
proposal to expand the city’s network has been submitted, which this step, both land use and soil maps are required, together
relies mostly on the Binahaan River catchment, towards the with their respective look-up tables that should be previously
northern relocation areas. Until the submission of this article, prepared if the user is not using the default ones. Step 3: Edit
this resettlement area that will accommodate 40% of Tacloban’s Inputs and Run SWAT, where the input tables containing
population in 2018, was still lacking a solid water supply climatic data are provided by the user. Step 4: Visualize, in
solution and other proposals were being evaluated—some which, to analyze the simulation outputs, the user may choose
advanced by the local government unit (LGU) of Tacloban. from static data, animation, or plotting. For more details on
QSWAT functioning, setup and outputs visualization options,
As a consequence, relocated families are fleeing back to the refer to the work of Dile et al. (2016).
city, rebuilding their homes in markedly dangerous flood-prone
areas, as Santos, Toda, Orduña, Santos, and Ferrão, (2016) have Data Sources
foreseen. The impacts of this typhoon also affected the region’s
vegetation (Carlos, et al., 2016) and water resources (Tolentino, QSWAT requires both maps and table data as inputs. In
et al., 2016). Step 1 of the setup, the watersheds are delineated based on
elevation layers and a stream network. The Digital Elevation
Therefore, the goal of this research is to quantify the impacts
Model (DEM) raster file was obtained from Earth Explorer,
of CC on the hydrological system on which Tacloban relies
a database from the US Geological Survey. The 30-m
for water supply purposes. A closer look was given to the
resolution grid was then clipped to show only Tacloban
three waterways currently considered for water extraction:
City and its surroundings (this allowed faster simulations;
the already explored Binahaan River and the Pongso and
hence, the model had a smaller area to process). As for the
Cabayugan Rivers, pointed out by the USAID as possible water
streams shapefile, it was obtained from the online database
sources to supply the new resettlement areas. Furthermore, this
DIVA-GIS and again clipped to show only the interest area.
work is expected to contribute to the enforcement of the role
After a few attempts, it was determined that a threshold of
of hydrological models in CC impact studies—specifically in
50 km2 for delineating watersheds was adequate to match
developing countries where data collection is scarcer—and
the created stream network, providing sufficient detail for
to support the resilience-enhancing practices on local water
the objectives of the study.
resources management.
Land use and soil maps were required in Step 2 for the
creation of HRUs. The land-use layer was downloaded from
Materials and Methods the WaterBase database, a service provided by the United
Nations University with the aim of advancing the practice
Hydrological Modeling with QSWAT of Integrated Water Resources Management in developing
countries, while the soil map used was one from the Food
In order to replicate Tacloban’s surrounding water lines and Agriculture Organization (FAO) of the UN. Only
behavior under CC impacts, a QGIS-based version of the clipping and reprojecting operations were then necessary to
Soil and Water Assessment Tool (SWAT) was used. This is have these maps ready for the simulation. The standard land
a widely used hydrological model from the US Department use and soil lookup tables used, compatible with the maps
of Agriculture, i.e., QSWAT. Since it has the nature of a from this database, were by default included on QSWAT.
continuous-time model (not as adequate to simulate single-
event but rather long-term changes), SWAT enables users to Climate variables for QSWAT simulations (introduced as
address gradual environmental problems—being increasingly table data) provide the moisture and energy inputs that affect
used for CC contexts all over the globe. For instance, the model the water balance and determine the relative importance of
is capable of simulating the impacts of temperature changes and its various components. Tacloban City weather station is
soil water deficit on vegetation growth and erosive response, installed 2.71 m above the sea level in the peninsula that
besides the effects of global CCs on hydrological balance and, hosts the city airport (11º12’32.21”N, 125º01’29.92”E)
therefore, on the processes that determine surface and base and is managed by PAGASA. Daily data records from this
flow generation (Serpa et al., 2015).
15
Volume 3 Issue 1 February 2018
station were collected for the following variables, according Scenarios and CC
to the model requirements: precipitation (in millimeters),
temperature (min and max, in degree Celsius), relative CC can be simulated with SWAT by manually manipulating
humidity (in percentage) and wind speed (in meters per the climatic daily data series and then running the
second). These series corresponded to the period of January model using them as input data to assess changes in the
1, 1979, to December 31, 2010. The solar radiation series hydrological system (Neitsch et al., 2011). This method
was extracted from the Global Weather database; hence, is frequently defined in specialized literature as the CF
these records were not available for Tacloban. This database method. In essence, this method modifies the observed
hosts data for any given location, which are already historical time series by affecting it with the projected
displayed in a compatible format. The “stations” on this differences between future and current climates, computed
database, however, do not correspond to real stations but by climate models, either GCMs or RCMs (Chen, et al.,
to points of a matrix of a weather generating model. The 2011). These perturbation factors are computed on a
point 1111250 was chosen as being closest to Tacloban’s monthly basis, for each future time period (Equations 1
PAGASA station (only 16 km apart and with an elevation of and 2; Minville et al., 2008). With regards to generating
4 m). It was assumed that using solar radiation data from a future temperature daily series (T fut,d), the CF method
different station would not have significant impacts on the involves adjusting the observed daily records (T obs,d) by
modeling outputs, considering the proximity of the stations, adding the difference in monthly averages (T fut,m and T obs,m)
similar elevation, and no relevant orographic occurrences predicted by the climate model between the future horizon
between them. Moreover, from all the climatic variables, and the reference period.
solar radiation is believed to be the one with the lowest
expected spatial variability and not as influenced by local
factors. Finally, the Weather Generator Tables (WGEN for
the different scenarios) were created according to the model
description of Neitsch, Arnold, Kinity, and Williams (2011; As pointed out by Chen et al. (2011) and Minville (2008),
for further detail on the creation of these tables, refer to for the precipitation series, the adjusted precipitation on
these authors’ work). the horizon period (Pfut,d) is obtained by affecting daily
observed records (Pobs,d) by the precipitation ratio between
It should be noted that QSWAT uses the SCS-CN method to monthly averages (Pfut,m and Pobs,m).
estimate superficial runoff, balancing for each hydrological
unit its soil characteristics, which loses some accuracy when
applied to larger basins. However, the lack of spatially
distributed rainfall data for the study area did not allow the
use of other methods that could enhance the accuracy of this Despite being frequently used to strategic-scale assessments
impact study. of hydrological impacts of CC, this method is not without
problems: statistically, the baseline and scaled series are
Statistical Validation of Precipitation Series different in their means, maxima, and minima, but all the
other properties of the data remain unchanged (such as
The rainfall records registered by a meteorological station range and variability). Thus, the CF method application to
are statistically defined as random variables (assuming that the precipitation series does not affect the number of rainy
rainfall events are independent from each other), thus being days—meaning that the temporal sequence of wet and dry
only useful if validated through a statistical analysis. If a days remains unchanged, limiting the method’s efficiency
sufficiently long series of rainfall records is available, one in circumstances of persistent droughts, for example,
can determine which probability distribution function is which are important in impacts assessment. Furthermore,
closer to the empirical distribution of the data. One of the the procedure fails by admitting that the spatial pattern
main advantages of adjusting such function is that it further of the present climate remains unchanged (Diaz-Nieto &
allows extrapolating values with a return period larger than Wilby, 2005). Nonetheless, assuming that it should not be
the one of the analyzed sample (with a comfortable confidence used to make projections of precipitation on a daily basis
degree, usually around 1-α=0.95). The normal distribution as the temporal structure of the series remains constant, the
(or Gaussian Law) is reportedly adjusted to annual average CF method is still largely used in quantifying the impacts
records, while the Gumbel Law is used to describe the of CC on hydrology, and its application is enforced when
distribution of extreme data (Lencastre & Franco, 2006). equivalent observational and GCM/RCM data exist for
comparison purposes.
The quality of fit of these adjustments was analyzed
recurring to analytical methods such as the Chi-squared CC scenarios for this modeling exercise were created
and the Kolmogorov–Smirnov tests. For more detail on consistently with PAGASA’s official projections for
these statistical procedures, refer to the work of Lencastre Leyte, as summarized on Table 1. These climate trends
and Franco (2006). were obtained via the UK Met Hadley Center RCM data
16
Climate, Disaster and Development Journal
Baseline (ºC) 2020 (ºC) 2050 (ºC) Baseline (mm) 2020 (%) 2050 (%)
DJF 24.6 +0.9 +1.8 689.5 +3.0 +9.4
MAM 27.8 +1.2 +2.3 342 -8.9 -18.9
(from the project Providing Regional Climates for Impacts This analysis comprised the referred modeling scenarios:
Studies or PRECIS). This downscaling to the Filipino baseline (1980–2010), CC2020 (2007–2036), and CC2050
territory was performed using a grid resolution of 25 km, (2037–2065).
considering a baseline period from 1971 to 2000. For
Leyte, rainfall is expected to increase in all seasons except As a resilience indicator, the number of days in which the
from March to May, and for the same period, temperature river flows were lower than the proposed water withdrawal
increases are expected to be the highest (PAGASA, 2011). was determined, expressed in a percentage of total days and
discretized into four quarters to assess possible seasonal
Taking the Tacloban’s weather station data series (for variations of water availability (Equation 3).
precipitation and maximum and minimum temperatures),
the daily records were affected by these variations to
obtain two new series for projected CC scenarios of 2020
and 2050, which are designated as CC2020 and CC2050,
respectively. For the remaining climatic variables (relative In accordance with the supply constraints that Tacloban
humidity, wind speed, and solar radiation), the same faces and the supply proposals currently considered, two
records as the original series were used due to lack of supply solutions were evaluated:
projections for future scenarios.
• the “business-as-usual” scenario, in which LMWD
Modeling Calibration and Validation keeps extracting water from Binahaan River to supply
the whole community (including Tacloban City and
The calibration and, if possible, the validation of any the seven supplied municipalities of Dagami, Palo,
modeling exercise outputs are essential steps to assure that Pastrana, Sta. Fé, Tabontabon, Tanauan, and Tolosa)
the chosen model is accurately describing the real system. plus the Tacloban North resettlement areas, considering
Thus, both calibration and validation are often processed the existing withdrawal rate of 0.72 m3/s; and
by comparing obtained data with real empiric observations.
• the USAID-proposed complementary solution of
The QSWAT modeling results for Tacloban were calibrated resorting to the northern Leyte rivers (with an extraction
using a series of observed monthly average flow gauge rate of 0.58 m 3/s), i.e., Pongso and Cabayugan, to
records for the Sapiniton Gauging Station, from 1985 to 2003, supply only the new resettlement areas, while the City
available on the referred USAID report (hence, there were of Tacloban and surrounding municipalities continue
no flow measurements available for Binahaan, Pongso, and to receive water from the Binahaan River.
Cabayugan Rivers). This gauging station (refer to Figure 1
for its location) corresponded to a drainage area of 204 km2.
Although relatively smaller (with only 147 km2), for calibration Results and Discussion
purposes, subbasin 3 was considered as correspondent. To
assess the similarity of observed and obtained results, simple Data Series and Statistical Validation
correlation functions were applied.
Tacloban’s station daily precipitation records for the baseline
Resilience Analysis series were arranged to obtain the total annual and annual daily
maximum series. The former was adjusted using a Gaussian
Quantitative indicators of water supply are limited by the distribution function, while the latter was adjusted using a
premise that there is either sufficient or insufficient water Gumbel Law function. The quality of these adjustments was
to suppress a certain demand. The modeling output daily validated through the Kolmogorov–Smirnov and Qui-squared
flows, from the three rivers relevant to the considered tests, the results of which are presented on Table 2.
supply solutions, were the subjects of a resilience analysis.
17
Volume 3 Issue 1 February 2018
Adjusted
Data Series Average Std Dev Distribution Kolmogorov-Smirnov Test Results Qui-squared Test
Function Results
18
Climate, Disaster and Development Journal
19
Volume 3 Issue 1 February 2018
JJA 20.3% 26.4% quarters, should be addressed. Likewise, considering land cover
SON 1.3% 1.2% changes could be also interesting, hence, as demonstrated by
TOTAL 9.7% 11.3% Tolentino et al. (2016), after Haiyan, the reduction of forests
DJF 0.7% 0.5% on these watersheds conducted to a decrease of a third of the
MAM 17.7% 18.1% groundwater infiltration—thus increasing the potential runoff.
CC2020
20
Climate, Disaster and Development Journal
Elis, J. (2014). Climate Resilience Indicator Literature Review, Teutsschbein, C. (2013). Hydrological Modeling for Climate
Canada: Columbia BasinTrust. Change Impact Assessment - Transferring Large-Scale
Information from Global Climate Models to the Catchment
Fowler, H. J. & Kislby, C. G. (2007). Using regional climate Scale. Stockholm: Department of Physical Geography and
modeling data to simulate historical and future river flows in Quaternary College, Stockholm University.
northwest England. Climatic Change, 80(3), 337-367
Tolentino, P. L. M., Geronia, M. C. M., Clutario, M. V. A. & David,
Gultiano, S., Balbarino, E. & Saz, E. (2003). Chapter 3: Geography C. P. C. (2016). Potential long-term impact of Typhoon Haiyan
21
Volume 3 Issue 1 February 2018
22