An analysis of unmet water demand under climate change scenarios in the Gualí River Basin, Colombia, through the implementation of Hydro-BID and WEAP hydrological modeling tools

Climate change can affect hydrological services in Andean basins, so a possible reduction in water supply can lead to not meeting the needs of different users, which has become a real challenge for decision-makers with regards to water management. This paper presents the results obtained from hydrological modeling exercises in the Gualí River Basin (Colombia) by combining the Hydro-BID modeling tool, which consists of an analytical hydrology database for Latin America and the Caribbean that provides a great advantage for countries with limited information and the Water Evaluation and Planning System (WEAP) modeling tool, in order to determine the potential impacts of climate change on unsatisfied demand for water in the basin. The results show a possible decrease in flow compared to current conditions; between 5.8% and 9.56% for CPR 2.6, and between 2.18% and 6.86% for CPR 8.5. The approach presented is useful to ensure that timely decisions are made to meet the demands of users under the conditions of climate change. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wcc.2019.118 ://iwaponline.com/jwcc/article-pdf/12/1/185/851804/jwc0120185.pdf Darwin Mena (corresponding author) Lina Restrepo Melissa Pimiento Miguel Cañón Faculty of Environmental Engineering, Universidad Santo Tomás, Bogotá, Colombia E-mail: darwinmena@usantotomas.edu.co Abel Solera Institute of Water and Environmental Engineering, Universitat Politècnica de València, Valencia, Spain Freddy Duarte Faculty of Civil Engineering, Escuela Colombiana de Ingeniería Julio Garavito, Bogotá, Colombia


INTRODUCTION
Integrated water resources management (IWRM) is a worldwide concept that seeks to advance proper water planning and management to ensure its sustainability and conservation, while promoting the generation and access to information in order to solve water-related problems, provide assistance and implement tools and resources that Worldwide consumption of water has increased eightfold since the 19th century and has doubled in the last 15 years. Moreover, demand from the agricultural, domestic, and industrial sectors has increased by 7, 20, and 12 times, respectively. Urban water consumption has increased 20 times in the last 100 years. Consequently, in 2050, approximately 55% of the world population will face a water crisis (Song et al. ). In the specific case of Colombia, despite being considered one of the countries with the largest natural water supplies in the world, with an approximate water yield of 56 L/s*km 2 , which greatly surpasses the average world yield of 10 L/s*km 2 , on the water scarcity index nearly 4% of the population suffers from a high scarcity, 7% mediumhigh, and 30% medium. It is estimated that in 2025, the population affected by water shortages could reach 39%, which is why actions aimed at conserving this resource must be prioritized (Domínguez et al. ). Modeling the water resource is a key tool in planning, forecasting, and evaluating variables, policies, and strategies, among many other factors. However, one of the main problems in Latin American and Caribbean countries is the scarcity of information that provides sufficient and quality data to use in hydrological models. Even though WEAP has a hydrological modeling module and has already been widely used to evaluate the climate's effect on water resources and optimize demand points in states such as California, Massachusetts, Georgia, as well as in African and Asian countries (Strzepek et al. ), the advantage of integrating Hydro-BID is that it uses an analytical hydrology database that reduces difficulties related to information scarcity in developing countries.
Hydro-BID software has been implemented to simulate hydrology and water resources management in the Latin American and the Caribbean (LAC) region, under climate change scenarios in countries such as Argentina, Peru, and Ecuador with pilot models in Colombia (Fekadu et al. ).
Consequently, for the purposes of this study, Hydro-BID will be implemented to generate a time series of flows for the current climate and future climate projections and WEAP will be used to simulate water availability and unmet demand.

Case study
The Gualí River Basin ( The basin is primarily characterized by a tropical climate with a bimodal rainfall pattern due to the Inter tropical Confluence Zone, with a maximum annual precipitation of 3,105.12 mm. The average temperature is 18.7 C, with a maximum of 20.3 C and a minimum of 17.7 C, varying throughout the basin due to changes in its elevation, with its relative humidity reaching a maximum of 90%, highlighting that this parameter behaves inversely to temperature. Likewise, the water supply measured in the basin is approximately 42.8 m 3 /s, with an average water yield of 67.75 L/s*km 2 (Cortolima 

Hydro-BID modeling tool
Hydro-BID is an integrated and quantitative system to simulate hydrology and water resource management in the LAC region under climate change scenarios. Moreover, it facilitates the evaluation of water quantity and quality, infrastructure needs, adaptation strategies, and project design to address these changes. This software is centered on an analytical hydrographic dataset (AHD), which analyzes information on basin topology, currents, land use, soil types, precipitation, temperature, and observed current flows for its calibration. The rainfall-runoff model  The Hydro-BID hydrological model requires input data such as precipitation, temperature, and river flow series, expressed in centimeters (cm), degrees Celsius ( C), and cubed meters per second (m 3 /s), respectively. Specifically, these data were entered on a daily basis given that Hydro-BID manages the step related to time (Nalesso ). Sixteen precipitation stations, five temperature and four flow stations were selected with a 23-year historical record (01/01/1989-1/30/2013) with a daily temporality, with which a homogeneity analysis was performed by using the double-mass curve methodology, which found that the correlation factor, R 2 , varied by 0.9931 and 0.9995. To account for this, no adjustment factor was applied and the estimation of missing data was performed using the arithmetic mean method.

Calibration and validation of the Hydro-BID model
The model was calibrated by sub-basins, in order to identify optimal parameter values, by which the simulated data are correctly adjusted to the observed series. This process was performed manually by trial and error, from the upstream to the downstream basin, relating the parameters obtained for the basin immediately preceding it, to the mouth of the main riverbed. To carry out this process, the period between 01/01/1993 and 02/28/2005 was selected, since the data in this time interval were more complete for the three climatological variables, which reduced error throughout the calibration process. The parameters used in the calibration are the curve number, which is used to characterize the type of soil of the basin and its hydrology, which ranges between 0 and 1; and the available water capacity (AWC) that represents the quantity of water that can be stored in the soil to be used by plants, affecting the infiltration towards groundwater. Moreover, the recession coefficient (R) is used, which calculates how groundwater found close to the surface contributes to river flows or surface streams after a rainfall event. This is followed by losses from the system, or seepage, which represents the exchange between subsurface water with deeper bodies of water, in which the quantity of water attributed to this parameter is considered an output of the model. There are also the grow season and dormant season ET evapotranspiration factors, which are used for when there are, or are not, crops in the basin.
These last two parameters are given a value of 1 when they are not taken into account, which was the case in this study, as a result of the diversity of crops that vary in type and seasonality in this zone (Fekadu et To evaluate the calibration, Hydro-BID integrates a series of metrics which include: the overall error volume, for which the closer the result is to zero, the better the model's performance; correlation metrics (R); the modified correlation; and the Nash-Sutcliffe efficiency, which range between 0 and 1, in which the closer the result is to 1, the greater the reliability of the simulation (Dawson et al. ).

General circulation model (GCM) and downscaling
In performing an analysis of the impacts of climate change on basins, having seasonal and spatial information at a smaller scale than those given by general circulation models (GCM) is indispensable. To this end, carrying out a downscaling process on general series is necessary, based on the assumption that large-scale climate conditions have a strong influence on climate at a local level (Maraun et al. ). In this study, a statistical downscaling method based on the chaos theory was employed, also referred to as downscaling. This method was chosen because existing scale reduction techniques, even though they have presented successful outcomes, are based on the assumption that climate systems are linear, while other studies have demonstrated their non-linear and chaotic behavior (Mihailović et al.

).
Two climate change scenarios, RCP 2.6 and RCP 8.5 which represent opposite projections, were selected. RCP 2.6 establishes a strict mitigation scenario where population growth decreases, the use of energy and fossil fuels is limited, and a strict climate change policy is created. This is designed so that, under these conditions, there is a greater probability of maintaining global warming at less than 2 C above pre-industrial temperatures. On the other hand, RCP Due to the fact that the Gualí River Basin is located within the project study area, the three most suitable GCMs were preselected by taking into account the relative bias for the Magdalena-Cauca Basin. The models that have the least amount of bias for this zone were MPI-ESM-MR, CCSM4, and NORESM1-M. Subsequently, the square root metric of the mean square error (RMSE) between the historical GCM, and precipitation and temperature station records for each of the models was calculated. Based on these results, the MPI-ESM-MR model was selected, which generated an RSME of 5.44 for precipitation and 6.91 for temperature.
The downscaling method implemented considers deterministic chaos concepts and the non-linear dynamic in climatological series. This method takes into consideration two main components: the evaluation of the presence of chaos, and the synchronization of global data series with data from weather stations, which for input data, requires precipitation and temperature series from weather stations, and a catalog that has the stations' coordinates and the GCM historical model, in addition to selected RCP data.
For outputs, it generates reduced series and an error that evaluates the performance of the model; RMSE in this case, for its validation (Duarte ).

WEAP water allocation module
As a water assessment and planning system, WEAP is a robust computational tool that evaluates and carries out IWRM. Its ultimate aim is to represent current and future water conditions in a determined study area, in order to calculate water supply and demand. This model was The WEAP modeling tool consists of several modules, including the hydrological module that contains a rainrunoff model which provides five different methods for calibration. Another module, which was used in this work, is the water allocation. The latter is a support system for decision-making to facilitate the management of water resources, which is based on optimizing the satisfaction of the demands of different users.
In this investigation, the first module mentioned (hydrological) was not used, because the objective was to test the versatility of being able to integrate the Hydro-BID hydrological model and the WEAP water allocation module, which was verified in the results.

Integration of WEAP and Hydro-BID models
The integration of WEAP and Hydro-BID was a three-step process. In the first, current demand was calculated and a projection was made for future demand. The second entailed the topology design in WEAP, and in the final step, flows produced by Hydro-BID were integrated.
To calculate the current demand of the basin, it was necessary to refer to records from the Regional Autonomous Corporation, Cortolima, in order to obtain the values corresponding to the concessions assigned within the study area that are classified as domestic, agriculture, or industrial.
Based on these data and complemented with information found in the River Basin Management Plan (also known by its acronym in Spanish, POMCA) of the Gualí River, the demand points along the study area were classified, depending on the current from which the catchment is per-

).
Based on this analysis, the initial losses were set at 35% for domestic demand points, 30% for industrial demand points, and 55% for agricultural points.
To determine the future demand, a forecasting exercise was performed for the consumption mentioned, which was based on historical gross domestic product (GDP) records Once the analysis and projection of water demand in the basin was completed, the monthly flows obtained from Hydro-BID were used as input data in the WEAP software  In the modeling carried out in the WEAP software, it was not necessary to perform any calibration process given that the program only used the analysis module and demand projections. In this sense, the flows obtained from the Hydro-BID program were integrated into the WEAP topology as the base flows for each sub-basin and as the primary information to obtain the results in each of the proposed scenarios.

Indicator approach for unmet demand and potential reduction measures
The indicator used to estimate the potential effect of climate change on water supply and its repercussions on demand is the unmet demand, which represents the total water deficit in m 3 /s. In order to decrease the identified levels of unmet demand, three management scenarios were considered that consist of decreasing the level of loss throughout the distri-

General circulation model (GCM) and downscaling
The performance of the downscaling exercise was evaluated with the RSME metric to estimate the correlation between GCM data before and after the downscaling process, with respect to historical series from the climatological stations.
As shown in Table 2, the correlation between the data is closer to the optimal value, which is zero for a fully adjusted model, after carrying out the downscaling process, verifying that the downscaling technique minimizes error levels that are generated by applying GCM (Fekadu et al. ).    A percentage increase or decrease in precipitation and temperature was calculated in each season in two periods, before the middle of the century (half of century (HOC)) and at the end of the century (end of century (EOC)). It was obtained that on average for the RCP 2.6 scenario in the first half of the century, an increase of 2.01% and 0.06% is expected for precipitation and temperature variables; and for the second half of the century, a possible increase of 0.95% and 0.1%, respectively, will be generated compared to the current conditions. Likewise, the RCP 8.5 scenario showed an average decrease of 0.68% for precipitation in the first half of the century, and in the second half an increase of 1.1% was obtained for the same variable; in addition, for the temperature, an increase of 0.06% and a decrease of 0.08% are expected in the first and second half of the century, respectively.

Calibration and validation of the Hydro-BID model
As a result of the calibration, the final parameter values were obtained for the output sub-basin, specifically 1.75 for the curve number, 0.6 for the water content in the soil, 0.02 for the recession coefficient, 0.00005 for the seepage, and 1 for the grow season and dormant season ET factor parameters. As a result of applying the parameters obtained, monthly performance metrics were calculated as follows: 4.51 for the general error volume, 0.65 for the R correlation, 0.45 for the modified correlation, and 0.42 for the Nash-Sutcliffe efficiency.
From these results it is evident that the calibration process was satisfactory, comparing the performance metrics before and after said process, as evidenced in Table 3, where the calibration process is carried out, and various metrics approximate the optimal ranges.
While the value of the Nash-Sutcliffe index performance metric is not ideal, it is acceptable. This behavior is the result of the study area being considered a pulse basin, where there are sudden increases in runoff volumes caused by thawing from the Nevado del Ruiz volcano (upstream). As this behavior was not considered in the calibration process, the adjustment of the results in the upper part of the basin only represents the atmospheric component, not additional inputs. In considering that the calibration was performed sequentially, this behavior caused the overall results from the same process to be accep- With respect to the sensitivity of the Hydro-BID calibration parameters, it was found that the most sensitive parameter is the curve number followed by the AWC, given that these coincide with the runoff quantity that is generated throughout the basin. This is mainly due to the curve number representing the waterproofing level that is present in the study area, due to natural or anthropogenic conditions, thus favoring, or not, infiltration processes from precipitation received on the soil. Furthermore, the AWC, by representing the soil's water content, determines the volume of water that is capable of being absorbed before the soil reaches its saturation point, in which from that  RCP 2.6 scenario and 0.0037 m 3 /s for the RCP 8.5 scenario, the results of which are shown in Table 4. Similarly, it is evident that the Mariquita and Fresno aqueducts are still the most affected and therefore present a greater risk of impacting the development of economic and cultural activities in the area. Lastly, an inversely proportional behavior was identified between the loss reduction percentage and the flow amount that is not completely supplied in the basin, in which the greater the reduction and improvement in the distribution system from the catchment point to the places of consumption, the lower the percentage of unmet demand.
Finally, it was found that by only reducing losses in the system as a measure, unmet demand can be reduced by up to 87.8% in the case of RCP 2.6 C, and up to 90.26% in the case of RCP 8.5 C. This is assuming that the concessions granted by the corporation will remain constant in the basin and that no additional catchment points will be created. Furthermore, carrying out field visits is considered pertinent in order to verify the information provided by Cortolima and identify possible catchment points not registered with the authority.

CONCLUSIONS
The downscaling process more accurately represents the real conditions of the study area, increasing the correlation between the observed data and the historical data from the GCM. This is reflected in the results from the RSME metric, which moved from an average of 7.79 cm before downscaling to 1.52 cm afterwards for RCP 2.6, and from 7.97 cm to 1.25 cm in RCP 8.5 for the precipitation variable.
In the same manner, the temperature metric moved from 8.36 C before downscaling to 3 C in RCP 2.6, and from 8.69 C to 2.94 C in RCP 8.5.
In general, accurately depicting future climatic conditions implies a high level of complexity, especially for the precipitation variable, because when simulations are based on GCM, uncertainty increases in projecting flows obtained from the modeling process. Therefore, the results should be taken as a guide for potential changes that may occur in the basin from now until 2100, rather than a true fact.
In the Gualí River Basin, for the period covering 2011-2100, under the RCP 2.6 and RCP 8.5 scenarios, available flow decreases between 5.8% and 9.56% are expected for RCP 2.6, and 2.18% and 6.86% for RCP 8.5, in which 2013-2040 will be the period with the greatest reduction in flow.
With respect to the conditions of the reference period between 1989 and 2013, there was no unmet demand for any of the established demand points, which is consistent with the studies carried out by the POMCA for the Gualí River Basin. However, the scenarios simulated have a critical period between 2071 and 2100, in which there is unmet demand throughout the basin.
The simulation results demonstrate that there is unmet demand in the Fresno and Mariquita aqueduct nodes, with an average of 0.0413 m 3 /s and 0.0386 m 3 /s, respectively.
Implementing measures to reduce loss throughout the distribution system can decrease unmet demand by up to 87.8% and 90.26% for the RCP 2.6 and RCP 8.5 scenarios, respectively. However, it is not possible to cover the demand of every point in their totality by only applying this measure. Furthermore, of the proposed mitigation strategies, Option C is the most effective as it has the highest loss percentage reduction at every demand point.
To reduce losses, it is recommended that supply system operators optimize the use of micro-measurement, actively control system leaks, and overhaul and/or replace networks.
Of the climate change scenarios considered, RCP 8.5 has the highest flow when compared with RCP 2.6, resulting in a smaller unmet demand throughout the analysis period for the former.
Lastly, based on the Hydro-BID results, it was found that there will be a possible loss of available flow in the