The Philippine Hydrological Model (Scheidegger et al. 2023) is the first national-scale hydrological model of the country. Its primary purpose is to quantify components of the hydrological cycle at the national level, with spatio-temporal patterns of precipitation, evapotranspiration, surface runoff, river flow, groundwater recharge, and groundwater flow as model outputs. The model has been developed using a version of the macro-scale hydrological modelling software, VIC, into which a gridded groundwater model has been added. Consequently, it simulates the integrated surface water and groundwater system. The model has been constructed using openly available global data sets and calibrated against local observations, principally of river flows. The resulting modelling framework provides a means to develop understanding of the water resources across the Philippines and aims to support future national water resources planning. We present a summary of the results of the model using monthly district averaged components of the water cycle, and briefly describe some of the potential benefits of the continued development and improvement of the model, and of its future use.

Over the past decade, the Philippines has improved their national water security index and lies now within the top third of Asian countries (Asian Development Bank 2020). Nevertheless, changing climate and increasing urban population are putting more stress on water resources. Decreasing rainfall during the dry season and more intense rainfall during the wet season will exacerbate both water availability during periods of drought and the magnitude of flood events during periods of heavy rainfall. To assess the current state of, and future changes to, water availability across the country there has been a need for a national-scale hydrological model. Development of the Philippine Hydrological model began under the ‘Philippines Groundwater Outlook (PhiGO)’ project - a collaboration between the British Geological Survey (BGS), Ateneo de Manila University (ADMU), and the Philippine’s National Water Resources Board (NRWB). The project was joint-funded under the NERC-Newton and DOST-PCIEERD programme, understanding the impacts of hydrometeorological hazards in Southeast Asia (project NE/S003118/1). Under PhiGO, models of Panay and Pampanga were developed. Using subsequent funding from BGS’s International Geoscience Research and Development (IGRD) programme the model was then expanded to cover the whole of the Philippines.

To simulate groundwater recharge in the Philippines, we use the integrated VIC hydrological model coupled to a lateral groundwater flow model (VIC-AMHAS), as developed by Scheidegger et al. (2021). VIC is a macro-scale hydrological model, which has been applied widely for water and energy balance studies (Hamman et al. 2018). The model describes full water and energy transport over a grid cell. When precipitation reaches the land surface, it is partitioned into runoff and infiltration. To accumulate flows at river gauging stations, routing of runoff and baseflow is performed by post-processing model output (Lohmann et al. 1996). Figure 1. VIC-AMBHAS model framework. The soil column in VIC is coupled using bi-directional exchange of water between the soil and the aquifer. The aquifer allows for river baseflow, abstraction and leakage. BGS © UKRI. VIC-AMBHAS model framework (figure 1). The soil column in VIC is coupled using bi-directional exchange of water between the soil and the aquifer. The aquifer allows for river baseflow, abstraction and leakage. The lateral groundwater model coupled to VIC is a distributed, one-layer, two-dimensional groundwater model driven by groundwater recharge and groundwater pumping. Groundwater recharge is derived from interaction of the groundwater model with the VIC soil by allowing bi-directional exchange of water between the aquifer and the soil. A full description of the lateral groundwater model and coupling to VIC is given by Scheidegger et al. (2021).

The model is run on a 1/60° (~2 km) grid across the country and is driven with openly available global datasets. The model is parameterised with spatially distributed parameters from a range of sources that describe the land surface, including soil properties and vegetation properties. The soil properties such as field capacity, plant available water, wilting point, saturated hydraulic conductivity, and residual saturation for the VIC model are taken from a global high-resolution map of soil hydraulic properties (Zhang and Marcel 2018). Quartz fraction and bulk density values are from SoilGrid1km (Hengl et al. 2014). Landcover vegetation parameters are taken from Modis (Friedl and Sulla-Menashe 2015), leaf area index and albedo from Copernicus (Smets et al. 2019), and vegetation height from LiDAR-derived Global Estimates of Forest Canopy Height (Healey et al. 2015). The groundwater part of the model requires values for hydraulic conductivity and specific yield, which are classified based on the groundwater availability map of the Philippines (Bureau of Mines and Geo-Sciences and Ministry of Natural Resources 1986). A full description of the model is given by Scheidegger et al. (2022). The VIC model is driven by meteorological forcing data using a gridded, sub-daily time-series of meteorological variables as input. Average air temperature, total precipitation, atmospheric pressure, incoming shortwave radiation, incoming longwave radiation, vapor pressure, and wind speed are required. For the historical simulation (1979 – 2018), Copernicus ERA5 hourly data from 1979 to present are used (Hersbach et al. 2018). The meteorological forcing data is at 0.25°, and hence a much coarser resolution than the soil and vegetation parameters. Therefore, the meteorological forcing data were downscaled to match the model grid using the delta method (Moreno and Hasenauer 2016). In order to find the best preforming model, the soil infiltration capacity parameter the soil thickness, the specific yield and the hydraulic conductivity of the aquifer were varied. The model was calibrated against observed river flows at gauging stations available from the National Hydrologic Data Collection Program (Department of Public Works and Highways 2016). In order to increase the model skill, the model could be subdivided into different catchments and calibrated separately for each catchment. In the future, other measures to evaluate the model could be included, such as soil moisture changes from remote sensing data, evapotranspiration, and groundwater levels. This would improve the confidence of the variables separate to runoff. Figure 2 Comparison of simulated mean river flows and observed river flows obtained from Department of Public Works and Highways (2016). Comparison of observed and simulated flow shows that 96% of the simulated river flows are within one order of magnitude of the observed river flow, that 81% are within half an order of magnitude of observed stream flow, and 55% within a quarter of order of magnitude. This means that for a hypothetical stream flow of 108 m3/month one order of magnitude would be simulated flows between 107 m3/month to 109 m3/month, a range of half an order of magnitude is between 3.2×107 m3/month to 3.2×108 m3/month and a range of a quarter order of magnitude 5.6×107 m3/month to 1.8×108 m3/month . Only 28% of the simulated river flows are within 26% of observed flow. Reasons for the discrepancy are many, from the model conceptualisation, the change in flow regime after the 1991 eruption of Mount Pinatubo and the change in the flow regime thereafter, the lack of representing water management operations and irrigation practices, or the coarse resolution of the meteorological driving data and model parameterisation. Figure 3 Median runoff in mm/month and simulated and observed stream flows (Department of Public Works and Highways, 2016) for eight gauging stations. BGS © UKRI.

This document summarises the development of the first integrated surface and groundwater model of the Philippines through a collaboration between the British Geological Survey, Ateneo de Manila University, and the Philippine’s National Water Resources Board. Whilst ongoing work is required to improve it (for example, representations of water use and irrigation practice need to be included) the results of this first version of the model are very promising. We consider that it has the potential to underpin improved understanding of the hydrology of the Philippines, how future climate and anthropogenic change could affect water resources, and how management will need to adapt. Our vision is that it becomes a ‘community model’, i.e., an openly accessible tool continually developed by a range of users, and used a variety of stakeholders in the Philippines, especially those responsible for regulating and managing the country’s water resources and environment. Whilst there are challenges in doing this, this vision is certainly achievable if we can build a community of interested researchers and stakeholders. If you’d like to know more please contact us.

The ‘Future Projections of the Hydrology of the Philippines’ provides projections of hydrological change for the period 1980–2089 generated using the UKCP18 climate change projections based on two greenhouse gas ‘representative concentration pathways’ (RCP): RCP2.6 and RCP 8.5 (Met Office Hadley Centre 2018). RCP2.6 represents a mitigation scenario aiming to limit the increase of mean global mean temperature to around 1.6C° by 2100 above preindustrial level for mid-range climate sensitivity. In contrast, RCP8.5 is a pathway where greenhouse gas emissions continue to grow unmitigated and lead to an estimated global average temperature rise of 4.3°C above preindustrial level by 2100. It is to note that the estimated ranges for future climate are conditioned on a set of modelling, statistical, and dataset choice assumption with expert judgement in methodological and data choices by the producers of the UKCP18 dataset. The probabilities of the projection are interpreted as an indication of a particular future climate outcome for a given representative concentration pathway and there is more evidence for outcomes near the centre of the distribution than in the tails (Fung et al. 2018). Each of the two climate projections is composed of a 15-member ensemble, representing the uncertainty in the simulations of future climate.