Comparing the performance of process-based models for drought simulation in Scotland

Abstract

<p class="paragraph">Scotland is increasingly vulnerable to periods of dry weather, impacting water users and the natural environment. In 2022, large parts of Scotland have experienced water scarcity, resulting in Scotland Environmental Protection Act (SEPA) suspending water abstractions for abstraction licence holders in some Scottish catchments. To understand and manage these water scarcity events in Scotland, we need to monitor and model the drought processes. This research is a part of a Scottish Government funded project ‘Understanding the vulnerabilities of Scotland’s water resources to drought’ which has been co-constructed with a range of national level stakeholders and aims to understand what the specific impacts of droughts are and what are the vulnerabilities that may apply to Scotland under future change. This includes the understanding of the spatial variability and characteristics of future hydrological drought events and short-term forecasting of drought duration to inform adaptive catchment management, while considering water resources requirements of different user sectors. As a first step towards constructing a national short-term drought forecasting framework, w<span xml:lang="EN-GB" data-contrast="none">e</span><span data-ccp-charstyle="normaltextrun"> have</span> reviewed the <span data-ccp-charstyle="normaltextrun">state-of-the art </span><span data-ccp-charstyle="normaltextrun">hydrological modelling approaches</span> <span data-ccp-charstyle="normaltextrun">currently</span> <span data-ccp-charstyle="normaltextrun">applied in</span><span data-ccp-charstyle="normaltextrun"> the UK</span>. <span data-ccp-charstyle="normaltextrun">Our review suggests</span> a lumped conceptual model, <span data-ccp-charstyle="normaltextrun">GR</span>6J and <span data-ccp-charstyle="normaltextrun">a </span><span data-ccp-charstyle="normaltextrun">distributed hydrological response unit-based model, </span><span data-ccp-charstyle="normaltextrun">HYPE,</span> <span data-ccp-charstyle="normaltextrun">are </span><span data-ccp-charstyle="normaltextrun">the most appropriate hydrological models for</span><span data-ccp-charstyle="normaltextrun"> both</span><span data-ccp-charstyle="normaltextrun"> simulating </span>and<span data-ccp-charstyle="normaltextrun"> short-term</span> forecasting of <span data-ccp-charstyle="normaltextrun">droughts</span>, <span data-ccp-charstyle="normaltextrun">based on </span><span data-ccp-charstyle="normaltextrun">the following </span><span data-ccp-charstyle="normaltextrun">criteri</span>a<span data-ccp-charstyle="normaltextrun">: </span><span data-ccp-charstyle="normaltextrun">openly</span> <span data-ccp-charstyle="normaltextrun">available</span> model code, <span data-ccp-charstyle="normaltextrun">proven ability at </span><span data-ccp-charstyle="normaltextrun">simulating and forecasting low flows</span><span data-ccp-charstyle="normaltextrun">, </span><span data-ccp-charstyle="normaltextrun">and widely</span> used and supported <span data-ccp-charstyle="normaltextrun">m</span>odel. <span data-ccp-charstyle="normaltextrun">In </span>next steps, <span data-ccp-charstyle="normaltextrun">w</span>e <span data-ccp-charstyle="normaltextrun">will </span><span data-ccp-charstyle="normaltextrun">design</span> <span data-ccp-charstyle="normaltextrun">a common modelling framework </span><span data-ccp-charstyle="normaltextrun">for drought simulation and forecasti</span><span data-ccp-charstyle="normaltextrun">ng </span><span data-ccp-charstyle="normaltextrun">in </span>Scotland. <span data-ccp-charstyle="normaltextrun">Using both H</span><span data-ccp-charstyle="normaltextrun">YPE and GR6J, </span><span data-ccp-charstyle="normaltextrun">w</span>e <span data-ccp-charstyle="normaltextrun">will </span><span data-ccp-charstyle="normaltextrun">set up and test</span> both<span data-ccp-charstyle="normaltextrun"> models</span> in <span data-ccp-charstyle="normaltextrun">a</span> <span data-ccp-charstyle="normaltextrun">medium size </span><span data-ccp-charstyle="normaltextrun">long-term monitoring </span><span data-ccp-charstyle="normaltextrun">test catchment in</span> Tarland <span data-ccp-charstyle="normaltextrun">in northeast Scotland (~70km</span><span xml:lang="EN-GB" data-contrast="none"><sup>2</sup></span><span data-ccp-charstyle="normaltextrun">) where we have good process understanding and recent </span>hydro climatological datasets<span data-ccp-charstyle="normaltextrun">.</span> <span xml:lang="EN-GB" data-contrast="auto">C</span><span data-ccp-charstyle="normaltextrun">omparison of </span><span xml:lang="EN-GB" data-contrast="auto">the model performances</span> of HYPE and GR6J will guide us to take a decision on which model to move forward with for upscaling in Scotland. Machi<span xml:lang="EN-GB" data-contrast="none"><span data-ccp-charstyle="normaltextrun">ne learning approaches </span><span data-ccp-charstyle="normaltextrun">for low-flow forecasting using </span><span data-ccp-charstyle="normaltextrun">long-short-memory networks will also be explored in </span><span data-ccp-charstyle="normaltextrun">develop</span>ing</span><span data-ccp-charstyle="normaltextrun"> a multi-model </span><span data-ccp-charstyle="normaltextrun">drought </span>forecasting ensemble. <span data-ccp-props="{"> </span></p> <p class="paragraph"><span data-ccp-props="{">Keywords: Drought, water scarcity, modelling, HYPE, GR6J, forecasting </span></p>