Evaluation of continental-scale ensemble hydrological forecasts from Environment and Climate Change Canada: a comparison with forecasts from the Global Flood Awareness System (GloFAS)

Abstract

<p><span class="TextRun SCXW169027254 BCX8" lang="EN-CA" xml:lang="EN-CA" data-contrast="auto"><span class="NormalTextRun SCXW169027254 BCX8">Environment and Climate Change Canada (ECCC) is in the process of deploying a continental-scale hydrological prediction system </span><span class="NormalTextRun SCXW169027254 BCX8">known as the</span><span class="NormalTextRun SCXW169027254 BCX8"> National Surface and River Prediction System (NSRPS). <span class="TextRun SCXW224052350 BCX8" lang="EN-CA" xml:lang="EN-CA" data-contrast="auto"><span class="NormalTextRun SCXW224052350 BCX8">Operating at a resolution of 30 arc seconds, </span></span></span><span class="NormalTextRun SCXW169027254 BCX8">NSRPS currently generates</span><span class="NormalTextRun SCXW169027254 BCX8"> Ensemble Streamflow Predictions</span> <span class="NormalTextRun SCXW169027254 BCX8">(</span><span class="NormalTextRun SCXW169027254 BCX8">ESPs</span><span class="NormalTextRun SCXW169027254 BCX8">)</span><span class="NormalTextRun SCXW169027254 BCX8"> for over </span><span class="NormalTextRun SCXW169027254 BCX8">four</span><span class="NormalTextRun SCXW169027254 BCX8"> million grid </span><span class="NormalTextRun SCXW169027254 BCX8">points</span><span class="NormalTextRun SCXW169027254 BCX8">. </span><span class="NormalTextRun SCXW169027254 BCX8">Issued once per day, the ensemble is composed of twenty members and provides 16-day forecasts.</span> <span class="NormalTextRun SCXW169027254 BCX8">NSRPS</span><span class="NormalTextRun SCXW169027254 BCX8"> differs from other hydrological forecasting </span><span class="NormalTextRun SCXW169027254 BCX8">systems </span><span class="NormalTextRun SCXW169027254 BCX8">in </span><span class="NormalTextRun SCXW169027254 BCX8">operational </span><span class="NormalTextRun SCXW169027254 BCX8">use in Canada by being continental in scope and by relying on an Earth System Modelling (ESM) approach </span><span class="NormalTextRun SCXW169027254 BCX8">for </span><span class="NormalTextRun SCXW169027254 BCX8">prediction</span><span class="NormalTextRun SCXW169027254 BCX8">. </span><span class="NormalTextRun SCXW169027254 BCX8">I</span><span class="NormalTextRun SCXW169027254 BCX8">n order to assess the value of </span><span class="NormalTextRun SCXW169027254 BCX8">forecasts </span><span class="NormalTextRun SCXW169027254 BCX8">issued</span><span class="NormalTextRun SCXW169027254 BCX8"> by NSRPS, </span><span class="NormalTextRun SCXW169027254 BCX8">a comparison is performed </span><span class="NormalTextRun SCXW169027254 BCX8">with </span><span class="NormalTextRun SCXW169027254 BCX8">a</span><span class="NormalTextRun SCXW169027254 BCX8"> simila</span><span class="NormalTextRun SCXW169027254 BCX8">r </span><span class="NormalTextRun SCXW169027254 BCX8">ESP</span><span class="NormalTextRun SCXW169027254 BCX8"> system</span><span class="NormalTextRun SCXW169027254 BCX8"> available over all of Canada: </span><span class="NormalTextRun ContextualSpellingAndGrammarErrorV2Themed SCXW169027254 BCX8">the</span><span class="NormalTextRun SCXW169027254 BCX8"> Global Flood Awareness System (</span><span class="NormalTextRun SpellingErrorV2Themed SCXW169027254 BCX8">GloFAS</span><span class="NormalTextRun SCXW169027254 BCX8">) from the European Centre for Medium-Range Weather Forecasts (ECMWF). The evaluation focusses on the Great Lakes and St. Lawrence watershed as well as the Nelson and Churchill watershed</span><span class="NormalTextRun SCXW169027254 BCX8">s</span><span class="NormalTextRun SCXW169027254 BCX8">, each over </span><span class="NormalTextRun SCXW169027254 BCX8">one</span><span class="NormalTextRun SCXW169027254 BCX8"> million km² in size. 393 streamflow stations are identified where NSRPS and </span><span class="NormalTextRun SpellingErrorV2Themed SCXW169027254 BCX8">GloFAS</span><span class="NormalTextRun SCXW169027254 BCX8"> agree </span><span class="NormalTextRun SCXW169027254 BCX8">on</span><span class="NormalTextRun SCXW169027254 BCX8"> the watershed delineation</span><span class="NormalTextRun SCXW169027254 BCX8">. The comparison</span><span class="NormalTextRun SCXW169027254 BCX8"> is limited to</span><span class="NormalTextRun SCXW169027254 BCX8"> the Spring, Summer and Fall of 2022</span><span class="NormalTextRun SCXW169027254 BCX8"> due to NSRPS forecast data availability</span><span class="NormalTextRun SCXW169027254 BCX8">. For most stations, NSRPS performs better than </span><span class="NormalTextRun SpellingErrorV2Themed SCXW169027254 BCX8">GloFAS</span><span class="NormalTextRun SCXW169027254 BCX8"> in terms of Continuous Ranked Probability Score (CRPS), but the median of the potential CRPS across the 393 stations is very similar for days 3-16.</span><span class="NormalTextRun SCXW169027254 BCX8"> Both systems suffer from a lack of </span><span class="NormalTextRun SCXW169027254 BCX8">spread</span><span class="NormalTextRun SCXW169027254 BCX8">,</span> <span class="NormalTextRun SCXW169027254 BCX8">particular</span><span class="NormalTextRun SCXW169027254 BCX8">ly</span><span class="NormalTextRun SCXW169027254 BCX8"> for</span><span class="NormalTextRun SCXW169027254 BCX8"> short lead times, but the problem is slightly more acute for </span><span class="NormalTextRun SpellingErrorV2Themed SCXW169027254 BCX8">GloFAS</span><span class="NormalTextRun SCXW169027254 BCX8">.<span class="NormalTextRun SCXW224052350 BCX8"> </span></span><span class="NormalTextRun SCXW169027254 BCX8">Bayesian Model Averaging (BMA)</span><span class="NormalTextRun SCXW169027254 BCX8"> is explored in order to obtain calibrated probabilistic forecasts that perform better than both NSRPS and </span><span class="NormalTextRun SpellingErrorV2Themed SCXW169027254 BCX8">GloFAS</span><span class="NormalTextRun SCXW169027254 BCX8">.</span></span><span class="EOP SCXW169027254 BCX8" data-ccp-props="{"> </span></p>