Ninth International Conference on Permafrost ... - IARC Research

Snowmelt in an Arctic Catchment: Application of the Hydrological ModelWATFLOOD in a Small Arctic Basin with Different Land Cover ClassesA. StrutzkePhilipps-University Marburg, GermanyCh. OppPhilipps-University Marburg, GermanyIntroductionThe world’s climate change affects arctic areas in manyways. Climate models predict considerable warming formost northern regions, especially the Western CanadianArctic. As a result, the western Arctic faces a shortenedsnow-covered season, changes in winter snow coverproperties, and changes in timing and volume of snowmeltwater runoff. Furthermore, the fragile ecosystem of theCanadian Arctic has to face a huge impact on its vegetationand animal species. Shrubs and trees will expand northwardsand to higher elevations, replacing existing plants. Animalsdependent on cold temperatures such as polar bears, seals,caribou, and reindeer are forced to move further north.The objective of this research project is to examinethe capability of the Canadian model WATFLOOD tosimulate the runoff in Arctic environments. WATFLOODwas developed for and, so far, is mainly used in southernCanadian river systems and large drainage basins. In arcticareas the physical conditions are very different from those inthe south: rivers freeze over completely in winter, the soil ispermanently frozen (permafrost), and the watersheds in mostcases are untouched by human influences. One of the mostimportant aspects of arctic watersheds is that over the longwinter, the precipitation falls almost exclusively as snow,which accumulates in the watershed. At the end of winter,snowmelt occurs, and most of the meltwater is releasedwithin days out of the basin. The end-of-winter snowmeltcreates the highest yearly runoff peak in spring, with littlerunoff in the short summer that follows.Modeling small watersheds in the Canadian Arctic isstill a young science. The Arctic itself is a data-poor areacompared to other well-researched areas in Canada or theworld. Applying a hydrological model to arctic areas relieson having measurement stations in the Arctic as well asdoing fieldwork to compare the modeled results with themeasured data in evaluating the model’s ability. The hopewith modeling in the Arctic is that scientists will be able tocome up with generalized parameter sets that can be usedelsewhere in the Arctic where observation stations do notexist.ResultsConsidering that WATFLOOD was developed for moretemperate regions than the Arctic, the validation runsshowed that WATFLOOD was able to simulate runoffin Hans Creek fairly accurately. WATFLOOD producedreasonable meltwater hydrographs in most years by onlyusing WATFLOOD’s relatively simple air temperatureindex algorithm to simulate snowmelt, together with thecalibrated soil and channel parameters for Hans Creek. Theaspect that WATFLOOD consistently estimates the daysof first runoff too early can be explained by two naturallyoccurring processes that are not specifically addressed in theWATFLOOD model: snow meltwater percolation throughthe snow cover and snow damming. Both of these processesdelay runoff by several days and are not accounted for bythe model, leading to the problems in simulating the first dayof runoff. Another major problem is the existence of manylakes in the Hans Creek watershed, where the outflow ofthe basin is located. This plays a huge role in the hydrologyof the study basin, slowing runoff considerably. The lakerichenvironment might also be an explanation for whyWATFLOOD was not able to simulate any runoff fromrainstorm events at the end of the summer, even though theywere clearly visible in the gauging station.The water storage in lakes and their particularly complexfreezing and melting patterns clearly were a problem forWATFLOOD. One reason for this could be that the landcover class “water,” as used for lakes in Hans Creek, wasbased on snow survey data from the generally very smalllakes in the neighboring Trail Valley Creek. Hans Creek,however, holds lakes in a huge variety of sizes, which wereall initiated in spring with the same amount of snow waterequivalent. This may lead to large modeling errors becauseof the extensive redistribution of snow during blowing-snowevents, which happen especially over lakes. Responsible isa lack of vegetation and the absence of melt/freeze cyclesthat could stabilize the snow cover. Additionally high windspeed events happen often over the long winter period in thisarea leading to frequent periods of blowing snow and highsublimation losses.In a subsequent model series, the model performancewas tested using climate data from a long-term observationstation in Inuvik, 60 km south of the Hans Creek watershed.While the overall average of total and peak runoff volumeswas fairly good, the individual hydrographs for the 15 yearsused in this series as well as the NSC clearly showed thatWATFLOOD could not simulate the runoff accurately formost of the years. Since WATFLOOD was able to predictthe runoff of the neighboring watershed Trail Valley Creek(Pohl et al. 2007) using Inuvik climate data, the reasonmust be sought in the differences between both watersheds.Likely reasons seem to be the bigger size of Hans Creek and305