The Grid-to-Grid model, or G2G, is used throughout the year to inform flood guidance given by the Flood Forecasting Centre (FFC) over England & Wales and by the Scottish Flood Forecasting Service (SFFS) over Scotland. A different approach to forecasting the snowmelt contribution to river flow is used by each organisation. The SFFS uses the G2G Snow Hydrology module with MetUM (Met Office Unified Model) inputs of precipitation and air temperature to form and melt the snowpack. In contrast, G2G for FFC uses snowmelt from the MetUM land surface scheme. These different configurations of G2G have been in use since 2011/12 and are in need of review, leading to the project reported on here on behalf of the FFC and SFFS.The report begins with a brief review of alternative snow hydrology methods to provide context for the methods used operationally, and to identify opportunities for improvement. An overview of the G2G Model is given, by way of background, and then the detail of the G2G Snow Hydrology module is set down. Functional extensions to G2G, made to help analyse and interpret model performance, are outlined and then applied to two catchments over Scotland. The G2G Snow Hydrology formulation is extended to use elevation zones within each G2G grid-cell. A separate air temperature adjustment for elevation is made for each zone using a lapse rate. Snowpack evolution is thereby represented in more spatial detail through the effect of temperature on rain/snow partitioning, melt and drainage. Model behaviour is analysed over the winter of 2009-10 in Scotland. The zonal formulation facilitated a closer comparison with observed snow depths at given locations. However, when snowpack properties are aggregated to the G2G grid-cell level, the values differ little to those from the operational G2G model. In turn, it has little effect on model performance at catchment scales, when assessed using gauged river flows.The approach used by the FFC for modelling snowmelt with G2G is then explained. It employs rainfall, snowfall and snowmelt as Gridded Post-Processing (GPP) products obtained from MetUM precipitation, wet-bulb temperature and snowmelt outputs. MetUM snowmelt employs the MOSES/JULES land surface model scheme; it is noted that an amended formulation to use multi-layers was introduced in March 2016. Inconsistencies between the MetUM outputs and GPP products that will impact on G2G performance at times of snow are identified.The alternative component methods, for rainfall/snowfall partitioning and for calculating snowmelt, forms the basis of a structured comparison of performance using different method combinations over three winters (2012 to 2015) in both Scotland and England & Wales. This evidence base of performance is used to draw general conclusions on choice of method. The highest performing methods were those in operational use, in the case of SFFS benefiting from a revised calibration made possible by the extended project dataset. The bias scores for the FFC operational configuration are poorer and possibly due to the inconsistencies referred to above. Throughout the report, models are assessed in simulation-mode (no data assimilation of river flow) and assuming foreknowledge of gridded raingauge rainfall with NWP precipitation used instead when snowfall is diagnosed (either through air temperature or NWP snowfall). This places emphasis on assessing the method formulation without being confounded by errors in the NWP forecasts.Inspection of observed and modelled hydrographs highlights that, even when accounting for snowmelt processes using the different methods, substantial uncertainties remain when simulating river flows at times of snow. This suggests that one focus of future work should be the use of weather model ensembles to quantify and better understand this apparent uncertainty.A final section provides an overall summary and closes with a set of recommendations, made separately for the FFC and SFFS. Some further considerations are discussed, including system constraints, product issues, ongoing plans and indicative scheduling/costing of proposed future activities.
PDF
download
878987797
028-85220240
