Principle Researchers: Eila Gendig Marwan Katurji Collaborators: Renan Le Roux Hervé Quénol VINEYARD SCALE CLIMATOLOGY Accuracy of forecasts and downscaling Brief Outline     Dynamical weather modelling Accuracy of the dynamic weather model (WRF: Weather Research and Forecasting system) Geo-statistical downscaling for enhanced spatial resolution Outlook Dynamical Weather Modeling      Weather Research and Forecasting Model (WRF) Hindcasting and forecasting Hourly data Range of climatological variables 1 km resolution W Soil/Surface type Topography Land use Global atmospheric Geographical boundary circulation data INPUT R F OUTPUT Aim  How good is the forecast? Observation   Air temperature, wind speed, wind direction,… Temporal Resolution     Model Seasonality? Diurnal changes? Meteorological phenomena Spatial Relationships   Coastal sites vs. inland sites Awatere Valley vs. Wairau Valley Automatic Weather Stations Validating Air Temperature I  1 growing season (8 months)  01/09/2013  to 30/04/2014 Daily mean air temperature (°C) Observation WRF Min Max Mean Min Max Mean 6.9 23.4 15.0 6.2 20.5 14.4 Bias = WRF – Observation = 14.4°C – 15.0°C = -0.6°C -5 0 5 Bias (°C) = ∆ Temperature (°C) Validating Air Temperature II Sunrise WRF too warm Sunset Wind Speed Bias Summary and Future Work  Weather model evaluation was done for one growing season. Mean daily air temperature bias from 11 weather stations was between -2.2 to 0.5 oC  Further model evaluation is required to relate model error to season, weather types and hour of the day  The latter will allow for a process based understanding of the model error. This is essential for tuning specific components of the weather modeling system Geostatistical Downscaling  Increasing spatial resolution  Enabling a closer look at the spatial variability of air temperature More accurate determination of growing degree days (GDDs)  Geostatistical Downscaling Landscape parameters Locality Slope Aspect Elevation Lower resolution temperature Multi-parameter regressive model Landscape parameters Higher resolution temperature Geostatistical Downscaling Air temperature 50 m grid resolution 1000 m grid resolution Summary and Future Work  Preliminary geo-statistical downscaling provides ultra-high spatial resolution of air temperature  Geo-statistical downscaling will be carried out for the entire growing season and evaluated in a similar way to the dynamical weather model  A combination of dynamical and statistical weather modeling will be coupled to the weather and GDD forecasts on wineclimate.co.nz for 2015.