Evaluation and Assimilation of Remotely- Sensed Lake Surface Temperature in the HIRLAM Weather Forecasting System H. Kheyrollah Pour 1, C.R. Duguay 1, L. Rontu 2 with contributions from Kalle Eerola 2, Ekaterina Kurzeneva 2 1 Interdisciplinary Centre on Climate Change (IC 3 ) and Department of Geography and Environmental Management, University of Waterloo, Ontario, Canada 2 Finnish Meteorological Institute (FMI), Helsinki, Finland Third Workshop on Parameterization of Lakes in Numerical Weather Prediction and Climate Modelling Finnish Meteorological Institute, Helsinki, September 18-20 2012
Lakes in regional weather and climate Consideration of lake-atmosphere interactions is an important issue in climate modeling and numerical weather prediction (NWP). Lakes have an important role in the surface radiation balance, heat and water vapor exchanges with the atmosphere. The presence (or absence) of ice cover on lakes in winter has an effect on the surrounding climate. Earlier/later freeze-up and break-up results in ice cover duration change, and this strongly influences the radiation and energy balance. A good representation of lake ice/temperature-atmosphere interactions is necessary to improve weather forecasting and climate modelling
Objective To improve the objective analysis of lake surface state and forecasting (local cloud, temperature and precipitation) in the operational HIRLAM forecasting system. Use in-situ observations (e.g. SYKE).
LWST In situ measurements 1970 The number of in situ sites reporting ice observations has plummeted since the 1980s. 1995 Very little in situ data are available for the North, and many lakes are in remote regions. IGOS Cryosphere Theme Report (2007) Except for a very limited number of studies, in situ Lake Water Surface Temperature (LWST) measurements are generally not available for northern lakes
Objective To improve the objective analysis of lake surface state and forecasting (local cloud, temperature and precipitation) in the operational HIRLAM forecasting system. Use in-situ observations (e.g. SYKE). Use a thermodynamical lake model(e.g. FLake). o FLake has shown quite realistic results in the simulation of lake freezing and provides a good background for independent analysis of LST in the operational HIRLAM. o Model has been found to be too sensitive during the melt period and to predict early melt.
Objective To improve the objective analysis of lake surface state and forecasting (local cloud, temperature and precipitation) in the operational HIRLAM forecasting system. Use in-situ observations (e.g. SYKE). Use a thermodynamical lake model (e.g. FLake). Use remote-sensing observations (e.g. MODIS, AATSR). How retrieved satellite LWST observations can be used to improve weather forecasting?
Data Assimilation Basic idea of data assimilation is to combine measurements and models to obtain an initial condition for NWP. The more accurate the estimate of the initial conditions, the better the quality of the forecast. -In situ -Satellite - Climatology - Previousanalysis - Lake models Observations Background Objective analysis Initial conditions Forecast model
Evaluation of MODIS LWST for Lake Paijanne 30 25 MODIS ( C) In-situ Temperature ( C) 20 15 10 5 0 Apr-07 Jul-07 Oct-07 Jan-08 Apr-08 Jul-08 Oct-08 Jan-09 Apr-09 Jul-09 Oct-09 MODIS-LWST ( C) Jan-10 Apr-10 Jul-10 Oct-10 Jan-11 Apr-11 Jul-11 Oct-11 30 25 20 15 10 5 0 Paijanne (2007-2011) Ia = 0.954 MBE = -1.81 RMSE = ± 2.34 n = 407 0 5 10 15 20 25 30 SYKE-LWT ( C)
MODIS versus AATSR for 15 Finnish lakes (2007-2009) MODIS-LWST ( C) 30 25 20 15 10 5 0 Tuusulanjärvi Pyhajarvi Pielinen Pesiojarvi Pielinen Paajarvi1 Ounasjarvi Nasijarvi Lappajarvi Langelmave Haukivesi Kevojarvi Kilpisjarvi Kuivajarvi Kyyvesi 0 5 10 15 20 25 30 SYKE-LWT ( C) MODIS I a = 0.942 MBE = -0.68 RMSE = ± 2.16 n= 3535 AATSR-LWST ( C) 30 25 20 15 10 5 0 Tuusulanjärvi Pyhajarvi Pielinen Pesiojarvi Pielinen Paajarvi1 Ounasjarvi Nasijarvi Lappajarvi Langelmave Haukivesi Kevojarvi Kilpisjarvi Kuivajarvi Kyyvesi AATSR 0 5 10 15 20 25 30 SYKE-LWT ( C) I a = 0.935 MBE = 1.03 RMSE = ± 2.34 n= 754
Evaluation of MODIS with in situ GSL Mid-June to September 2003 25 T (MODIS - In situ) In-situ (0m) MODIS LST 20 I a = 0.94 MBE = - 0.90 RMSE = ±1.36 Temperature ( C) 15 10 5 0-5 -10 Station 5 June July August September Kheyrollah Pour, H., Duguay, C. R., and Yerubandi, R. R., Analysis of lake and land surface temperature patterns during the open water and ice growth seasons in the Great Bear and Great Slave Lake region, Canada, from MODIS (2002-2010), in preparation.
Evaluation of MODIS LWST data 10 9 T (MODIS-In situ) In-Situ MODIS Nighttime_ST 1 Nighttime Temperature ( C) 8 7 6 5 4 3 2 1 I a = 0.58 MBE = - 0.64 RMSE = 1.27 0-1 -2-3 -4 July Aug. Sep. Daytime Temperature ( C) 10 9 8 7 6 5 4 3 2 1 0 T (MODIS-In situ) In-Situ MODIS Daytime_ST 1 I a = 0.50 MBE = - 0.14 RMSE = 1.24-1 -2-3 -4 July Aug. Sep. Kheyrollah Pour, H., Duguay, C. R., and Yerubandi, R. R., Analysis of lake and land surface temperature patterns during the open water and ice growth seasons in the Great Bear and Great Slave Lake region, Canada, from MODIS (2002-2010), in preparation.
MODIS versus 3-D Princeton Ocean Model (POM) MODIS LWST (Aug.-Sep., 2008) Modelled LWST (Aug.-Sep., 2008)
Limitation of cloud cover Ladoga December 17, 2009 Visible Image Thermal Image
Bands 7-2-1 Lake Ice Fraction 25 April 2011 Analysis 6 UTC Bands 3-6-7 NOFLAK Background: Previous analysis Observation: MODIS + SYKE KARLAK Background: FLake model Observation: MODIS + SYKE Differences
Experiment set-up KARLAK and NOFLAK April 25, 2011 KARLAK experiment Pixel 15 Background: FLake Observation: MODIS + in situ (SYKE) Analysis result KARLAK May 8, 2011 NOFLAK Background: Previous analysis Observation: MODIS + in situ (SYKE) Analysis result NOFLAK experiment
Ice fraction for Lake Ladoga derived from MAGIC classification of SAR scenes 22 February 2009 MODIS derived LWST -11.85-7.41-0.89-4.49-3.03-6.35-4.19-8.19
Discussion MODIS LST data used in HIRLAM experiments resulted in a more realistic lake analysis outcome, which can be considered as a significant step towards improving future weather forecasts in cold regions. Improved analysis does not lead to an improved weather forecast as long as there is no real connection between the analyzed (observed) and predicted state of lakes. Climate models use observations only for calibration and validation, not as data assimilation. But it is possible to benefit from the use of satellite lake observations to improve parameterization and therefore climate models.
Future Plan Update MODIS observations, continuing HIRLAM experiments, analyzing the results systematically. Run a new HIRLAM experiment using MODIS observations for all major Scandinavian lakes within the HIRLAM domain. Work on structure functions in optimal interpolation to consider lake depth and elevation in the analysis. Validate forecasting results with weather station observations using 2-m air temperature, cloud cover, precipitation, and surface energy budget to test the sensitivity of the forecasts by changing initial conditions.
Lake Louise, Banff National Park, April 2011 Thanks