Meteorological Forcing Dataset
The main effort during the first interim period has been to update our global macroscale modeling to near-realtime over Africa. Previously we have developed a meteorological forcing dataset for global land areas for 1950-2000 to force our VIC model. The challenge has been to update the forcing dataset (and thence the VIC simulation) to near-realtime using available data streams. For 1950-2000, the meteorology was derived from a blending of reanalysis (NCEP/NCAR) and gridded observation-based datasets including the Climatic Research Unit's TS2.0 monthly precipitation and temperature dataset, the NASA Tropical Rainfall Measurement Mission (TRMM) 3-hourly precipitation products and the NASA Surface Radiation Balance (SRB) short- and long-wave datasets. In effect the observation datasets are used to spatially downscale the reanalysis, which is available at high temporal resolution, and at the same time remove biases in the reanalysis. This work is described in detail in Sheffield et al. (2006).
To bridge the data gap between the beginning of 2001 and near'realtime we have extended these methods to blend reanalysis with available observations. Although reanalysis data are available up to realtime, most observation-based datasets are generally only available some months of even years behind realtime. Therefore for 2001-realtime we have used a number of different datasets depending on their availability. For 2001-2006, we have used the recently updated (to 2006) monthly gridded precipitation and temperature dataset of Willmott and Matsura. This matches well the CRU dataset (used for 1950-2000) over their overlap period at large scales. From the beginning of 2007, we have used the Global Precipitation Climatology Project (GPCP) monthly dataset which is available a few months off realtime. Ongoing work is looking at the differences between these various datasets during their overlap periods and methods to ensure temporal consistency. For the last few months up to realtime, we are relying on realtime precipitation products (PERSIANN data from University California Irvine, TRMMM data from NASA) and gauge telemetry (Global Telecommunication System (GTS) gauge data from NOAA). These products are being downloaded on a daily basis and are blended into a forcing dataset for VIC over Africa.
With initial versions of the meteorological forcings in place we have extended the VIC simulation up to near-realtime and are currently running it operationally. This website is now disseminating these data and drought-specific products. These include available fields of meteorological variables from the realtime data streams, and hydrological variables generated by the VIC model: surface temperature, soil moisture (at various levels), evapotranspiration, runoff and streamflow. These fields are provided in a consistent framework in terms of temporal and spatial resolution, and format through the web portal.
Our immediate objectives are to finalize the data streams for the realtime running of the VIC model. This involves assessing the availability of input data, given the problems with operational monitoring, and develop fall-back methods for when data are unavailable or fail quality control checks. As the realtime meteorological data are likely biased, we will periodically re-run the VIC model up to a few months off realtime when the long-term gridded observation-based products (which are our best estimates of precipitation and temperature) are updated, to avoid drift in the land surface states.
We will implement drought screening based on recent work by Andreadis et al (2005) and Sheffield et al (2004). These approaches are based on comparison of current conditions for model-simulated soil moisture and runoff with an historic climatology in a way that maps the physical values to empirical probabilities based on the long-term global retrospective simulation described above. We estimate probability distributions of total column soil moisture and runoff for each grid cell and each month, against which current conditions can be compared. The screening tools account for drought areal extent and duration using concepts adapted from Andreadis et al (2005), which involve a form of spatial cluster analysis to identify drought patterns from gridded model output. Based on the historic analysis, we will establish a set of severity-area-duration thresholds that can be used to screen evolving droughts. Within the real time monitoring framework, we will monitor where drought thresholds are crossed for either soil moisture or runoff. Once the prescribed drought thresholds have been crossed, we will continue to track drought evolution in time (i.e., in subsequent forecasts) until our nowcasts indicate that the drought has dissipated. Drought dissipation will be evaluated in comparison with severity-area-duration thresholds estimated using an approach similar to the one used to establish drought screening thresholds.
Andreadis, K.M., E.A. Clark, A.W. Wood, A.F. Hamlet, and D.P. Lettenmaier, 2005. 20th Century drought in the conterminous United States, J. Hydromet. 6, 985-1001
Sheffield, J., G. Goteti, F. Wen, and E.F. Wood, 2004. A simulated soil moisture based drought analysis for the USA. J. Geophys. Res., 109, D24108, doi:10.1029/2004JD005182.
Sheffield, J., G. Goteti, and E.F. Wood, 2006. Development of a 50-yr, high resolution global dataset of meteorological forcings for land surface modeling. J. Climate, (13), 3088-3111.