2.4 The MAD-WRF solar irradiance nowcasting model: model overview and evaluation of the cloud initialization system.
Jimenez, P.A , G. Thompson, J. Dudhia and J.A. Lee, National Center for Atmospheric Research
Short term
predictions of up to a few hours (i.e. nowcasting) of solar irradiance require
accurate cloud initialization and representation of cloud evolution. Standard
nowcasting methodologies include the use of retrievals from earth observing
satellites to detect and advance the clouds. This methodology faces limitations
when microphysical processes control the cloud evolution or in cases wherein
clouds are anchored to terrain features. These processes are better represented
in numerical weather prediction (NWP) models. However, nowcasting systems based
on NWP models do not always include accurate satellite-based cloud
initialization. In addition, the clouds may dissipate if the initial conditions
are not favorable to sustain the clouds. To overcome these limitations, we are
integrating the Multi-Advection and Diffusion Nowcast (MADCast) methodology of
using a satellite-based initialization and WRF to advect and diffuse
microphysics quantities, with the physics of the WRF-Solar NWP model to create
an improved end-to-end solar irradiance forecast system, called MAD-WRF.
This presentation will provide an overview of the MAD-WRF model and an
evaluation of the cloud initialization system. The MAD-WRF cloud initialization
is based on observations of the cloud base / top height combined with a cloud
initialization parameterization. The parameterization relies on relative
humidity to identify the location of clouds and estimate the liquid water and
ice content. Alternatively, the parameterization can be used on top of
hydrometeors, if available, to enhance the initial cloud field. The performance
of the initialization is enhanced by assisting the cloud identification scheme
with cloud base / top height observations. The benefits of this cloud
initialization system are herein quantified with hourly WRF-Solar simulations
spanning the month of April 2018. The benefits of using cloud top height
retrievals from GOES16 and cloud base height observations from METAR stations
are also investigated. Results are compared against global horizontal
irradiance observations from the SURFRAD network and illustrate the added value
of the MAD-WRF cloud initialization system.