Introducing Land Surface Perturbations in the Global Ensemble Forecast System

The current Global Ensemble Forecast System (GEFS) is under-spread near the surface. This is a widespread deficit

of ensemble forecast systems, and is especially the case for near-surface variables including surface temperature.

Here, we seek methods to perturb surface parameters and initial conditions that increase spread and are physically

based. The surface perturbations introduced here address two major sources of uncertainty in ensemble prediction.

To address initial condition uncertainty in the land model, we apply perturbations based on EOFs of differences

between normalized soil moisture and temperature states from different LSMs. To address uncertainty within the

land model, perturbations of parameters, including roughness lengths for heat and momentum, parameters related

to soil hydraulic conductivity, stomatal resistance, vegetation fraction and albedo, are applied, with the amplitude

and perturbation scales based on previous research. We also extend the stochastically perturbed physical tendencies

(SPPT) to include soil moisture and soil temperature. We find that overall surface perturbations have a modest

impact on near-surface temperature and other variables. In particular, soil initial condition perturbations have the

most impact on near surface temperature spread in arid and semi-arid regions. When considering bias-corrected

RMSE (i.e. with mean model error removed), the combined surface perturbations increase the spread to match RMSE

in the Tropics and the summer hemisphere. The results indicate that surface perturbations, through their impact on

near-surface spread, have a positive impact on the skill of short-range ensemble forecasts.