Testbed: Ensemble Representations of Rainfall Observation and Analysis Uncertainty

The HEPEX Ensemble Representations of Rainfall Observation and Analysis Uncertainty Testbed was proposed at the 3rd HEPEX workshop, and formally launched during a side meeting of the 2009 EGU General Assembly. This cross-cutting testbed was established to look at the problem of representing uncertainty in precipitation observations through ensemble simulation approaches.

It is a common feature of precipitation products, whether interpolated raingauge measurements, radar or satellite retrievals, that their associated errors are both spatially and temporally correlated. Point rainfall error models thus give an incomplete representation of rainfall uncertainty. One mechanism that has been explored to overcome this problem is to generate an ensemble of modelled rainfall fields, each member of which is consistent with the available input data while variation between ensemble members is indicative of the associated uncertainties. Each ensemble member is additionally constrained to display ‘realistic’ rainfall structures in both space and time. In most work to date, these structures have been imposed through the use of geo-statistical models although other conditional modelling approaches could conceivably be applied. Ensemble observation techniques may be regarded as limiting cases of both ensemble downscaling approaches (taken to the limit of no change in resolution) and ensemble nowcasting algorithms (with the lead time reduced to zero).

The testbed launch meeting coincided with a dedicated themed session in the main EGU hydrology programme. Ensemble uncertainty representation has continued to be represented in the EGU programme to date, although following the first year this has been as a strand in broader themed sessions.

As with the other HEPEX testbeds, the original idea was to develop a collaborative scientific community working together on cross-cutting projects. In actuality, the testbed became more of a coordinating framework to report related work. This particular testbed was concerned with a relatively new field which is still only being researched by a handful of groups. Moreover, the related technologies are still under development, with a substantial research issues still unanswered before broad application becomes feasible.

The topic itself is thus still moving towards the ‘take-off’ point and experience with this testbed may therefore not directly relate to other testbeds. However, there remains enormous unexplored potential in this topic which should be fostered through the revised HEPEX structure and elsewhere.