Back page

ASSESSMENT OF THE BIAS, SPATIAL PATTERN AND TEMPORAL VARIABILITY OF ERRORS IN THE DIFFERENT SOURCES OF AREAL PRECIPITATION ESTIMATES

Workpackage number :

9

 

Start date or starting event:

+6

 

Partner code: Person-months

P1: 0

P2: 6

P3: 0

P4: 6

P5: 0

P6: 0

 

P7: 0

P8: 19

P9: 13

     

Objectives:

Areal precipitation estimates can be obtained from stochastic models, rain gauges, radar, satellite and NWP models. Each has its own levels of bias, spatial and temporal sensitivity. These must be assessed before any realistic procedure for combining the estimates or using one source to "correct" the other can be designed. This objective of this WP is to make that assessment.

Furthermore, the sensitivity of small and medium scale distributed hydrological models to bias, spatial and temporal variability in precipitation estimates will be assessed.

 

Methodology/work description :

The Radar, NWP forecasts, raingauge and satellite data are used to estimate rainfall fields from which areal rainfall is calculated. The estimates from the different sources are compared with a "baseline" estimate in terms of (i) total rainfall amounts (ii) spatial pattern and correlation structure. At Partner 4 the "baseline" will be the existing meso-scale analysis procedure. At Partner 9 the "baseline" will be a stochastic model, based only on previous rainfall which represents the best that can be achieved without concurrent measurements or telemetry. At Partner 2 the "baseline" will be represented by the identification of a Kriging model using past rain-gauge rainfall data. Spatial anomaly patterns will be identified for each source of estimates and compared with physical features of the catchment, e.g. its orography in the case of radar and raingauges or characteristic scales of the numerical model in the case of NWP.

Rainfall estimation fields from the 5th Framework Programme Project EURAINSAT (EVG1 CT-2000-00030) will be provided by Partner 8 as a contribution to the workpackage. Satellite-derived fields are particularly suited for this purpose given their space-time repetition. In particular the MSG-derived rainfall, which represents a key product of EURAINSAT, will be upgraded every 15 min. The bias of satellite rainfall fields, however, is expected to be higher than that of other sources. The trade-off between effective coverage and low bias is thus to be carefully checked in terms of the desired output, especially for flood monitoring and forecasting.

For each source, the rainfall estimates will be used to drive an hydrological model to verify their adequacy in both estimating and forecasting catchment runoff. In the case of Partner 4 will be used a the state of the art semi-distributed model. In the case of the smaller Partner 9 catchment, both the simpler lumped models (both catchment models and unit hydrographs) and the more complex distributed models (already developed at Partner 2) will be used. In all cases a comparison of the estimated with the measured floods will be used to assess any systematic areal bias in the precipitation estimates.

At partner 9, a Monte-Carlo stochastic procedure will be used to assess the sensitivity of distributed catchment models to the observed patterns of bias, spatial and temporal variability. Models of different complexity will be used, to investigate any relationship between model complexity and sensitivity to errors in the inputs Lumped catchment models will be compared with a fully distributed physically based model. A large number of realisations of forecast error patterns will be generated and the outflows from the models calculated. A detailed statistical analysis of the errors in these will be undertaken.

Basic minimum useful accuracy tolerances for the flood peak will be established in consultation with the end-user group.

 

Deliverables

  1. Comparison of precipitation estimates with mesoscale analysis, stochastic model and Kriging. Delivery date: 12.
  2. Comparison of flood estimates and forecasts using the semi-distributed model in the Swedish catchment Delivery date: 18.
  3. Comparison of flood estimates and forecasts using the simpler lumped models in the Irish catchment Delivery date: 18.
  4. Comparison of flood estimates and forecasts using the more complex distributed model in the Irish catchment Delivery date: 18.
  5. Assessment of precipitation errors - Delivery date: 36.

 

Milestones

M1

M2

M3 Delivery of a report on precipitation estimate assessment

M4

M5 Delivery of a report on flood estimation and forecasting assessment

M6

Back page