1 / 14

Geom and Fake Case Verification with SAL Contribution from U Mainz

This study presents the concept of the SAL 3-component quality measure for verifying GEOM and FAKE cases, considering QPF in a specified region. The study includes the definition, examples, and findings of the SAL measure.

jdevries
Download Presentation

Geom and Fake Case Verification with SAL Contribution from U Mainz

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Verification of GEOM and FAKE cases with SAL Contribution from U Mainz Christiane Hofmann, Matthias Zimmer, Heini Wernli Kindly presented by Christian Keil (DLR) April 2008

  2. The concept of SAL 3-component quality measure that considers QPF in pre-specified region (e.g. river catchment): S structure -2 … 0 … +2 objects too perfect objects too small/peaked large/flat A amplitude -2 … 0 … +2 averaged precip perfect averaged precip underestimated overestimated L location0 … +2 perfect wrong location of total center of mass (TCM) and/or of objects rel. to TCM Exact definition, examples etc: Wernli et al. 2008 (MWR, in press, see AMS early online webpage)

  3. GEOM cases OBS GEOM 0 S A L MOD GEOM 0 0 0 0 perfect forecast GEOM 1 -0.01 -0.01 0.11 small displacement GEOM 2 -0.03 -0.03 0.42 large displacement

  4. GEOM cases OBS GEOM 0 S A L MOD GEOM 0 0 0 0 perfect forecast GEOM 1 -0.01 -0.01 0.11 small displacement GEOM 2 -0.03 -0.03 0.42 large displacement “noise” induced by construction of cases (interpolation leads to small changes in total precipitation amount)

  5. GEOM cases OBS GEOM 0 S A L MOD GEOM 0 0 0 0 perfect forecast GEOM 1 -0.01 -0.01 0.11 small displacement GEOM 2 -0.03 -0.03 0.42 large displacement GEOM 3 1.19 1.19 0.27 large overestimation of amount and size, intermediate displacem. GEOM 4 -0.02 -0.02 0.27 intermediate displacem. - no information about orientation of object! GEOM 51.55 1.55 0.28 very large overestimation amount and size, intermediate displacem.

  6. FAKE cases OBS FAKE 0 S A L MOD FAKE 0 0 0 0 perfect forecast FAKE 3 -0.02 0 0.03 small displacement due to interpolation -> slightly different choice of threshold for object identification -> weak spurious signal in S

  7. FAKE cases OBS FAKE 0 S A L MOD FAKE 0 0 0 0 perfect forecast FAKE 3 -0.02 0 0.03 small displacement FAKE 5 0.02 -0.16 0.15 large displacement

  8. FAKE cases OBS FAKE 0 S A L MOD FAKE 0 0 0 0 perfect forecast FAKE 3 -0.02 0 0.03 small displacement FAKE 5 0.02 -0.16 0.15 large displacement part of precipitation is shifted out of domain -> correctly identified as underestimation of amount

  9. FAKE cases OBS FAKE 0 S A L MOD FAKE 3 -0.02 0 0.03 small displacement FAKE 5 0.02 -0.16 0.15 large displacement FAKE 6 -0.03 0.38 0.03 overestimation of amount, small displacem. FAKE 7 -0.54 -0.300.04 underestimation of amount, small displacem.

  10. FAKE cases OBS FAKE 0 S A L MOD FAKE 3 -0.02 0 0.03 small displacement FAKE 5 0.02 -0.16 0.15 large displacement FAKE 6 -0.03 0.38 0.03 overestimation of amount, small displacem. FAKE 7 -0.54 -0.300.04 underestimation of amount, small displacem. S is sensitive to uniform reduction of precip values (in contrast to uniform scaling, cf. FAKE 6)!

  11. FAKE cases 3 vs. 7 Threshold contour for identification of objects: max. value in domain/15. FAKE 3/6: larger max. value -> larger threshold -> 1 large object FAKE 7: smaller max. value -> smaller threshold -> several objects -> S < 0

  12. R R threshold for object identification threshold for object identification • Summary • GEOM cases: • SAL results are OK • weak point: SAL does not provide information about orientation of objects (GEOM 4) • FAKE cases: • SAL results are also OK • interesting difference FAKE 6 vs. FAKE 7: uniform scaling (FAKE 6) does not lead to error in S, but uniform reduction (FAKE 7) does! Explanation: uniform reduction can lead to identification of more and smaller objects: • all except FAKE 7: 1 large object FAKE 7: two small objects

  13. S A L - Definition of the components A = (D(Rmod) - D(Robs)) / 0.5*(D(Rmod) + D(Robs)) D(…) denotes the area-mean value (e.g. catchment) normalized amplitude error in considered area A  [-2, …, 0, …, +2] L = |r(Rmod) - r(Robs)| / distmax+measure of distance of objects to r(…) r(…) denotes the centre of mass of the precipitation field in the area normalized location error in considered area L  [0, …, 2] S = (V(Rmod*) - V(Robs*)) / 0.5*(V(Rmod*) + V(Robs*)) V(…) denotes the weighted volume average of all scaled precipitation objects in considered area normalized structure error in considered area S  [-2, …, 0, …, +2]

More Related