1. Adjoint Models as Analytical Tools Dr. Ronald M. Errico Goddard Earth Sciences and Technology Center (UMBC) Global Modeling and Assimilation Office (NASA)

2. Outline 1. Sensitivity analysis: The basis for adjoint model applications 2. Examples of adjoint-derived sensitivity 3. Code construction: an example 4. Nonlinear validation 5. Efficient solution of optimization problems 6. Adjoint models as paradigm changers 7. Singular vectors 8. Observation sensitivity 9. Other applications 10. Other considerations 11. Summary

3. Sensitivity Analysis: The basis for adjoint model applications Impacts vs. Sensitivity

6. xixi xjxj

8. A single impact study yields exact response measures (J) for all forecast aspects with respect to the particular perturbation investigated. A single adjoint-derived sensitivity yields linearized estimates of the particular measure (J) investigated with respect to all possible perturbations. Adjoint Sensitivity Analysis Impacts vs. Sensitivities

12. Examples of Adjoint-Derived Sensitivities

13. Contour interval 0.02 Pa/m M=0.1 Pa/m Example Sensitivity Field Errico and Vukicevic 1992 MWR

14. Lewis et al. 2001

15. From Errico and Vukicevic 1992 J=average surface pressure in a small box centered at P J=barotropic component of vorticity at point P

16. Development of Adjoint Model From Line by Line Analysis of Computer Code

17. Development of Adjoint Model From Line by Line Analysis of Computer Code

18. Y = X * (W**A) Z = Y * X Ytlm = Xtlm * (W**A) + Wtlm *A* X *(W**(A-1)) Ztlm = Ytlm * X + Xtlm * Y Xadj = Xadj + Zadj * Y Yadj = Yadj + Zadj * X Xadj = Xadj + Yadj * (W**A) Wadj = Wadj + Yadj * X *(W**(A-1)) Development of Adjoint Model From Line by Line Analysis of Computer Code Parent NLM : TLM : Adjoint :

19. Y = X * (W**A) Z = Y * X Ytlm = Xtlm * (W**A) + Wtlm *A* X *(W**(A-1)) Ztlm = Ytlm * X + Xtlm * Y Xadj = Xadj + Zadj * Y Yadj = Yadj + Zadj * X Xadj = Xadj + Yadj * (W**A) Wadj = Wadj + Yadj * X *(W**(A-1)) Development of Adjoint Model From Line by Line Analysis of Computer Code Parent NLM : TLM : Adjoint :

20. Development of Adjoint Model From Line by Line Analysis of Computer Code Automatic Differentiation TAMC Ralf Giering (superceded by TAF) TAF FastOpt.com ADIFOR Rice University TAPENADE INRIA, Nice OPENAD Argonne Others www.autodiff.org See http://imgi.uibk.ac.at/MEhrendorfer/ work_7/present/session3/giering.pdf

21. Development of Adjoint Model From Line by Line Analysis of Computer Code 1. TLM and Adjoint models are straight-forward to derive from NLM code, and actually simpler to develop. 2. Intelligent approximations can be made to improve efficiency. 3. TLM and (especially) Adjoint codes are simple to test rigorously. 4. Some outstanding errors and problems in the NLM are typically revealed when the TLM and Adjoint are developed from it. 5. It is best to start from clean NLM code. 6. The TLM and Adjoint can be formally correct but useless!

22. Nonlinear Validation Does the TLM or Adjoint model tell us anything about the behavior of meaningful perturbations in the nonlinear model that may be of interest?

23. Linear vs. Nonlinear Results in Moist Model 24-hour SV1 from case W1 Initialized with T’=1K Final ps field shown Contour interval 0.5 hPa Errico and Raeder 1999 QJRMS

24. Non-Conv Precip. ci=0.5mm Convective Precip. ci=2mm Convective Precip. ci=2mm Non-Conv Precip. ci=0.5mm Linear vs. Nonlinear Results in Moist Model

25. Non- Convective Convective. Comparison of TLM and Nonlinearly Produced Precip Rates 12-Hour Forecasts with SV#1 Errico et al. QJRMS 2004 Linear Nonlinear Contours: 0.1, 0.3, 1., 3., 10. mm/day

26. Linear vs. Nonlinear Results In general, agreement between TLM and NLM results will depend on: 1.Amplitude of perturbations 2.Stability properties of the reference state 3.Structure of perturbations 4.Physics involved 5.Time period over which perturbation evolves 6.Measure of agreement The agreement of the TLM and NLM is exactly that of the Adjoint and NLM if the Adjoint is exact with respect to the TLM.

27. Efficient solution of optimization problems

33. M Contours of J in phase (x) space P Gradient at point P

34. Adjoint models as paradigm changers

35. Sensitivity field for J=p s with respect to T for an idealized cyclone From Langland and Errico 1996 MWR 100 hPa 1000 hPa 500 hPa

36. Contour 0.1 unitContour 1.0 unit Adjoint of Nudging Fields Bao and Errico: MWR 1997

37. t=0 t=48 hours Errico et al. Tellus 1993 0.4 units 5.0 units

38. Non-Convective Forecast Precipitation Rate Contour interval 2 cm/day Errico, Raeder, Fillion: 2003 Tellus What is the modeled precipitation rate sensitive to?

39. Errico et al. 2003 Tellus Impacts for adjoint- derived optimal perturbations for forecasts starting indicated hours in the past. Vort optimized Rc optimized Rn Optimized

40. Errico et al. QJRMS 2004 Initial u, v, T, ps PerturbationInitial q Perturbation 12-hour v TLM forecasts Perturbations in Different Fields Can Produce the Same Result

41. Singular Vectors

42. Gelaro et al. MWR 2000

43. Bred Modes (LVs) And SVs Gelaro et al. QJRMS 2002 Results for Leading 10 SVs

44. Vertical modes of the 10-level MAMS model (from Errico, 2000 QJRMS) Balance of Singular Vectors

45. Errico 2000 Balance of Singular Vectors t=0 t=24 hours E=E_t, K=KE, A=APE, R=R mode E, G=G mode E A,G E,K,R

46. The Balance of Singular Vectors Initial R mode Initial G mode Contour 2 units Contour 1 unit Errico 2000

47. Singular Value Squared Mode Index Errico et al. Tellus 2001 EM E-norm Moist Model ED E-norm Dry Model TM R-norm Moist Model TD R-norm Dry Model How Many SVs are Growing Ones?

48. From Novakovskaia et al. 2007 and Errico et al. 2007

49. Sensitivity to Observations The use of an adjoint of a data analysis algorithm

51. Sensitivity to analyzed potential temperature at 500 hPa Sensitivity to raob temperatures at 500 hPa From Gelaro and Zhu 2006 J = mean squared 24 hour forecast error using E-norm

52. Baker 2000 PhD. Thesis

54. (J/Kg) …all observing systems provide total monthly benefit Impacts of various observing systems Totals GEOS-5 July 2005 Observation Count (millions) (J/Kg) NH observations SH observations From R. Gelaro

55. AMSU-A (15 ch)AIRS (153 ch) H 2 O Channels Diagnosing impact of hyper-spectral observing systems Channel (J/Kg)Forecast Error Reduction (J/Kg) GEOS-5 July 2005 00z Totals Forecast degradation (J/Kg) -0.6-7.0 …some AIRS water vapor channels currently degrade the 24h forecast in GEOS-5… 00 From R. Gelaro

56. Other Applications 1.4DVAR (P. Courtier) 2.Ensemble Forecasting (R. Buizza, T. Palmer) 3.Key analysis errors (F. Rabier, L. Isaksen) 4.Targeting (R. Langland, R. Gelaro)

57. Problems with Physics

58. Consider Parameterization of Stratiform Precipitation R 0 qqs NLM TLM Modified NLM

59. Sensitivity of forecast J with respect to earlier T in lowest model model level Time= -3 hours; contour int.=0.00025Time= -9 hours; contour int.=10000. From R. Errico, unpublished MAMS2 development Example of a failed adjoint model development

60. Tangent linear vs. nonlinear model solutions Errico and Raeder 1999 QJRMS TLM 60x small pert NLM NLM

61. RAS scheme ECMWF scheme BM scheme Fillion and Mahfouf 1999 MWR Jacobians of Precipitation

63. Problems with Physics 1.The model may be non-differentiable. 2.Unrealistic discontinuities should be smoothed after reconsideration of the physics being parameterized. 3. Perhaps worse than discontinuities are numerical insta- bilities that can be created from physics linearization. 4. It is possible to test the suitability of physics components for adjoint development before constructing the adjoint. 5. Development of an adjoint provides a fresh and complementary look at parameterization schemes.

64. Other Considerations Physically-based norms and the interpretations of sensitivity fields

65. Sensitivity of J with respect to u 5 days earlier at 45 O N, where J is the zonal mean of zonal wind within a narrow band centered on 10 hPa and 60 O N. (From E. Novakovskaia) 10 hPa 0.1 hPa 1000 hPa 100 hPa 1 hPa - 180 + 180 0 Longitude

66. Continuous vs. grid-point representations of sensitivity

67. Rescaling options for a vertical grid Delta log p Delta p 500 hPa 100 hPa 10 hPa 1 hPa 850 hPa

68. 1000 hPa 10 hPa 0.1 hPa Mass weightingVolume weighting 2 Re-scalings of the adjoint results From E. Novakovskaia

69. Summary

71. Unforeseen Results Leading to New Paradigms 1. Atmospheric flows are very sensitive to low-level T perturbations. 2. Evolution of a barotropic flow can be very sensitive to perturbations having small vertical scale. 3. Error structures can propagate and amplify rapidly. 4. Forecast barotropic vorticity can be sensitive to initial water vapor. 5. When significant R is present, moist enthalpy appears a key field. 6. Nudging has strange properties. 7. Relatively few perturbation structures are initially growing ones. 8. Sensitivities to observations differ from sensitivities to analyses. 9. A TLM including physics can be useful.

72. Misunderstanding #1 False: Adjoint models are difficult to understand. True: Understanding of adjoints of numerical models primarily requires concepts taught in early college mathematics.

73. Misunderstanding #2 False: Adjoint models are difficult to develop. True: Adjoint models of dynamical cores are simpler to develop than their parent models, and almost trivial to check, but adjoints of model physics can pose difficult problems.

74. Misunderstanding #3 False: Automatic adjoint generators easily generate perfect and useful adjoint models. True: Problems can be encountered with automatically generated adjoint codes that are inherent in the parent model. Do these problems also have a bad effect in the parent model?

75. Misunderstanding #4 False: An adjoint model is demonstrated useful and correct if it reproduces nonlinear results for ranges of very small perturbations. True: To be truly useful, adjoint results must yield good approximations to sensitivities with respect to meaningfully large perturbations. This must be part of the validation process.

76. Misunderstanding #5 False: Adjoints are not needed because the EnKF is better than 4DVAR and adjoint results disagree with our notions of atmospheric behavior. True: Adjoint models are more useful than just for 4DVAR. Their results are sometimes profound, but usually confirmable, thereby requiring new theories of atmospheric behavior. It is rare that we have a tool that can answer such important questions so directly!

77. What is happening and where are we headed? 1.There are several adjoint models now, with varying portions of physics and validation. 2. Utilization and development of adjoint models has been slow to expand, for a variety of reasons. 3. Adjoint models are powerful tools that are under-utilized. 4. Adjoint models are like gold veins waiting to be mined.

78. Recommendations 1.Develop adjoint models. 2.Include more physics in adjoint models. 3.Develop parameterization schemes suitable for linearized applications. 4. Always validate adjoint results (linearity). 4.Consider applications wherever sensitivities would be useful.

79. Adjoint Workshop The 8 th will be in fall 2008 or spring 2009. Contact Dr. R. Errico rerrico@gmao.gsfc.nasa.gov to be put on mailing list