1. Mesoscale dynamics of Southern California’s Climate Mimi Hughes

2. Mesoscale dynamics of Southern California’s Climate Mimi Hughes Outline:  Motivation  Datasets used  Precipitation distribution  Santa Ana winds Thanks to:  Alex Hall  Rob Fovell

3. Limitations of previous climate studies ~30km ~150km

4. Data

5. release 3.6.0 boundary conditions: Eta model analysis resolution: domain 1: 54 km, domain 2: 18 km, domain 3: 6 km 23 vertical levels. time period: May 1995 to April 2006 (re-initialized every 3 days) Parameterizations: MRF boundary layer Simple ice microphysics Clear-air and cloud radiation Kain-Fritsch 2 cumulus parameterization in coarse domains, only explicitly resolved convection in 6 km domain MM5 Configuration

6. One can think of this as a reconstruction of weather conditions over this time period consistent with three constraints: (1) our best guess of the large-scale conditions, (2) the physics of the MM5 model, and (3) the prescribed topography, consistent with model resolution. release 3.6.0 boundary conditions: Eta model analysis resolution: domain 1: 54 km, domain 2: 18 km, domain 3: 6 km 23 vertical levels. time period: May 1995 to April 2006 (re-initialized every 3 days) Parameterizations: MRF boundary layer Simple ice microphysics Clear-air and cloud radiation Kain-Fritsch 2 cumulus parameterization in coarse domains, only explicitly resolved convection in 6 km domain MM5 Configuration

7. Model Validation: Precipitation and winds Spatial Correlation: 0.87 Regression: slope = 1.13 intercept = 0.39 cm/month Correlation of simulated and observed daily mean wind anomalies at 18 stations. From Conil and Hall (2006)

8. Model Validation: Precipitation and winds Spatial Correlation: 0.87 Regression: slope = 1.13 intercept = 0.39 cm/month Correlation of simulated and observed daily mean wind anomalies at 18 stations. From Conil and Hall (2006)

9. North American Regional Reanalysis Eta model downscaling of NCEP; 32-km horizontal resolution (Mesinger 2006)

10. Precipitation Distribution

11. How can topography change the distribution of precipitation?

12. Flow over? (mechanical lifting…) Precipitation (grayscale) and topography (contours) for an idealized numerical study. From Jiang (2003) Wind As air moves over topography it is forced to rise, causing moisture to condense and fall out: P: Precipitation qU: Moisture flux h(x,y): Terrain See Smith (1979), Roe (2005), etc.

13. As air moves over topography it is forced to rise, causing moisture to condense and fall out: P: Precipitation qU: Moisture flux h(x,y): Terrain See Smith (1979), Roe (2005), etc. Flow over? (mechanical lifting…) Precipitation (grayscale) and topography (contours) for an idealized numerical study. From Jiang (2003) Is this too simple? Wind

14. Or Flow around? (aka blocked flow) Precipitation (grayscale) and topography (contours) for an idealized numerical study. From Jiang (2003) If the air approaching a barrier does not have enough kinetic energy to surmount it, the flow will be blocked (Smolarkiewicz and Rotunno, 1990; Pierrehumbert and Wyman, 1985). This can enhance precipitation upwind of the barrier. Wind

15. Case studies: Blocking influencing precipitation Medina and Houze (2003) compared two synoptic events during the mesoscale alpine program and found a substantial difference in precipitation and wind between them. –Less stable, higher wind speed case => winds uniform with height and precipitation greatly enhanced on the windward slope –More stable, lower wind speed case => winds shear in the lowest layers and precipitation more evenly distributed Neiman et. al. (2004) found that orographic blocking affected the propagation of the fronts during a storm from the 1997/98 season, substantially impacting the distribution of precipitation

16. Motivation: Approach: To investigate what processes are essential to predicting the distribution of precipitation in complex topography Systematic study using a hierarchy of models

17. Precipitation observations Cooperative Observation Precipitation measurements: average of daily rainfall from May 1995 to April 2006. Black contours show topography.

18. Winds during rain Vectors show wind speed and direction; colored contours show wind speed, in m/s.

19. Solid line shows linear regression. Large pale blue bullet is GPCP open-ocean average (119.5W- 121.5W, 31.5N- 32.5N) Upslope Model?

20. Questions I’ll address… Does orographic blocking occur during raining hours in Southern California? Does blocking significantly impact the climatological distribution of precipitation? Is there a simple way to get a quantitative estimate of the impact of blocking on precipitation?

21. Diagnosing Blocking

22. Brunt-Väisälä frequency: Depends on the moisture content of the atmosphere. When not saturated: When close to saturation (Durran and Klemp, 1982): Computing a bulk Froude number Average open ocean wind speed Barrier height: 1 km

23. Composite maps of normalized precipitation rate for rainy hours binned by Fr 2. Separation by Fr 2 : Precipitation

24. Separation by Fr 2 : Precipitation

25. How are the Froude number and the distribution of precipitation related?

26. Adapted from Roe (2005) High U 2 Small N 2 High Fr 2

27. Low Fr 2 Adapted from Jiang (2003) Low U 2 Large N 2

28. Vectors show wind speed and direction, normalized by open-ocean speed. Separation by Fr 2 : Surface winds

29. Quantifying the effect of blocking on precipitation

30. Linear model of orographic precipitation Relates the precipitation to the gradient of the terrain, with the additional complexity of three shifting terms to account for upstream tilted vertically propagating gravity waves, and advection of water droplets during condensation and fallout. (Smith 2003, Smith and Barstad 2004)

31. In Fourier space: Linear model of orographic precipitation Fourier transform of the terrain. Moisture coefficient Intrinsic frequency Depth of moist layer Hydrometeor fallout time Moisture conversion time Vertical wavenumber

32. In Fourier space: Where is the Fourier transform of the terrain. The inverse transform of gives the spatial distribution of precipitation once negative values are truncated and background rate is added. Linear model of orographic precipitation

33. Linear model: applied Spatial Correlation = 0.83 Precipitation distribution predicted by the Linear Model (LM) and the MM5 composite for the conditionally unstable hours.

34. Linear model: limitation Precipitation distribution predicted by the LM and the MM5 composite for the hours with lowest Fr 2.

35. Extent to which blocking affects precipitation distribution Spatial correlation of the LM with MM5 precipitation for different ranges of Fr 2

36. Extent to which blocking affects precipitation distribution Regression lines of MM5 precipitation/ slope relationship for different ranges of Fr 2.

37. Separation by Fr 2 : Percentage of precipitation

38. Summary We use a hierarchy of models to identify the processes essential for predicting precipitation distribution in complex topography. –Upstream blocking significantly modifies precipitation distribution in Southern California, contributing a substantial percentage of total precipitation, particularly at low elevation coastal locations. –Defining a bulk Froude number based on the ambient atmospheric conditions provides a useful measure of the extent to which blocking is affecting precipitation distribution. Exclusion of blocking effects is the main shortcoming of the linear model (LM), and including a term based on bulk Fr 2 might make the LM accurate for all cases.

39. Dynamics of the Santa Ana winds

40. Arrows show wind and colors show wind speed for average winds of the Santa Ana cluster defined in Conil and Hall (2006). Slope of regression onto SW component = SAt

41. Composite synoptic conditions on SA days Previous studies suggest Santa Anas form because of a synoptic scale pressure anomaly manifest at the surface as a high sea level pressure in the Great Basin (Raphael 2003; Sommers 1976).

42. A recent study by Gabersek and Durran (2006) showed how this pressure anomaly, which would cause strong ~offshore winds aloft, could cause strong surface winds. Synoptic forcing => strong surface winds

43. As strong upper level winds impinge on their idealized topography, the winds are accelerated through the gap.

44. The winds are primarily accelerated through downward momentum transport.

45. Composite synoptic conditions on SA days

46. Identify SA days by their synoptic conditions?

48. Many SA days have very weak synoptic forcing.

49. Synoptically controlled SAt? Offshore 2km desert wind (MM5) The increase in likelihood for large SAt with high gph anomaly is primarily due to the associated increase in mid-tropo. wind speed.

50. Synoptically controlled SAt? Large scale conditions only weakly constrain SAt, even if variations of mid- tropo. wind direction are considered. However, strong onshore 700 hPa winds do prevent strong offshore surface winds.

51. Local thermodynamic forcing Something else is the primary control on SAt. Conil and Hall (2006) found SAs were associated with anomalously cold desert surface temperatures: is this a clue?

52. Local thermodynamic forcing This large ocean-desert temperature gradient would cause an associated pressure gradient, forcing offshore winds at the surface.

53. Offshore 2km desert wind (MM5) Locally controlled SAt SAt is strongly related to the desert-ocean temperature gradient. As ∆T increases (=> cold desert), SAt increases.

54. Locally versus synoptically controlled SAt Correlation of gph anomaly and SAt = 0.5; correlation of ∆T and SAt = 0.78.

55. Offshore 2km desert wind (MM5) Conditional sampling based on ∆T and 2 km u Small u Large u Small ∆TLarge ∆T

56. Conditional sampling based on ∆T and u: Vertical structure

57. Conditional sampling based on ∆T and u: Vertical structure

58. Conditional sampling based on ∆T and u: Vertical structure Vertical wind profiles for the three composites show that the large ∆T cases have strong offshore winds trapped close to the surface.

59. Conditional sampling based on ∆T and u: Surface flow

60. Conditional sampling based on ∆T and u: Surface flow

61. Conditional sampling based on ∆T and u: Surface flow

62. Conditional sampling based on ∆T and u: Vertical structure Vertical wind profiles for the three composites show that the large ∆T cases have strong offshore winds trapped close to the surface. Can we quantify the contribution of these two processes to SAt?

63. Multivariate Regression model for SAt SAt, predicted = 0.041*u + 0.084*∆T + 0.22 Variance in SAt:

64. Conclusions Two features of Southern California’s climate are crucially dependent on mesoscale processes: These two processes depend on resolving the region’s complex topography, and thus are probably not well- represented in coarse resolution climate models. Precipitation distribution is strongly controlled by whether or not the air approaching the coastal mountains is blocked. Santa Ana winds develop primarily because of a local temperature gradient between the desert surface and the air over the ocean at the same altitude, with synoptic forcing playing an ancillary role.

65. Thanks!