1. What is land-atmosphere feedback on precipitation? Precipitation wets the surface... …causing soil moisture to increase... …which causes evaporation to increase during subsequent days and weeks... …which affects the overlying atmosphere (the boundary layer structure, humidity, etc.)... …thereby (maybe) inducing additional precipitation Lecture 9 Land and Climate: Modeling Studies

2. Perhaps such feedback contributes to predictability. Short-term weather prediction with numerical models (e.g., those shown on the news every night) are limited by chaos in the atmosphere. Establish atmospheric state Initialize model with that state; integrate into future Short-term (~several days) weather prediction days Relevance of initial conditions Decay reflects short timescale of atmospheric “memory” Atmosphere Saturday’s forecast for Tuesday (March 23, 2004): sunny, high of 46F (8C).

3. For longer term prediction, we must rely on slower moving components of the Earth’s system, such as ocean heat content and soil moisture. Establish ocean state, land moisture state Initialize model with those states; integrate into future Long-term (~weeks to years) prediction of ocean and/or land states Associated prediction of weather, if weather responds to these states months Relevance of initial conditions Ocean Land For soil moisture to contribute to precipitation predictability, two things must happen: 1. A soil moisture anomaly must be “remembered” into the forecast period. 2. The atmosphere must respond in a predictable way to the remembered soil moisture anomalies

4. Observational soil moisture measurements give some indication of soil moisture memory. Soil moisture timescales of several months are possible. “The most important part of upper layer (up to 1 m) soil moisture variability in the middle latitudes of the northern hemisphere has … a temporal correlation scale equal to about 3 months.” (Vinnikov et al., JGR, 101, 7163-7174, 1996.) Vinnikov and Yeserkepova, 1991 Part 1: Soil Moisture Memory

5. Delworth and Manabe (1988) analyzed soil moisture memory in the GFDL GCM and came up with a Markovian framework for characterizing it. We will discuss Delworth and Manabe’s soil moisture memory analysis further during the round- table discussion. D&M’s memory analysis was recently furthered at Goddard...

6.   Koster and Suarez, 2001

7. Memory equation:

8. Seasonality term:  wn /  wn+1

9. Evaporation term: cR n /C s (and equivalently, the runoff term: aP n /C s )

10. (the covariance term)

11. This analysis allows us to examine soil moisture memory in terms of both large-scale forcing and inherent LSM behavior (e.g., through a and c terms, which describe the sensitivity of evaporation and runoff to soil moisture). The memory equation reduces to that of Delworth and Manabe under several simplifying assumptions.

12. Recent idealized experiment to analyze soil moisture memory (Sarith Mahanama, GSFC) -- A perpetual July experiment was performed to investigate the effect of precipitation and net radiation on soil moisture memory. Two different LSMs (the Mosaic LSM and the NSIPP-Catchment LSM) were given identical water holding capacities, vegetation type, soil type etc. and were forced under a variety of artificially generated climates. -- The imposed climates had: average monthly precipitations ranging from 15 to 500mm average monthly net radiation ranging from 20 to 400mm (Water equivalent) -- Essentially, within the idealized framework, the intermodel differences of soil moisture memory result solely from intermodel differences in the sensitivity of evaporation and runoff to soil moisture variations. -- A total of 400 different “climates” were imposed on each LSM. The simulation associated with each climate was a 200-month perpetual July simulation. Sub-monthly distributions of the variables followed those of the PILPS2c 1979-July forcing data for a chosen region.

13. Idealized Experiment to analyze soil moisture memory Autocorrelation of soil moisture (  ) in different climates:

14. Idealized Experiment to analyze soil moisture memory Differences in autocorrelation of soil moisture (  ) in different climates:

15. Superposition of ISLSCP-I July net radiation and precipitation on memory difference plot:

16. Supplemental analysis of globally simulated soil moisture memory with the two different models. When the Mosaic and Catchment LSMs are given the same soil moisture holding capacities, the Catchment LSM indeed shows higher memory for intermediate dryness index. When the Mosaic and Catchment LSMs are given their own, model-specific soil moisture holding capacities, the memory differences are seen to be largely a func- tion of the difference in capacity. (I.e., to some extent, a larger water holding capacity implies a larger memory.)

17. Part 2: Atmosphere’s Response to Soil Moisture Anomalies Three ways of looking for evidence of atmospheric response: 1. Examine observational data. Very difficult. (See next lecture.) 2. Simple analytical models. 3. AGCM studies. Useful for several reasons: (a) full set of diagnostic out- puts, (b) inclusion of nonlinearities, and (c) ability to do sensitivity studies. Advantage: feedbacks can be quantified and easily understood. Disadvantage: ignores some nonlinearities and complexities of system. Examples: Rodriguez-Iturbe et al., WRR, 27, 1899-1906, 1991. Brubaker and Entekhabi, WRR, 32, 1343-1357, 1996. Liu and Avissar, J. Clim, 12, 2154-2168, 1999.

18. AGCM evidence goes way back... Shukla and Mintz (1982) provide one of the first AGCM studies demonstrating the impact of land moisture anomalies on precipitation: Questions that can be addressed with an AGCM: How large is the impact of a land anomaly on the atmosphere? What are the relative roles of ocean variability, land variability, and chaotic atmospheric dynamics in determining precipitation over continents?

19. Studies examining the impact of “perfectly forecasted” soil moisture on the simulation of non-extreme interannual variations. Some examples: Delworth and Manabe, J. Climate, 1, 523-547, 1988. Dirmeyer, J. Climate, 13, 2900-2922, 2000. See GSWP lecture Douville et al., J. Climate, 14, 2381-2403, 2001. Dry conditions 1988 conditions1987 conditions Wet conditions Koster et al., J. Hydromet., 1,26-46, 2000. Simulations used: Ensemble 1: Sixteen 45-year simulations at 4 o X5 o with Interactive land surface processes Prescribed interannual-varying SST Ensemble 2: Sixteen 45-year simulations at 4 o X5 o with Fixed land surface processes (but with realistic interannual variations) Prescribed interannual-varying SST Round-table discussion

20. Description of this last study... # of Total Exp. simulations Length years Description A 4 200 yr 800 AL 4 200 yr 800 AO 16 45 yr 720 ALO 16 45 yr 720 Prescribed, climatological land; climato- logical ocean Interactive land, climato- logical ocean Prescribed, climatological land, interan- nually varying ocean Interactive land, interan- nually varying ocean SSTs set to seasonally-varying climatological means (from obs) SSTs set to interannually-varying values (from obs) LSM in model allowed to run freely Evaporation efficiency (ratio of evaporation to potential evaporation) prescribed at every time step to seasonally-varying climatological means Koster et al., J. Hydromet., 1, 26-46, 2000

21. Analysis of the simulation output shows that land and ocean contribute differently to continental precipitation variability. Annual precipitation variances Seasonal precipitation variances (from a similar 1995 study)

22. Simulated precipitation variability can be described in terms of a simple linear system:   ALO =   AO [ X o + ( 1 - X o ) ]   ALO   AO Total precipitation variance Precipitation variance in the absence of land feedback Fractional contribution of ocean processes to precipitation variance Fractional contribution of chaotic atmospheric dynamics to precipitation variance Land-atmosphere feedback factor

23. Contributions to Precipitation Variability

24. A variable  is defined that describes the coherence between the different precipitation time series. In an additional ensemble, every member of the ensemble is subject to the same time series of evaporation efficiency. Does the precipitation respond coherently to this signal? More from Koster et al. (2001) Results for SST control over precipitation coherence:

25. Koster et al. (2001) (cont.) Boreal summer Boreal winter Results for SST and soil moisture control over precipitation coherence Differences: an indication of the impacts of soil moisture control alone

26. Why does land moisture have an effect where it does? For a large effect, two things are needed: a large enough evaporation signal a coherent evaporation signal – for a given soil moisture anomaly, the resulting evaporation anomaly must be predictable. Both conditions can be related to relative humidity: The dots show where the land’s signal is strong. From the map, we see a strong signal in the transition zones between wet and dry climates. Koster et al. (2001) (cont.) Evap. coherence

27. Why does land-atmosphere feedback occur where it does? One control: Budyko’s dryness index variance amplification factor The results of this study could be highly model-dependent. A critical question about a critical issue: how does the atmosphere’s response to soil moisture anomalies vary with AGCM? We address this with...

28. Part 1: Establish a time series of surface conditions (Simulation W1) Step forward the coupled AGCM-LSM Write the values of the land surface prognostic variables into file W1_STATES Step forward the coupled AGCM-LSM Write the values of the land surface prognostic variables into file W1_STATES time step ntime step n+1 (Repeat without writing to obtain simulations W2 – W16) Part 2: Run a 16-member ensemble, with each member forced to maintain the same time series of surface prognostic variables (Simulations R1 – R16) Step forward the coupled AGCM-LSM Throw out updated values of land surface prognostic variables; replace with values for time step n from file W1_STATES Step forward the coupled AGCM-LSM time step ntime step n+1 Throw out updated values of land surface prognostic variables; replace with values for time step n+1 from file W1_STATES Coupled large scale Ongoing experiment: GLACE, a follow-on to a pilot coupled model intercomparison experiment. (K02: Koster et al., Comparing the degree of land-atmosphere interaction in four atmospheric general circulation models, J. Hydromet., 3, 363- 375, 2002.) … the GLACE Experiment Part 3: Same as Part 2, but only reset the deep soil moisture variables.

29. How does GLACE build on K02? Participation from a wider range of models. The idea is to generate a comprehensive “table” of coupling strengths, a table that can help in the interpretation of the published results of a wide variety of models. Separation of the effects of “fast” and “slow” reservoirs. The K02 results largely reflect the specification of the “fast” reservoirs (e.g., surface temperature). They thus may have little relevance to issues of seasonal prediction. Effect on air temperature. Ignored in the K02 study is the effect of the specification of surface variables on the evolution of air temperature. (This is a particularly interesting issue when only the “slow” soil moisture reservoirs are specified.) Correction of miscellaneous technical issues. Lessons learned from the K02 study can be applied immediately to GLACE.

30. 5. NCAR Kanae/Oki2. U. Tokyo w/ MATSIRO Xue12. UCLA with SSiB Koster11. NSIPP with Mosaic Lu/Mitchell10. NCEP/EMC with NOAH Taylor9. Hadley Centre w/ MOSES2 Sud8. GSFC(GLA) with SSiB Gordon7. GFDL with LM2p5 Verseghy6. Env. Canada with CLASS Kowalczyk4. CSIRO w/ 2 land schemes Dirmeyer3. COLA with SSiB McAvaney/Pitman1. BMRC with CHASM ContactModel Participating Groups Status submitted

31. All simulations in ensemble respond to the land surface boundary condition in the same way  is high Simulations in ensemble have no coherent response to the land surface boundary condition  is high

32. Ω p (R - W): GFDL GEOS CSIRO-CC3NSIPPCCCma NCEPBMRCHadAM3 UCLACOLA CSIRO-CC4 Impact of all land prognostic variables on precipitation

33. Ω p (S - W): GFDL GEOS CSIRO-CC3NSIPPCCCma NCEPBMRCHadAM3 UCLACOLA CSIRO-CC4 Impact of sub-surface soil moisture on precipitation

34. In principle, imposing land surface boundary states should decrease the intra-ensemble variance of the atmospheric fields. Idealized pdf of precipitation at a given point, across ensemble members corresponding pdf when land boundary is specified We look at the variance ratios:  2 P (S)  2 P (W)  2 P (R)  2 P (W) and

35. Variance(R) / Variance(W): Impact of all land prognostic variables on precipitation GFDL GEOS CSIRO-CC3NSIPPCCCma NCEPBMRCHadAM3 UCLACOLA CSIRO-CC4

36. Variance(S) / Variance(W): GFDL GEOS CSIRO-CC3NSIPPCCCma NCEPBMRCHadAM3 UCLACOLA CSIRO-CC4 Impact of sub-surface soil moisture on precipitation

37. Ω T (S - W): Impact of sub-surface soil moisture on temperature GFDL GEOS CSIRO-CC3NSIPPCCCma NCEPBMRCHadAM3 UCLACOLA CSIRO-CC4

39. Do models show any agreement regarding where land-atmosphere interaction is important?

40. Experiment website: http://glace.gsfc.nasa.gov

41. Rind, Mon. Weather Rev., 110, 1487-1494. June 1 initialized dry Beljaars et al., Mon. Weather Rev., 124, 362-383…. Wet initial- ization Dry initial- ization Differ- ences Such studies include Oglesby and Erickson, J. Climate, 2, 1362-1380, 1989. Also: How about AGCM studies that only initialize the soil moisture? (I.e., studies that don’t prescribe soil moisture throughout the simulation period?)

42. Impact of Soil Moisture Predictability on Temperature Prediction (darker shades of green denote higher soil-moisture impact) Predictability Timescale Estimate (via memory) Actual Predictability Timescale (diagnostics of precipitation show a much weaker soil-moisture impact) …and a study by Schlosser and Milly (J. Hydromet., 3, 483-501, 2002), in which the divergence of states in a series of parallel simulations was studied in detail: for soil moisture Some recent studies have examined the impact of soil moisture initialization on forecast skill (relative to real observations). These will be discussed in the next lecture.