AMIP II Diagnostic Subproject 12:
Land-Surface Processes and Parameterizations
(a joint AMIP/PILPS
Tom Phillips1, Ann Henderson-Sellers2, Parviz Irannejad2, Kendal McGuffie3 , Huqiang Zhang4
1Program for Climate Model Diagnosis and Intercomparison,
LLNL, California, USA
2 Environment, Australian Nuclear Science and Technology Organisation, Menai NSW, Australia
3Department of Applied Physics, University of Technology, Sydney, Australia
4 Bureau of Meteorology Research Centre, Melbourne, Australia
Last Update: 28 March, 2002
Because the land surface is the locus of most human activities, realistic simulation of continental weather and climate is of crucial importance. Land-surface parameterizations in today's atmospheric models exhibit a wide range of complexity: classic "bucket" models (e.g., Manabe 1969) and detailed soil-vegetation-atmosphere transfer schemes (SVATs) (e.g., Dickinson et al. 1993, Sellers et al. 1996) occupy opposite ends of the spectrum, while most such schemes are SVATs that exemplify an intermediate degree of complexity (e.g., Henderson-Sellers 1996).
The perceived need to systematically analyze these diverse schemes motivated the World Climate Research Programme to launch the Project for the Intercomparison of Land-surface Parameterization Schemes (PILPS) in 1992. PILPS diagnoses the behaviors of participating land-surface schemes (hereafter, "PILPS schemes") in controlled experiments implemented in 4 phases. The first two phases include studies of scheme behavior when driven in "off-line" (one-way feedback) mode by atmospheric forcings prescribed from GCM output (Phase 1) or from varied observational data sets (Phase 2). PILPS Phase 3 entails the diagnosis of land-surface schemes coupled to their "home" atmospheric host models, while Phase 4 concerns the analysis of results from coupling different land-surface schemes to a common host (Henderson-Sellers et al. 1996).
In practice, Phase 3 has involved the analysis of land-surface schemes in the AMIP models, organized as Diagnostic Subproject No. 12 (DSP12). This joint AMIP/PILPS proposal elaborates the rationale for continuing this subproject in AMIP II. Supporting details are provided on our Web home page http://www.cic.mq.edu.au/pilps-rice/.
In AMIP I, we diagnosed the continental climatologies of some 10 models with PILPS schemes (for a review of this work, see Henderson-Sellers 1999). Analysis of large-scale climatic differences revealed the presence of outlier variables in all the simulations (Love and Henderson-Sellers 1994, Love et al. 1995). Diagnosis of regional-scale surface moisture partitioning indicated that the inter-model scatter in midlatitudes was comparable to that of the analogous PILPS off-line experiment, despite of the presence of two-way feedbacks in the AMIP runs (Irannejad et al. 1995). Subsequent diagnosis of two cases of repeated AMIP I experiments made with simple bucket vs SVAT schemes in the same atmospheric model (Qu and Henderson-Sellers 1996, Phillips 1997) revealed widespread statistically significant differences in continental temperature, pressure, and turbulent fluxes--results in accord with those of analogous studies (e.g., Sato et al. 1989, Thompson and Pollard 1995, Yang et al. 1995). Thus, all our AMIP I analyses highlight the substantial differences in simulated continental climate that result from coupling diverse land-surface parameterizations to atmospheric models.
We should note, however, that the makeup of many AMIP I models imposed certain constraints on our analysis. First, the spectrum of land-surface scheme complexity was not well represented, in that most models employed bucket schemes rather than SVATs. Also, a number of AMIP I simulations were compromised by nonconservation of energy and moisture and/or pronounced trends in soil moisture and snow mass ( Love and Henderson-Sellers 1994 , Robock et al. 1998).
The AMIP II models are likely to show significant improvements in both these respects, in that a richer variety of land-surface schemes that have been tested in PILPS off-line experiments will be represented, and greater attention will be paid to model spin-up and related experimental procedures. Thus, a more meaningful investigation of the nature and import of differences in the simulation of continental climate should be possible.
Our AMIP I analysis also was limited by the dearth of suitable validation data. Now, with the availability of several reanalyses that span the AMIP II years and supplementary land-surface data that cover at least part of this period, validation is a more feasible, though still problematical, undertaking. The problems are not only limited to data issues. For example, Gates et al. (1996) note the current absence of a definitive formalism for the validation of coupled (in the AMIP context, land + atmosphere) climate models. They recommend, however, that model validation proceed along three paths:
(i) evaluation of the results of coupled model simulations
(ii) tests of individual components of coupled models
(iii) sensitivity studies of component coupling
Validation path (i) implies that a necessary, but not sufficient, condition for confidence in coupled model results is "adequate simulation" of the current climate--a concept that has not been well defined or implemented to date. The participation of PILPS in AMIP II presents a unique opportunity to formulate a definition of "adequate simulation" of land-surface climate, and to evaluate the performance of current-generation GCMs accordingly. In such an evaluation, the relationship between model performance and land-surface scheme complexity will be of particular interest. This is our first objective.
Path (ii) of the validation of land-surface scheme prediction entails testing whether essential processes are captured. From PILPS Phase 2 off-line studies, we have found that many schemes are able to simulate such processes to varying degrees, but that in general the performance of any individual scheme is inferior to that of the ensemble mean when a range of time and space scales are considered (e.g., Shao and Henderson-Sellers 1996). In addition, while Koster and Milly (1997) demonstrate that essential hydrological processes in almost all land-surface schemes can be simply represented at monthly time scales, Qu et al. (1998) show that the detailed behaviors and degree of sensitivity of these processes to increased surface air temperature are very different. Moreover, from preliminary experience pursuing validation path (iii) in the coupled experiments of PILPS Phase 4 (e.g. Peylin et al. 1997, Schulz et al. 1997), we have found that differences in the simulation of continental climate tend to play out most strongly in particular regions. These results highlight the need to evaluate land-surface scheme behavior in select locations where processes critical to prediction capability can be carefully studied. When this is attempted in the context of AMIP II, the analysis methodology must take account of substantial inter-model differences in surface forcings. This is our second objective.
Finally, it should be noted that the different phases of PILPS are organized
to yield results along all three of these validation paths. Our subproject
addresses path (i) and identifies regions in which path (iii) should be
pursued; Phases 1 and 2 mainly follow path (ii), while Phase 4 emphasizes
path (iii) and, to a lesser extent, path (i). As much as possible, therefore,
we will draw insights from, and connections to, other PILPS initiatives
in order to deepen our insights on the workings of land-surface schemes
in the AMIP II models. This is our third objective.
To summarize, our objectives in AMIP II are:
1) Validate the simulated continental climates, and assess the relationship of model performance to land-surface scheme complexity.
2) Identify key regions where continental processes are most sensitive to the choice of land-surface scheme, and analyze inter-model differences there.
3) Draw insights from, and connections to, other phases of PILPS.
Here we describe a methodology designed to meet our objectives concerning the validation and analysis of differences in AMIP II simulations of land-surface variables. We also briefly describe the relationship of our subproject to the activities of other phases of PILPS.
We propose to validate the large-scale continental climate of each AMIP II model, recognizing the problematical nature of such an undertaking: in spite of recent advances in deriving global observational datasets, many land-surface variables are not directly observable. Such model validation, although inherently provisional, will serve as an initial assessment of the performance of today's GCMs in simulating continental climate. We anticipate that recommended additions to the WGNE list of standard model diagnostics for land-surface variables also will result from this effort.
In the context of AMIP II, validation of a model's continental climate reflects not only on the performance of the land-surface scheme, but on that of the coupled system as a whole. We will place highest priority on validating variables that reflect the response of the land surface more than the atmospheric forcings (while acknowledging that forcing and response are not cleanly separable in coupled mode). These "response variables" include the surface latent and sensible heat fluxes, the ground temperature, and hydrological variables that include the evaporation + sublimation, runoff + drainage, and moisture stores in soil and snowpack. In order to interpret the land-surface response, we also will need to validate selected "forcing variables" such as the net downwelling radiative fluxes, liquid/solid precipitation, near-surface gradients of temperature and humidity, and surface wind stresses. At lowest priority, validation of precipitable water, cloud cover, and surface pressure may clarify the nature of inter-model differences in forcing or response variables. (Our ultimate choice of validation variables will attempt to avoid duplicating the efforts of other diagnostic subprojects as well.)
As initial reference data, we will use the ECMWF, NCEP/NCAR, and NASA reanalyses, since these include estimates of the land-surface variables of interest and span the entire AMIP II period (1979-1995). Moreover, differences among the reanalyses will serve as estimates of the observational uncertainties in each validation variable. However, because the reanalyses reflect the biases of the respective analysis models (e.g., Betts et al. 1997), wherever possible we also will use other global validation references that are based primarily on gauge or satellite observing systems for applicable portions of the AMIP period. Some examples include:
- ISLSCP temperatures, radiative fluxes, and hydrometeorology: currently available for 1987-1988, eventually for 1987-1996
- SRB surface short-wave radiation data: currently available for 1985-1988
- CMAP and GPCP merged gauge-satellite monthly precipitation data: available for the entire AMIP period
- GRDC river runoff data: available for varied time periods
As objective summaries of the goodness-of-fit of each model climatology to the reference dataset, we will compute standard statistics such as pattern correlations, RMS differences, and variance ratios, but also multivariate measures of the overall spatio-temporal fidelity of each simulation (e.g. Wigley and Santer 1990, Taylor 2001). We will supplement these numerical summaries with selective use of zonal mean plots and maps of seasonal and/or annual climatologies.
As an outcome of our validation effort, we will note systematic model biases in the simulation of particular land-surface variables. We also will examine the relationship (or lack thereof) between overall model performance and land-surface scheme complexity, as defined by various metrics (e.g., Henderson-Sellers et al. 1995). In this regard, the collective performance of AMIP II models with bucket schemes will be especially instructive: although buckets were outliers in the PILPS off-line experiments, in coupled mode the presence of compensating feedbacks may drive their associated climatologies within the envelope of observational uncertainties. We also will assess whether models using modified bucket schemes (e.g., inclusion of spatially varying field capacity, stomatal resistance, etc.) outperform those retaining the classical formulation of Manabe (1969).
At the large-scale, diagnosis of inter-model differences will come as a byproduct of model validation. In addition, we will examine land-surface processes in several key regions, so as to elucidate detailed differences in scheme behaviors. This regional analysis of differences will entail more qualitative interpretation, and less statistical computation, than our validation effort. We will focus initially on those AMIP II models with PILPS schemes to formulate initial hypotheses on possible reasons for regional differences, reserving the remainder of the models/land-surface schemes for hypothesis testing.
The primary criterion for selection of a key region is that the local response variables must exhibit significant inter-model differences as measured, for example, by a battery of multivariate statistical tests. Secondary considerations will include whether the region shows sensitivity to changes in land-surface scheme when coupled to the same atmospheric host (see Relationship to Other Phases of PILPS), and whether the region overlaps a GEWEX Continental Experimental Projects (e.g. BALTEX, GAME, GCIP, LBA, MAGS) that can provide a reality check on the simulations. From our AMIP I analysis and initial experience in coupling different SVATs to the same atmospheric host (e.g. Peylin et al. 1997, Schulz et al. 1997 ), candidate regions include Amazonia, Saharan Africa, Tibetan Plateau, Siberia/Northeast Asia, Northern Canada, and Central/Western USA. After examining the AMIP II data, we will make a final selection of several regions that are diverse in both climate and biome.
Of all the regional processes of potential interest, we will pay particular attention to surface evaporation, since this variable link the moisture and energy budgets. However, because of inter-model forcing differences that complicate single-variable comparisons, we will concentrate on differences in the relationships between surface evaporation and other variables. We may track these connections, for example, by plotting the average month-to-month regional variation in surface evaporation (latent heat flux) against those of other moisture (energy) variables for each model. Here, our objective will not be to "explain" the inter-model differences (probably an unattainable goal for such coupled nonlinear systems), but only to trace the associated chain of processes/feedbacks these differences reflect. An example: The relatively high sensible heat flux in model A is consistent with reduced snow melt -> reduced soil moisture -> reduced evaporation -> reduced latent heat flux.
We also may diagnose differences in the temporal variability of surface evaporation that are likely to depend on the land-surface scheme (Koster and Suarez 1994, 1995). For example, lagged autocorrelations of monthly evaporation can quantify differences in climatic persistence (Scott et al. 1995). Contingent on the availability of optional high-frequency data, we may investigate differences in the variability of surface evaporation at daily time scales as well. These diagnostics can corroborate whether the variability of surface evaporation for bucket schemes in coupled mode is generally skewed toward lower frequencies than that for SVATs (Scott et al. 1997), and whether there are substantial differences in variance structure among the SVATs.
In addition, we will investigate inter-model differences in regional partitioning of moisture and energy (e.g., runoff and Bowen ratios) and/or in the local moisture convergence (P - E). These basic "signatures" of a land-surface scheme (Chen et al. 1997, Koster and Milley 1997, Koster and Suarez 1994, Laval et al. 1996) are likely to be less sensitive to inter-model forcing differences than are single variables.
Where it is necessary to compare single response variables across the models, we will explore the utility of dividing by the appropriate forcing variable (e.g., net downwelling shortwave radiation or precipitation), as a zeroth-order scaling technique. This dimensionless-variable approach has shown promise in analyzing repeated AMIP I runs with different LSSs coupled to the same atmospheric model with (Phillips 1997) and in climate-change experiments with multiple models/LSSs (Gedney et al. 2000).
In our analysis, we will strive to draw insights from other phases of PILPS. For example, the results of the PILPS Phase 1 and 2 off-line experiments will provide general guidance for interpreting the regional behavior of the same schemes in coupled mode. These insights are likely to be most pertinent where our selected regions overlap the geographical locations of the off-line experiments (e.g. Pitman et al. 1993, Henderson-Sellers 1996, Chen et al. 1997, Schlosser et al. 1997, Wood et al. 1997).
In addition, complementary PILPS Phase 4 studies of different land-surface
schemes coupled to the same atmospheric host model will be especially relevant.
Thus, by coordinating our diagnosis with that of Phase 4, we expect to
deepen our insights on coupled land-atmosphere interactions.
We will require the following AMIP II model
Table 2: Single-level low frequency (monthly mean) output
Surface (2 m) air temperature
Mean sea-level pressure
Total precipitation rate
Snowfall rate (water equivalent)
Convective precipitation rate
Total soil frozen water content
Surface soil water content (upper 0.1 m)
Total soil water content
Total runoff (including drainage)
Snow depth (water equivalent)
Surface (10 m) eastward wind
Surface (10 m) northward wind
Surface (2m) specific humidity
Surface sensible heat flux
Surface latent heat flux
Surface evaporation plus sublimation rate
Eastward surface wind stress
Northward surface wind stress
Surface incident shortwave radiation
Surface reflected shortwave radiation
Surface downwelling longwave radiation
Surface upwelling longwave radiation
Daily maximum/minimum surface (2m) temperatures
Total cloud amount
Table 3: High-frequency (6-hourly) output
Total precipitation rate
Table 4: Time series of daily global averages (area-weighted)
Net downward energy flux at surface
Evaporation and sublimation
Total snow-covered area
Snow depth (water equivalent)
Average SST over open ocean
Table 5: Fixed geographic fields
Total soil moisture field capacity
Surface (upper 0.1 m) soil moisture field capacity
Surface latent heat flux
Surface (2 m) temperature
Betts, A.K., P. Viterbo, A. Beljaars, H-L. Pan, S-Y. Hong, M. Goulden, and S. Wofsy, 1997: Evaluation of the land-surface interaction in the ECMWF and NCEP/NCAR reanalysis models over grassland (FIFE) and Boreal Forest (BOREAS). Proceedings of the First International Conference on Reanalyses, 27-31 October 1997, Silver Spring, Maryland.
Chen, T.H., A. Henderson-Sellers, P.C.D. Milly, A.J. Pitman, A.C.M. Beljaars, J. Polcher, F. Abramopoulos, A. Boone, S. Chang, F. Chen, Y. Dai, C.E. Desborough, R.E. Dickinson, L. Dümenil, M. Ek, J.R. Garratt, N. Gedney, Y.M. Gusev, J. Kim, R. Koster, E.A. Kowalczyk, K. Laval, J. Lean, D. Lettenmaier, X. Liang, J.-F. Mahfouf, H.-T. Mengelkamp, K. Mitchell, O.N. Nasonova, J. Noilhan, A. Robock, C. Rosenzweig, J. Schaake, C.A. Schlosser, J.-P. Schulz, Y. Shao, A.B. Shmakin, D.L. Verseghy, P. Wetzel, E.F. Wood, Y. Xue, Z.-L. Yang, and Q. Zeng, 1997: Cabauw experimental results from the Project for Intercomparison of Land-surface Parameterizations Schemes. Journal of Climate, 10, 1194-1171.
Dickinson, R.E., A. Henderson-Sellers, and P.J. Kennedy, 1993: Biosphere-Atmosphere Transfer Scheme (BATS) Version 1e as coupled to the NCAR Community Climate Model. NCAR Technical Note NCAR/TN-383+STR, National Center for Atmospheric Research, Boulder, Colorado, 72 pp.
Gates, W.L., A. Henderson-Sellers, G.J. Boer, C.K. Folland, A. Kitoh, b.J. McAvaney, F. Semazzi, N. Smith, A.J. Weaver, and Q. Zeng, 1996: Climate Models-Evaluation. In Climate Change 1995: The Science of Climate Change. Contribution of Working Group 1 to the Second Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, pp. 233-284.
Henderson-Sellers, A., B. Henderson-Sellers, D. Pollard, J.M. Verner, and A.J. Pitman, 1995: Applying software engineering metrics to land surface parameterization schemes. Journal of Climate, 8, 1043-1059.
Irannejad, P., A. Henderson-Sellers, Y. Shao, and P.K. Love, 1995: Comparison of AMIP and PILPS off-line landsurface simulations. Proceedings of the First AMIP Scientific Conference, 15-19 May 1995, Monterey, California, WMO/TD-No. 732, pp. 465-470.
Love, P.K., and A. Henderson-Sellers, 1994: Land-surface climatologies of AMIP-PILPS models and identification of regions for future investigation (PILPS Phase 3a). GEWEX Report, IGBPO Publication Series No. 13, 48 pp.
Love, P.K., A. Henderson-Sellers, and P. Irannejad, 1995: AMIP Diagnostic Subproject 12 (PILPS Phase 3): Land-surface processes. Proceedings of the First AMIP Scientific Conference, 15-19 May 1995, Monterey, California, WCRP-92, WMO/TD-No. 732, 101-106.
Peylin, P., J. Polcher, G. Bonan, and D. Williamson, 1997: Comparison of two complex land surface schemes coupled to the National Center for Atmospheric Research general circulation model. Journal of Geophysical Research, 102, 19413-19431.
Phillips, T.J., 2001: On the predictability of seasonal land-surface climate. Program for Climate Model Diagnosis and Intercomparison (PCMDI) report No. 68. Accessible at http://www-pcmdi.llnl.gov/publications/pdf/rpt68.pdf .
Pitman, A.J., A. Henderson-Sellers, F. Abramopoulos, R. Avissar, G. Bonan, A. Boone, J.G. Cogley, R.E. Dickinson, M. Ek, D. Entekhabi, J. Famiglietti, J.R. Garratt, M. Frech, A. Hahmann, R. Koster, E. Kowalczyk, K. Laval, L. Lean, T.J. Lee, D. Lettenmaier, X. Liang, J-F. Mahfouf, L. Mahrt, C. Milly, K. Mitchell, N. de Noblet, J. Noilhan, H. Pan, R. Pielke, A. Robock, C. Rosenzweig, S.W. Running, C.A. Schlosser, R. Scott, M. Suarez, S. Thompson, D. Verseghy, P. Wetzel, E. Wood, Y. Xue, Z-L. Yang, , L. Zhang, 1993: Results from the Off-line Control Simulation Phase of the Project for Intercomparison of Landsurface Parameterisation Schemes (PILPS), GEWEX Tech. Note, IGPO Publ. Series, 7, 1993, 47pp.
Qu, W., and A. Henderson-Sellers, 1996: Surface energy balance of the revised AMIP simulations from BMRC and LMD GCM--A comparison of the effects of the bucket hydrology and the biophysically based land-surface scheme on the simulated climate. Unpublished manuscript.
Qu, W., A. Henderson-Sellers, A. Pitman, T-H. chen, F. Abramopoulos, A. Boone, S. Chang, F. Chen, Y. Dai, R.E. Dickinson, L. Dümenil, M. Ek, N. Gedney, Y.M. Gusev, J. Kim, R. Koster, E.A. Kowalczyk, J. Lean, D. Lettenmaier, X. Liang, J-F. Mahfouf, H-T. Mengelkamp, K. Mitchell, O.N. Nasonova, J. Noilhan, A. Robock, C. Rosenzweig, J. Schaake, C.A. Schlosser, J-P. Schulz, A.B. Shmakin, D.L. Verseghy, P. Wetzel, W.F. Wood, Z-L. Yang, and Q. Zeng, 1998: Sensitivity of latent heat flux from PILPS land-surface schemes to perturbations of surface air temperature. Journal of the Atmospheric Sciences (in press).
Sato, N., P.J. Sellers, D.A. Randall, E.K. Schneider, J. Shukla, J.L. Kinter II, Y-T. Hou, and E. Albertazzi, 1989: Effects of implementing the Simple Biosphere Model in a general circulation model. Journal of the Atmospheric Sciences, 46, 2757-2782.
Sellers, P.J., D.A. Randall, G.J. Collatz, J. Berry, C. Field, D.A. Dazlich, C. Zhang, and L. Bounoua, 1996: A revised land-surface parameterization (SiB2) for atmospheric GCMs. Part 1: Model formulation. Journal of Climate, 9, 676-705.
Shao, Y., and A. Henderson-Sellers, 1996: Validation of soil moisture simulation in landsurface parameterisation schemes with HAPEX data. Special issue on soil moisture simulation. Global and Planetary Change, 13(1), 11-46.
Wood, E.F. , D.P. Lettenmaier, X. Liang, D. Lohmann, A. Boone, S. Change, F. Chen, Y. Dai, R.E. Dickinson, Q. Duan, M. Ek, Y.M. Gusev, F. Habets, P. Irannejad, R. Koster, K.E. Mitchell, O.N. Nasonova, J. Noilhan, J. Schaake, A. Schlosser, Y. Shao, A.B. Shmakin, D. Verseghy, K. Warrach, P. Wetzel, Y. Xue, Z-L. Yang, and Q. Zeng, 1997: The Project for Intercomparison of Land-surface Parameterization Schemes (PILPS) Phase-2(c) Red-Arkansas River Basin Experiment: 1. Experiment description and summary intercomparisons. Global and Planetary Change (accepted).
Xie, P., and P.A. Arkin, 1997: Comparison between a global precipitation analysis based on in-situ and satellite observations and precipitation from reanalyses. Proceedings of the First International Conference on Reanalyses, 27-31 October 1997, Silver Spring, Maryland.
Yang, Z.L., A.J. Pitman,
B. McAvaney, and A. Henderson-Sellers, 1995: The impact of implementing
the bare essentials of surface transfer land surface scheme into the BMRC
GCM. Climate Dynamics, 11, 279-297.
For further information, contact Tom Phillips .