Intercomparison of US Land Surface Hydrologic Cycles from Multi-analyses & Models NOAA 30th Annual Climate Diagnostic & Prediction Workshop, 27 October, 2005, State College, PA Yun Fan & Huug van den Dool CPC/NCEP/NOAA
Outline Motivation Data Soil moisture annual cycle & long-term variability over Illinois Spatial & temporal correlations over CONUS Annual land surface hydrologic cycles CFS land surface predictability Summary
Motivation Soil Moisture (SM): one of key factors in environmental processes, such as meteorology, hydrology & et al. Accurate SM is important for Weather & climate prediction. Long-term large-scale in situ measurement not yet established Remote sensing – promising but immature Calculated SM: depends on quality of forcing & models Questions: Skills of soil moisture data sets Land surface hydrologic predictability of CFS Existing problems & possible reasons
8 Land Surface Datasets: 2.Three 50+ Year Retrospective Offline Runs 3.Three Reanalysis Datasets 1.Observations 18 Illinois soil moisture observation sites (1981- present) S.E. Hollinger & S.A. Isard, 1994 RR - North American Regional Reanalysis ( present) F. Mesinger et al, 2003, 2005 R1 – NCEP-NCAR Global Reanalysis I ( present) E. Kalnay et al, 1996 & R. Kistler et al 2001 R2 – NCEP-DOE Global Reanalysis II ( present) M. Kanamitsu et al, 2002 Noah - Noah LSM Retrospective N-LDAS Run ( ) – present Y. Fan, H, van del Dool, D. Lomann & K. Mitchell, 2003 VIC - VIC LSM Retrospective N-LDAS Run ( ) E. Maurer, A. Wood, J. Adam, D. Lettenmaier & B. Nijssen, 2002 LB - CPC Leaky Bucket Soil Moisture Dataset J. Huang, H. van den Dool & K. Georgakakos, 1996, Y. Fan & H. van den Dool, NCEP Climate Forecast System (CFS) Datasets S.Saha et al 2005
VICLBRRR2R1Obstemporal Noah VIC LB RR R R1 Temporal anomaly correlations averaged over Illinois 0.61 ERA40
dW(t)/dt: soil water storage change P(t): precipitation E(t): evaporation R(t): surface runoff G(t): subsurface runoff Res=P-E-R-G-dW/dt
Spatial & temporal anomaly correlations averaged over US spatialNoahVICLBRRR2R1temporal Noah VIC VIC LB LB RR RR R R2 R R1
Summary I. By overall mean annual cycle & interannual variability 1. Offline retrospective runs are generally better than reanalyses Noah VIC LB RR R2 R1 Good > poor 2. All other models (except Noah) either too dry and or too large annual cycle 3. Three reanalyses (RR > R2 > R1) shown steadily improvements II. RR has not reached its potential III. CFS (land surface soil moisture) 1. Good prediction skill (cr > 0.6, against to R2) for up to 5 months 2. Dry bias increase & delayed anomalies with lead time increase IV. Looking forward to R3