Toward improving the representation of the water cycle at High Northern Latitudes

Transcription

1 Toward improving the representation of the water cycle at High Northern Latitudes W. A. Lahoz a, T. M. Svendby a, A. Griesfeller a, J. Kristiansen b a NILU, Kjeller, Norway b Met Norway, Oslo, Norway wal@nilu.no

2 Outline Introduction Background: rapid warming at Northern Latitudes Difficulties of measuring soil moisture at Northern High Latitudes Data sources for Northern High Latitudes: satellite/in situ Data in situ/satellite (different characteristics) Comparison in situ and satellite soil moisture Norway Time-series: Extreme events: droughts/floods Summary and future work (e.g., data assimilation) Example of early results from data assimilation

3 Introduction Rapid warming Northern Latitude regions in recent decades -> lengthening of growing season, greater photosynthetic activity & enhanced carbon sequestration by ecosystem (Barichivich et al & references therein) Changes likely to intensify summer droughts, tree mortality & wildfires Potential major climate change feedback is release of carbon-bearing compounds from soil thawing - e.g., methane, carbon dioxide (Woods Hole Research Center, Holmes et al., 2015, Policy Brief)

4 Example of changes at northern high latitudes: PDSI = Palmer Drought Severity Index (sc self-calibrating) PET = potential evapotranspiration: red/blue (with/without interannual PET changes) = Mean summer T anomaly wrt Proxy regional summer droughts

5 Problems with satellite data at N. high latitudes AMSR-E SMOS Surface soil moisture - SSM Low corr AMSR-E product SMOS L3 product Wrt to reference: DAS2 ECMWF SSM from Land DA (ASCAT SM + 2m T/RH) 03/ /2011 High RMSD +ve/-ve High bias Al Yaari et al. 2014

6 Important to have information on land surface (soil moisture & temperature) at high northern latitude regions Availability of soil moisture measurements from satellites -> opportunity to address issues associated with climate change: assessing multi-decadal links between increasing temperatures, snow cover, soil moisture variability & vegetation dynamics (Barichivich et al & references therein) Relatively poor information on water cycle parameters for biomes at northern high latitudes (boreal humid Scandinavia) make it important efforts made on improving representation of water cycle at these latitudes (Al Yaari et al. 2014; van der Schalie 2015)

7 Why soil moisture retrieval difficult over Norway SMOS, soil moisture (m 3 /m 3 ) forest; rocks; moderate topography; strong topography; wet snow; open water fraction > 10%

8 Data: in situ/satellite In situ data network for Norway NVE (Norway Water Authority)) Wetland fraction (ASCAT data) describes the coverage of inundated and wetland areas from 0 to 100 % Topographic index (ASCAT data) is derived from GTOPO30 data; the standard deviation of the elevation is calculated and the result is normalized to values between 0 and 100 % NVE has total of 18 stations

9 Time-series of satellites (1978- present) ESA CCI soil moisture website:

10 Satellite-derived soil moisture ASCAT (Advanced SCATterometer): Launched October 2006 Active scatterometer C-Band, GHz : 1 cm depth Descending orbit 9:30, ascending orbit 21:30 SMOS (Soil Moisture and Ocean Salinity): Launched November 2009 Passive radiometer L-Band, 1.4 GHz : 5 cm depth Ascending orbit 06:00, descending orbit 18:00 Co-location criterion in situ/satellite: Make comparison if in situ data falls within lat-long grid associated with orbit swath: 0.3 o x0.3 o SMOS 0.1 o x0.1 o ASCAT (latitude dependent)

11 NVE in situ data Data provided by NVE (unofficial product) Six NVE stations: Ås, Kise, Øverbygd, Særheim, Værnes and Kvithamar. For Værnes, soil moisture not measured in 2013 and For Kvithamar no data exist in 2009 and after 2012 SMOS: no satellite data exist in grids covering Ås, Øverbygd & Særheim. For Særheim, satellite overpass data from a neighbouring grid used, but these data less representative ASCAT: no overpass data retrieved for Kvithamar Griesfeller et al. (2015) considered period for AMSRE-/NVE comparison

12 Satellite data used ASCAT: Product version WARP 5.5, Release 2.2. Data downloaded from H-SAF FTP server (ftp://ftphsaf.meteoam.it), H-SAF website SMOS: Soil Moisture (SM) products between and (tagged RE02). Operational data after (tagged OPER). Downloaded from Note: SMOS data version not uniform , focus on Data from 2014 are not reprocessed Statistics based on filtering & normalization method used for analysis (Griesfeller et al., 2015)

13 Data preparation ASCAT soil moisture data provided as degree of saturation from 0 % (dry) to 100 % (saturated); for comparison with volumetric in situ data convert the ASCAT data to volumetric soil moisture values (m 3 m -3 ) Exponential filter first described by Wagner et al. (1999); this filter is used to estimate root-zone soil moisture (specified as SWI, m 3 m 3 ) - Can use land data assimilation to estimate root zone soil moisture ASCAT and SMOS data are normalized using the mean and standard deviation of in situ data (Brocca et al. 2013) other approaches possible (e.g., CDF-matching)

14 Comparison Results: Figures/1 Ås Øverbygd Year

15 Figures/2 Værnes Kvithamar Year

16 Figures/3 Kise Year 2013

17 Comparison: Summary Tables SMOS Asc «better» ASCAT Des «better»

18 Extreme events for Norway Floods see image: May 2013 Southern & Eastern Norway (Kise & Ås NVE stations) Sep-Oct 2014 Western Norway (No NVE stations) May 2015 Eastern Norway (Kise NVE station) Droughts (Southern & Eastern Norway) see maps: Jan-Feb 2014 extremely light precipitation Mar 2014 positive temperature anomaly Kise: only 1 st half Jan/2 nd half Feb Øverbygd: March Ås:Throughout

19 24 May 2013 Eastern Norway Source: Dagbladet

20 Extreme weather map Jan-Feb Global temperature anomaly 4 th Mar 2014 ClimateReanalyzer.org

21 Results extreme events Figures NVE ASCAT & SMOS Year

22 Difference in spatial scales: in situ/satellite Flood 23 May 2013; Star: Ås; o N o E

23 Summary Norway one of most challenging regions for measurements of soil moisture First time soil moisture data from satellite platforms evaluated over Norway - extends the work in Griesfeller et al., 2015 Averaged correlations of satellite/in situ data over Norway relatively high for some stations (particularly for ASCAT) this result is not expected Capture of extreme events over Norway (in situ/satellite data) Satellite soil moisture information over Norway is useful! SMOS: ascending generally better than descending orbit ASCAT: descending generally better than ascending orbit ASCAT generally performs better than SMOS

24 Future work Extend collaboration with NVE improved data use, error characterization Land data assimilation of soil moisture, with focus on flood/drought events: (i) benefit for forecasts of hydrological cycle (NILU); and (ii) benefit for NWP forecasts (Met Norway) preliminary land data assimilation of ASCAT, SMOS and AMSR-E data for 1-month in 2011 (July) Land DA statistics (e.g., SMOS): consistent with 0.04 m 3 m -3 error for SMOS Example (EnKF): test vs independent data over Europe e.g., Urgons, France SMOSMANIA Replicate for Norway

25 Extra slides

26 Tables/1

27 Tables/2

28 Data assimilation study Jul 2011

29 Satellite data ASCAT converted to m 3 m -3 using % -> (0,1) Assumes max/min values are 100%, 0% (approx.) Satellite information unscaled, July top: mean; bottom: std SMOS drier ASCAT more variable

30 Model data Model SURFEX (Le Moigne 2012): July top: mean; bottom: std Scale satellite data to model data account for bias & variability: Linear re-scaling (Brocca et al., 2013): SAT : satellite; OBS : model Satellite data: same mean & std as model over July 2011 Focus on satellite anomalies Look at selected days & time series N.B. CDF-matching inappropriate length of time series is too short future work

31 Scaling data Satellite information 5 July top: unscaled; bottom: scaled Scaling: dries AMSR-E, ASCAT, moistens SMOS White areas: either no data, or data off-scale

32 Data assimilation Combine obs & model information + errors Lahoz & De Lannoy, Surv. Geophys., 2014 Focus on July 2011 European domain short period so care with stats EnKF (variants) use ensemble square root EnKF (Sakov and Oke, 2008) Model spin-up (1 month) Model forcing from WRF (NCAR FNL data) check representation of precipitation Five ensemble members (can choose other sizes) Perturbation of superficial & mean volumetric water content - precipitation forcing available but not used; mean of ensemble = 0 Scale observations to model (linear re-scaling; other options) Test observational errors (chi-square approach) Test system using self-consistency (O-F vs O-A differences) Test results against independent data (ISMN in situ data) also ESA CCI data Land DA results are preliminary & illustrative

33 Tests Observations: self-consistency tests; evaluation of errors Chi square approach applied to corrected satellite data N = no. of obs (July); F = forecast; A = analysis: Chi-sq(A) = (1/N) * SUM[(O-A)^2 / R] Chi-sq(F) = (1/N) * SUM[(O-F)^2 / R] 1. O-A differences should be smaller than O-F differences self-consistency test; passed 2. Chi-sq values should be close to 1 observational error information SMOS (N=547431) YERROBS=0.1 - Chi-sq(A) = 8.88 YERROBS=0.1 - Chi-sq(F) =64.45 YERROBS=0.3 - Chi-sq(A) = 2.86 YERROBS=0.3 - Chi-sq(F) = 6.79 YERROBS=0.6 - Chi-sq(A) = 1.11 YERROBS=0.6 - Chi-sq(F) = 1.69 AMSR-E (N=949842) YERROBS=0.3 - Chi-sq(A) = 2.71 YERROBS=0.3 - Chi-sq(F) = 6.38 ASCAT (N= ) YERROBS=0.3 - Chi-sq(A) = 2.72 YERROBS=0.3 - Chi-sq(F) = 6.45 SURFEX code - observational error defined as R = (YERROBS*COFSWI)^2 YERROBS, parameter set in input file: typically use 0.3 COFSWI=(W fc -W wilt ) typical range Error associated with SMOS anomalies is in range m 3 m -3 when YERROBS=0.6 Consistent with a SMOS error of 0.04 m 3 m -3 Kerr et al., 2010

34 July, SMOS, Differences: analyses model Left: unscaled SMOS; Right: scaled SMOS Regions of larger impact in unscaled version replicated in scaled version - e.g., France/Germany/England

35 Test v independent data Time series: analyses vs ISMN data July 2011 Thanks Morgan Kjølerbakken SMOSMANIA - Urgons 43.54N, 0.43W 145 masl SMOSMANIA - France