Advanced Usage of the WRF Preprocessing System

Transcription

1 * Advanced Usage of the WRF Preprocessing System Michae Duda *NCAR is sponsored by the Nationa Science Foundation

2 Outine The GEOGRID.TBL fie What is the GEOGRID.TBL fie? Ingesting new static fieds Exampes: Using high-resoution and use and topography data The METGRID.TBL fie What is the METGRID.TBL fie? Exampe: Defining interpoation options for a new fied Exampe: Using the METGRID.TBL fie for a rea-time system Utiity programs exampe: fixing hot akes 2

3 The GEOGRID.TBL Fie GEOGRID.TBL is the fie that determines which fieds are interpoated by geogrid at runtime Each entry in GEOGRID.TBL corresponds to one data source When new data sources are invoved, or when the defaut treatment of fieds is inadequate, user may want/need to edit GEOGRID.TBL However, defaut GEOGRID.TBL is sufficient to initiaize a WRF simuation 3

4 The GEOGRID.TBL Fie Format of GEOGRID.TBL fie is simpe text, with specifications of the form keyword=vaue Exampe entry for a 30 anduse data set: =============================== name=landusef # Houston, TX urban data priority = 1 dest_type = categorica z_dim_name = and_cat interp_option = 30s:nearest_neighbor abs_path = 30s:/users/duda/Houston/ =============================== For a compete ist of possibe keywords See p

5 The GEOGRID.TBL Fie Using the GEOGRID.TBL, we can Change the method(s) used to interpoate a fied Appy smoothing fiters to continuous fieds Derive fieds from others E.g., dominant category or sope fieds Add new data for geogrid to interpoate 5

6 New Fieds in GEOGRID.TBL There are three basic types of new data to be added through the GEOGRID.TBL fie: 1) Competey new fieds - fieds that were previousy not processed by geogrid 2) Different resoution data sets for an existing fied - Such sources do not need to be suppemented by existing data - E.g., Adding a 90-meter resoution topography data set 3) Aternative sources for a fied that must be used in addition to an existing source - E.g., A new soi category data set exists, but covers ony South Korea 6

7 1) Competey new fieds Competey new fieds: For a new fied, simpy add an entry in GEOGRID.TBL for that fied. ============================ name = MY_NEW_FIELD_NAME priority = 1 dest_type = continuous interp_option = four_pt abs_path ============================ = /data/duda/mydata/ Name of fied that this entry is for Priority of this data source compared with other sources for same fied How to interpoate this fied Where on disk to find the data for this fied See p

8 2) Different resoution data set Different resoution data sets for an existing fied : Specify the path to the new data set and which interpoation methods shoud be used for the new resoution in the existing entry for that fied. =============================== name = HGT_M priority = 1 dest_type = continuous smooth_option = smth-desmth interp_option = 30s:specia(4.0)+four_pt interp_option = my_res:four_pt interp_option = defaut:four_pt re_path= 30s:topo_30s/ re_path= my_res:new_topo_directory/ re_path= defaut:topo_2m/ =============================== 8

9 3) Aternative data sources Aternative sources for a fied that must be used in addition to an existing source : Add a new entry for the fied that has the same name as the fied s existing entry, but make priority of new entry higher. =============================== name = HGT_M priority = 2 dest_type = continuous interp_option = defaut:four_pt re_path = defaut:some_path/ =============================== name = HGT_M priority = 1 dest_type = continuous interp_option = defaut:four_pt re_path = defaut:topo_2m/ =============================== 9

10 Preparing new geogrid data sets To add a new data source, we need to 1) Write the data in the proper binary format - See Chapter 3: Writing Static Data to the Geogrid Binary Format - Can make use of read_geogrid.c and write_geogrid.c 2) Create an index metadata fie for the data set - This tes geogrid about the projection, coverage, resoution, type, and storage representation of the data set 3) Add/edit entry for the data in the GEOGRID.TBL fie - The change to GEOGRID.TBL wi foow one of the three cases mentioned before 10

11 The Geogrid Data Format The geogrid format is a simpe binary raster Eements of a rectanguar array of data are written, row by row, to a fie No record markers or any type of metadata are written to this fie See p A fie containing a N M array, with each eement represented using K bytes, shoud have size exacty N*M*K bytes! 11

12 The Geogrid Data Format Since the contents of the fie contain ony the vaues from the array, care must be taken if using Fortran to write the array - Fortran unformatted writes add record markers to the beginning and end of each record - So, rather than X 1 X 2 X 3 X n-1 X n we get RX 1 X 2 X 3 X n-1 X n R, where R is a record marker Instead of Fortran, the C routines read_geogrid.c and write_geogrid.c may be used to read and write binary fies - these may be caed from either Fortran or C 12

13 The Geogrid Data Format The fienames of geogrid binary fies shoud have the form: xxxxx-xxxxx.yyyyy-yyyyy where xxxxx XXXXX yyyyy YYYYY is the starting x-index is the ending x-index is the starting y-index is the ending y-index E.g., For a binary fie containing an array with 500 coumns and 750 rows, the fie name woud be

14 The Geogrid Data Format If the data are not avaiabe in a singe tie (array), mutipe fies may be used to store the data A ties must have the same x-dimension A ties must have the same y-dimension If necessary, a tie can be padded with missing vaues to expand it to the same size as other ties in the data set 14

15 The Geogrid Data Format If the data do not cover a rectanguar region, areas with no data are simpy fied with a missing vaue so that the overa data set is rectanguar The particuar missing vaue used in the data set is specified in the index metadata fie for the data set 15

16 Exampe: Houston LU Data Set Given dataset for new Houston urban and use categories Reguar at/on projection, 30 resoution; categories 31, 32 & 33 Urban areas (back) using USGS 24-category data set Area of Houston data tie in reation to mode domain white=missing data and bue=vaid data 16

17 Exampe: Houston LU Data Set To make use of the new data, we do the foowing: 1) Write the data to the binary format used by geogrid 2) Create an index fie for the data type=categorica category_min=31; category_max=33 projection=reguar_ dx= ; dy= known_x=1.0; known_y=1.0 known_at= known_on= wordsize=1 tie_x=157; tie_y=143; tie_z=1 missing_vaue = 0. units="category" description="3-category urban LU" Data set has categories 31 through arc second resoution Geographic ocation of data set Treat 0 as no data See p

18 Exampe: Houston LU Data Set 3) Define an entry for the data in GEOGRID.TBL =============================== Give this data source priority name=landusef priority = 2 over defaut data sources dest_type = categorica z_dim_name = and_cat interp_option = defaut:nearest_neighbor abs_path = defaut:/users/duda/houston/ =============================== How to interpoate this data source, and where to find it on disk 18

19 Exampe: Houston LU Data Set 4) Run geogrid.exe Any gridpoints covered by Houston data wi use it; otherwise defaut USGS data wi be used Urban areas (back) using USGS 24-category data set Augmented urban areas (red shades) using new LU data set 19

20 Exampe: South Korea Shutte Radar Topography Mission (SRTM) 3 arc second topography data DX=1km DX=333m We woud ike to use the SRTM data, especiay for domains 2 and 3. Foow steps for adding a new resoution for an existing data set (case 2) 130x x232 DX=111m 361x391 20

21 Exampe: Seou To use the SRTM topography data, we 1) Write data to geogrid binary format 2) Create an index fie for the data set 3) Modify the GEOGRID.TBL entries for HGT_M, HGT_U, and HGT_V =============================== name = HGT_M priority = 1 dest_type = continuous interp_option = 30s:specia(4.0)+four_pt interp_option = SRTM:four_pt re_path = 30s:topo_30s/ re_path = SRTM:SRTM/ =============================== 4) Specify that we shoud interpoate from SRTM in nameist by setting geog_data_res = 30s, SRTM+30s, SRTM+30s 21

22 Exampe: Seou Domain 3 (DX=111m) using defaut 30 USGS topography Domain 3 (DX=111m) using 3 SRTM topography 22

23 Another Exampe: Los Angees For Los Angees, we have a 30-meter resoution, 3 urban and use category data set Urban areas (back) using USGS 24-category data set Augmented urban areas (red shades) using new LU data set 23

24 Outine The GEOGRID.TBL fie What is the GEOGRID.TBL fie? Ingesting new static fieds Exampes: Using high-resoution and use and topography data The METGRID.TBL fie What is the METGRID.TBL fie? Exampe: Defining interpoation options for a new fied Exampe: Using the METGRID.TBL fie for a rea-time system Utiity programs exampe: fixing hot akes 24

25 The METGRID.TBL Fie The METGRID.TBL fie contros how meteoroogica fieds are interpoated Unike GEOGRID.TBL, METGRID.TBL does not determine which fieds wi be processed, ony how to process them if they are encountered Every fied in intermediate fies wi be interpoated If no entry in METGRID.TBL for a fied, a defaut interpoation scheme (nearest neighbor) wi be used It is possibe to specify in METGRID.TBL that a fied shoud be discarded 25

26 The METGRID.TBL Fie Suitabe entries in METGRID.TBL are provided for common fieds Thus, many users wi rarey need to edit METGRID.TBL When necessary, different interpoation methods (and other options) can be set in METGRID.TBL Interpoation options can depend on the source of a fied 26

27 The METGRID.TBL Fie Exampe METGRID.TBL entry (for soi moisture 0-10 cm ) ======================================== name=sm interp_option=sixteen_pt+four_pt+average_4pt masked=water interp_mask=landsea(0) fi_missing=1. fag_in_output=flag_sm ======================================== 27

28 Exampe: A new METGRID.TBL entry Suppose we have a 1000x1000 domain over Houston (dx=500 m) This is the same domain as in the urban and use exampe Meteoroogica data come from 1-degree GFS Note that we wi be interpoating 1-degree data onto a 500-m grid! We want to create an entry for a new soi moisture fied, SM

29 Exampe: A new METGRID.TBL entry Initiay, we run metgrid.exe and get the message: INFORM: Entry in METGRID.TBL not found for fied SM Defaut options wi be used for this fied! The resuting SM fied ooks very coarse We need to create a METGRID.TBL entry so metgrid wi know how to interpoate this fied! GFS puts -1E30 in water areas (LANDSEA=0) 29

30 Exampe: A new METGRID.TBL entry We add an initia entry in METGRID.TBL for SM000010: =========================== name = SM masked = water interp_mask = LANDSEA(0) interp_option = sixteen_pt + nearest_neighbor fi_missing = 0. =========================== Specify that the fied shoud not be interpoated to mode water points Specify that metgrid shoud not use points in source where LANDSEA fied equas 0 Fi mode points that don t receive an interpoated vaue (ike water) to 0 For a compete ist of possibe keywords See p

31 Exampe: A new METGRID.TBL entry Now, after running metgrid.exe again, the SM fied ooks ike 16-point Nearest Neighbor Interpoated SM fied (sixteen_pt + nearest_neighbor) Which interpoator was used at each mode grid point 31

32 Exampe: A new METGRID.TBL entry But, the interpoated fied sti ooks bad near the coastine Shoud be sufficient data to use 4-point interpoation in these areas Interpoated SM fied (sixteen_pt + nearest_neighbor) Mode grid points here shoud be adjacent to at east one vaid GFS point (though not nearest) 32

33 Exampe: A new METGRID.TBL entry Update the METGRID.TBL entry for SM =========================== name = SM masked = water interp_mask = LANDSEA(0) interp_option = sixteen_pt + four_pt + average_4pt fi_missing = 0. =========================== If 16-pt doesn t work, then try 4-pt before reverting to a 4-point average Note that 4-point average wi work anywhere nearest_neighbor woud (missing/masked vaues not counted in the average) 33

34 Exampe: A new METGRID.TBL entry The resuting fied, beow-eft: 16-point 16-point 4-point Nearest Neighbor 4-pt average Interpoated SM fied (sixteen_pt + four_pt + average_4pt) Which interpoator was used at each mode grid point 34

35 Exampe: A new METGRID.TBL entry By using wt_average_4pt instead of average_4pt: sixteen_pt + four_pt + average_4pt sixteen_pt + four_pt + wt_average_4pt 35

36 METGRID.TBL: Rea-time System Exampe Suppose we have a rea-time system that: Uses GFS for initia and boundary conditions When possibe (i.e., if the fies are avaiabe soon enough) uses soi moisture and soi temperature fieds from AGRMET In our system, it may occasionay happen that the AGRMET fies are not ready when we want to start our WRF run Because system is rea-time, we want to proceed using just the GFS and surface fieds! 36

37 METGRID.TBL: Rea-time System Exampe We aready know how to run ungrib on mutipe sources of data to get GFS:YYYY-MM-DD_HH and AGRMET:YYYY-MM-DD_HH intermediate fies, and specify fg_name = GFS, AGRMET, in the &metgrid nameist record to use both sources See p

38 METGRID.TBL: Rea-time System Exampe Without further changes, what happens if: Ony GFS data are avaiabe when we run metgrid Metgrid runs and warns that no AGRMET data fies were found: Processing _00 GFS AGRMET WARNING: Coudn't open fie AGRMET: _00 for input. Metgrid wi finish, but wi ony use GFS data! 38

39 METGRID.TBL: Rea-time System Exampe And the 0-10 cm soi moisture fied (SM000010) ooks ike: 39

40 METGRID.TBL: Rea-time System Exampe However, what happens if: Both GFS and AGRMET fies are avaiabe when we run metgrid? Our SM fied ooks ike: We get unreasonabe vaues with magnitude ~1E30 near and-water boundaries! 40

41 METGRID.TBL: Rea-time System Exampe Why are there bad vaues near coastines? What went wrong? In both Vtabe.GFS and Vtabe.AGRMET, the and-sea mask fied is named LANDSEA - In METGRID.TBL, our entry for SM says: ======================================== name=sm interp_option=sixteen_pt+four_pt+wt_average_4pt+search masked=water interp_mask=landsea(0) fi_missing=1. fag_in_output=flag_sm ======================================== 41

42 METGRID.TBL: Rea-time System Exampe ======================================== name=sm interp_option=sixteen_pt+four_pt+wt_average_4pt+search masked=water interp_mask=landsea(0) fi_missing=1. fag_in_output=flag_sm ======================================== After metgrid reads in LANDSEA from GFS fie to use as an interpoation mask, it ignored the LANDSEA fied from AGRMET for use as a mask. - So, metgrid used the GFS LANDSEA mask even when interpoating AGRMET data! 42

43 METGRID.TBL: Rea-time System Exampe When metgrid interpoated SM000010, it used the GFS andmask for a fied masked by the AGRMET andmask! GFS LANDSEA fied AGRMET LANDSEA fied Note the disagreement between the two data sources near coastines. 43

44 METGRID.TBL: Rea-time System Exampe Soution: - Rename LANDSEA to AGR_LAND in Vtabe.AGRMET - Rename LANDSEA to GFS_LAND in Vtabe.GFS - Create separate entries in METGRID.TBL one for GFS SM fied another for AGRMET SM fied 44

45 METGRID.TBL: Rea-time System Exampe ======================================== name=sm000010; from_input=gfs interp_option=sixteen_pt+four_pt+wt_average_4pt+search masked=water interp_mask=gfs_land(0) fi_missing=1. fag_in_output=flag_sm ======================================== ======================================== name=sm000010; from_input=agrmet interp_option=sixteen_pt+four_pt+wt_average_4pt+search masked=water interp_mask=agr_land(-1.e30) fi_missing=1. fag_in_output=flag_sm ======================================== 45

46 METGRID.TBL: Rea-time System Exampe With modified Vtabes and METGRID.TBL: The SM fied when ony GFS fies are avaiabe The SM fied when both GFS and AGRMET fies are avaiabe 46

47 Outine The GEOGRID.TBL fie What is the GEOGRID.TBL fie? Ingesting new static fieds Exampes: Using high-resoution and use and topography data The METGRID.TBL fie What is the METGRID.TBL fie? Exampe: Defining interpoation options for a new fied Exampe: Using the METGRID.TBL fie for a rea-time system Utiity programs exampe: fixing hot akes 47

48 Motivating Probem The Hot Lake probem: Inand water bodies that are not resoved by SST data sets often receive extrapoated vaues from nearby oceans or other resoved water bodies. Mývatn Lake Mývatn region Above eft: Skin temperature fied (TSK) for Iceand and surrounding ocean on 26 January UTC from NCEP GFS and RTG SST data. Above right: TSK in the Mývatn region. SST for Mývatn Lake is ~277 K! 48

49 Approach In WRF v3.3 and ater, et the rea preprocessor know which water points are inand water bodies, and provide it a more accurate estimate of SST to be used ony over these water bodies. 1) Identify inand water bodies in the and cover data set 2) Provide a suitabe proxy for SST fied over inand water bodies E.g., Average surface air temperature for X days prior, 273 K for frozen akes, etc. 3) Modify the SST fied in the WRF input fie Use new capabiity in v3.3 rea.exe program 49

50 Identifying Lakes Some data sets aready identify akes with separate categories MODIS, CORINE For others, we need a way to do this Shoud be automated don t want to spend ong hours cicking on pixes for each data set Shoud be tunabe what constitutes a ake wi naturay depend on what our SST data set is abe to resove Ideay, woud not require auxiiary data In nameist.wps, set: geog_data_res = usgs_akes+30s for USGS and use (16=ocean, 28=ake) geog_data_res = modis_akes+30s for MODIS and use (17=ocean, 21=ake) 50

51 Creating a Proxy SST Fied The avg_tsfc.exe utiity program may be used to compute the average 2-m air temperature fied for any number of fu diurna cyces Number of cyces determined by avaiabe intermediate fies and date range in nameist The resuting TAVGSFC intermediate fie may be provided to the metgrid program 51

52 Test case: Lake Mývatn To confirm that everything is working as expected, try correcting the temperature for Lake Mývatn in the winter Grid ID Resoutio n Size 1 16 km 99x km 208x km 136x m 160x160 Ics + BCs from NCEP GFS Sea surface temperatures from RTG SST Initia time: 26 January 2011, 1200 UTC 52

53 Test case: Lake Mývatn 26 January 2011, 12 UTC ~277.5 K ~273.8 K Initia skin temperature fied using extrapoated SST vaues Initia skin temperature fied using previous 5-day average of 2-m air temperature for ake SST 53

54 Test case: Lake Mývatn Time series of sensibe heat fux in the center of the ake show a significant decrease when using a more reaistic SST (TAVGSFC) Latent heat fux time series from simuation using TAVGSFC for SST aso shows a decrease from RTG SST time series as we 54

55 Summary In this ecture, we ve seen What the GEOGRID.TBL and METGRID.TBL fies do How to use new geographica data sources in the WPS High-resoution and use and topography data How to use the METGRID.TBL fie to correct two types of interpoation-reated probems How utiity programs can be used to improve simuations For other features of the WPS, see Chapter 3 of the User s Guide For more information about using high-resoution topography data or urban and use data (over the U.S.), see how_to_hires.htm 55

56 Questions? 56