Testof enhancedsoiltypedetermination in Kai Sattler Danish Meteorological Institute ksa@dmi.dk H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
a medium soil type with medium soil water capabilities. see experiment showed an improvement of Tm by to Introduction&Motivation HIRLAM climate generation includes processing of type (Bringfelt et al. 99), and the data base it refers soil One reason for this is that the three classes are good repe- of the soil characteristics. sand () and clay sentatives represent two extremes, whereas loam () represents () medium. is taken from the FAO-UNESCO soil database (FAO- to 987). This soil data classifies soil types into UNESCO, categories: the limitations of the implemented soil type remapping However, became clear some years ago during spring time sand. loam. loam & clay 6. peat 7. the western part of Denmark, when HIRLAM's m temperature in forecasts where found to grow too much during clay. sand & loam. ice 8. rock 9. compared to the observed values. daytime significant contribution to this effect was found to come A. sand & clay. missing data the lack of soil water content, which in turn was found from be connected to the definition of the soil type in the low to ISBA surface scheme in HIRLAM, however, does make use of these categories. Instead the ISBA not sub-surface tile of the HIRLAM grid, e.g. soil vegetation initialization during a cold start depends strongly moisture code for the climate files (steered by script preparation includes a remapping of the FAO soil types Preps_ISBA) the underlying soil type. on large part of western Denmark (Jutland) was hereby A classes, which are a sub-set of the ISBA soil types to 996, table A.). three soili types are: (Bringfelt to be classified by soil type sand. Even though this found type occurs at coastal locations, it cannot be seen as a soil sand. loam. soil type over the area. FAO soil type data base major not supply these details. does clay. make up the sub-surface soil type soili used in y (Bringfelt 996, table A.). representation HIRLAM snow surface (SN) is hereby not used in recent HIRLAM of experiment was then performed, where the soil type An set to loam instead of sand, in order to make HIRLAM was versions. though the remapping of the soil types and reduction Even only three categories is a significant simplification, the to has worked reasonably well. proceduce locally in Jutland. A quick fix was then applied in degrees DMI-HIRLAM. operational H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
Refiningsoil typedetermination F A O s o il ty p e,7,6 8 9 T ile n o v e g e ta tio n the issue of soil type definition for HIRLAM ISBA Recently was taken up again, when extreme surface tem- scheme were observed at the coast of Greenland. Even peratures these problems had other causes and were not though v e g i = 8 ( d e s e r t) v e g i = ( ic e ) they revealed a code bug concerning the no vegetation related, sub-surface tile. As a consequence, the potential T ile lo w b e g e ta tio n v e g i = ( c r o p ) of soil type definitions in the other two subsurface insufficiencies tiles low vegetation and forest were discussed. v e g i = ( s h o r t g r a s s ) v e g i = 7 ( ta ll g r a s s ) v e g i = 9 ( tu n d r a ) v e g i = ( ir r ig a te d c r o p ) discussions took up the idea to refine the soil type se by utilizing the sub-surface type vegi in addi- determination to the FAO soil type not only under sub-surface tile no tion but also under low vegetation and forest (As vegetation, v e g i = ( s e m i- d e s e r t) v e g i = ( b o g /m a r s h ) v e g i = 6 ( e v e r g r e e n s h r u b ) A. in Bringfelt (996) shows, vegi has been utilized Table tile, but not in tiles and ). in v e g i = 7 ( d e c id u o u s s h r u b ) T ile fo re s t combination of using vegi and FAO soil type has potential to refine the ISBA soil type determination, the v e g i = ( e v e r g r e e n n e e d le tr e e ) v e g i = ( d e c id u o u s n e e d le tr e e ) v e g i = ( d e c id u o u s b r o a d le a f tr e e ) the vegetation type may give an indication on the because soil type. This way the crude information on soil underlying v e g i = 6 ( e v e r g r e e n b r o a d le a f tr e e ) v e g i = 8 ( m ix e d w o o d la n d ) in the FAO data can be enhanced by making use of type about the vegetation. information v e g i = 9 ( fo r e s t/fie ld s ) new mapping of FAO soil type and vegi into ISBA subsurface soil type is formulated in the list to the right, difference between the current and the new ISBA type determination in HIRLAM can be seen from the soil each of the three sub-surface tiles no vegetation (), low for () and forest (): vegetation pictures below. H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
current new H ir la m A ll S ta ff M e e tin g 8 th A la d in W o r k s h o p 7 - A p r il 8, B r u s s e ls
scores Overall impact of the refinement of ISBA soil type deter- Impact onsurfaceparameters Verification against observations EXP: RCR7 RCR7T Time: 7-78 Domain: EWG Forecast from soili was tested in two experiments. first mination made use of the soil types as in the current experiment system, and the second one used the newly HIRLAM soil types as described above. defined Both experiments were run over the following three periods:. 7 to 78. 7- to 7-- Mean and RMS error (hpa) Mean sea level pressure Mean and RMS error (K) -metre temperature. 7-6 to 7-7 first period is to represent winter conditions, and the period includes shifting weather condition, which second sunny high pressure situations over Europe. include period contains a summer period with many thunder- third storms. experiments are based on HIRLAM trunk version as of Both [67], still including the fix of changeset [69]. changeset experiments were configured to run on the RCR-7. and using DVAR analysis. Else default specifications domain, were used. standard verification was applied to the experiment HIRLAM runs. Overall impact of the soili refinement on surface MSLP, Tm, RHm and Wm is shown in the parameters figures (EWGLAM station list). following Mean and RMS error (m/s) - - - metre wind speed Mean and RMS error (%) - -metre relative humidity H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
Verification against observations EXP: RCR7 RCR7T Time: 7-7 Domain: EWG Forecast from Verification against observations EXP: RCR7 RCR7T Time: 76-77 Domain: EWG Forecast from Impact onsurfaceparameters Mean and RMS error (hpa) Mean sea level pressure Mean and RMS error (K) -metre temperature Mean and RMS error (hpa) Mean sea level pressure Mean and RMS error (K) -metre temperature Mean and RMS error (m/s) - - - metre wind speed Mean and RMS error (%) - -metre relative humidity Mean and RMS error (m/s) - - - metre wind speed Mean and RMS error (%) - -metre relative humidity the figures above show the impact of the refined determination As of soili on surface parameters is very small. these tendencies seem to be rather counter-productive, RHm, however. re is no visible impact on m wind definitions have slight tendency to higher m humidity, new in accordance to this, a slight tendency to lower m and For mean sea level pressure, a very small improvement speed. of the negative bias in the April period is visible (upper temerature. When considering the model bias for Tm and left figure of the left panel on this sheet). H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
series Time look at the time evolvement of the verification scores A Impact onsurfaceparameters the slight impact almost at all times. Respective confirms are shown for the m dew point to the right and figures below (bias and rms). further dew point parameter experiences the biggest influ- but it a very small one. ence, th still th is a short period around the 7 and 8 of is exception An 7, where the refined soili settings have a February postitive impact on the RMS of the dew point (see visible lower figure to the right. period between 6 th and th of February was characterized by a warming over 9 European continent, especially the eastern and some the the northern parts. following two maps of metre of illustrate this: temperature m te m p e ra tu re a n d m e a n s e a le v e l p re s s u r e a t 6 th a n d 9 th F e b r u a r y 7 H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
Impact onsurfaceparameters H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
distribution Spatial verification scores of the surface parameters MSLP, the curent soili settings in HIRLAM used on the left hand side of each panel respec- shown tively RCR7T. Impact onsurfaceparameters figures show that small local differences occur, usually with improvements in one region and deterioration at RHm and Wm compare very similarly between the Tm, experiments, when regarding the stationwise scores two another location at the same time. the two HIRLAM experiments. following figures from the distributions of bias (panels to the left) and rms show is interesting to notice, still, that differences also occur It the open sea. This may indicate that the differences over (panels to the right) valid for 8h forecast length. scores shown are: experiments surface parameter scores induced by the change in soili of over land tiles have propagated through the model fields RCR7. domain within the time period of the forecast range. the effect of changing soili does not remain a local So but it spreads over the model domain within the usual one, range. It can also be expected that the effect has forecast impact on the daily cycle in an operational setup of (small) the refined soili settings used on the right hand side of each panel respec- shown tively HIRLAM. Each row in a panel refers to one of the three periods. H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
Impact onsurfaceparameters(meansealevel pressure) H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
Impactonsurfaceparameters(mtemperature) H ir la m A ll S ta ff M e e tin g 8 th A la d in W o r k s h o p 7 - A p r il 8, B r u s s e ls
Impactonsurfaceparameters(mdewpointtemperature) H ir la m A ll S ta ff M e e tin g 8 th A la d in W o r k s h o p 7 - A p r il 8, B r u s s e ls
Impact onsurfaceparameters(mwindspeed) H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
SummaryandConclusions work tries to investigate the effect of changing the This soil type fields (soili) in HIRLAM on the verification ISBA show that refining the soili determination in Results has a small effect on the vefification scores. HIRLAM of surface parameters mean sea level pressure, m scores m humidity and m wind speed. temperature, effect consists mainly in the tendency towards slightly This humidity and slightly decreased temperature increased new soili fields are based on a refined determination the soil type. This refinement takes both the FAO soil of to the surface. close cases where the temperature bias of HIRLAM is nega- In or where humidity bias is positive, the refinement of tive, has a rather negative influence. soili as well as the sub-surface type vegi into account when type one of the three ISBA soil types for a certain land defining the same way, in case of positive temperature bias, or In humidity bias in HIRLAM, the soili refinement has negative of the HIRLAM grid. This methodology has been in use tile the sub-grid surface tile (no vegetation land) since the for slight positive effect. a spring case also indicates a slight improvement of scheme was introduced into HIRLAM, but it was never ISBA for the sub-grid surface tiles (low vegetation land) applied sea level pressure bias. mean differences in scores occur, as well as there are Local and (forest). differences in bias scores, so the effect of the soili local changes from region to region. An improvement refinement refined methodology was compared to the current in HIRLAM by running HIRLAM experiments for procedure verification scores for surface parameters due to soili of can therefore hardly be stated in general. refinement periods of several weeks duration each. periods three winter season, spring and summer conditions. represent single case within the first comparison period indicates, A an improved rms in dew point during some days however, comparison comprises standard verification scores for surface parameters mentioned above. the of the winter period. H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s
References B., Gustafsson N., Vilmusenaho P., Jarvenoja Bringfelt S. 99: Updating the HIRLAM physiography and climate data base. HIRLAM Tech. Rep., 9. Bringfelt B. 996: Tests of a new Land Surface Treatment in HIRLAM. HIRLAM Tech. Rep.,. 987: Soils of the World. Food and Agriculture Organization and United Nations Educational, FAO-UNESCO and Cultural Organization. Technical Report, Elsevier Science Publishing Co. Inc., New York. Scientific H i r l a m A l l S t a f f M e e t i n g 8 t h A l a d i n W o r k s h o p 7 - A p r i l 8, B r u s s e l s