# The climatological variance of albedo, soil moisture in weather models

Looking at WRF currently but I am sure this applies to most weather models there is a file (specifically LANDUSE.TBL) that specifies USGS derived albedo, soil moisture and other parameters on a biannual basis - one for summer and one for winter.

Why would not these parameters vary on a more seasonal basis and/or on a annual basis? Why are these these variables assumed to be constant? Would not changing vegetation or land cover modifications affect these values on a regional basis as well as their impact on regional climate?

Also would it make sense to model these values on a regional basis after deriving them from remote sensing as indicated in the below mentioned article http://web.maths.unsw.edu.au/~jasone/publications/evansetal2012a.pdf

• Sometimes, uncertainties are so large that the most general assumption is to assume something to remain constant. It might be what you get if you average many models that disagree widely about amplitude and phase of the seasonal cycle. I can't say if that is what's going on here, though. – gerrit Feb 9 '15 at 16:53
• @gerrit - how about snow cover over large areas i.e. regional scales ? Would that not vary seasonally ? – gansub Feb 10 '15 at 1:39
• It certainly should. I really don't know enough to answer your question, I hope others do. – gerrit Feb 10 '15 at 4:20
• Most of those values (except surface roughness) will get overwritten with the VEGPARM.TBL values if you use Noah or the RUC LSM. Noah uses monthly values as far as I know. RUC might write values on its own... not really sure if it uses the VEGPARM.TBL or not. There are many groups that use the monthly MODIS-derived values. – f.thorpe Feb 11 '15 at 22:43
• @gansub model development is done such that you can run the model independently before you start adding more complex options to ingest observations. A "summer" and "winter" lookup table is provided for users, but I don't think anyone in the WRF community would argue those defaults are appropriate for the best retrospective weather simulations. Forecasting can use old satellite data (MODIS land-surface data is several days if not weeks old by the time it is processed) but that has error just as a seasonal average would. – f.thorpe Feb 12 '15 at 17:31

WRF model development is done in such a way that users can run the model independently before you start adding more complex options to ingest observations. There is even an "ideal" mode that new users can take advantage of to learn how the system works (not for simulating real Earth situations). In "real" mode, there are typically two types of simulations done: a forecast or a retrospective (historical) simulation. It's important to have real observations ingested into a retrospective simulation, since the intent is to model the event as best you can. However, this is not possible in forecast mode, since the observations do not yet exist. So, it is important to have default lookup tables that characterize your unknowns.

Regarding the values in the LANDUSE.tbl, there are many groups that use the monthly MODIS-derived values to ingest into their simulations that replace some of these values. A "summer" and "winter" lookup table is provided for users in WRF, but those defaults are not appropriate for the best retrospective weather simulations. Forecasting, on the other hand, can use old satellite data (MODIS land-surface data is several days if not weeks old by the time it is processed) but that has error just as a seasonal average would. The values in the LANDUSE.tbl (except surface roughness) will get overwritten with the VEGPARM.TBL values if you use Noah or the RUC Land Surface Model (LSM). Noah uses monthly values as far as I know. RUC might write values on its own... not really sure if it uses the VEGPARM.TBL or not. I'm sure there are experienced WRF modelers out there that can give you more detailed information, e.g. on the WRF modelers forum.

For soil mosture you can initialize the model with a dataset that provides that information. If you are doing historical retrospective you are likely going to initialize from a reanalysis dataset and if you are doing a forecasting case then you are likely initializing from GFS, RAP, HRRR or some other model.

For the case of initialization from GFS, the grib files you can get from NCEP have soil temperature and moisture for 4 levels 0-0.1 m, 0.1-0.4 m, 0.4-1 m and 1-2m below ground. The North American Regional Reanalysis (NARR) data a single layer for subsurface soil moisture and temperature available every 3 hours and from daily or monthly means. The ECMWF reanalysis (ERA-interim) has 4 layers of soil temperature and mositure and a surface roughness.