Water holding capacities in the water satisfaction index

[tamuka]

I am currently working on the Water Satisfaction Index (WSI), and have noted that default soil water holding capacity inputs for agrometshell are around 50mm. Would anyone happen to know how these water holding capacities were derived? I would greatly appreciate any insights from you on this matter.

I have been comparing the agrometshell-based soil water holding capacities with those from the FEWS NET version of the WRSI. The FEWS NET version features WHC values typically between 125mm and 200 mm (per metre of soil depth) in most. As far as I understand, they were derived from pedo-transfer functions which were applied on the FAO Soil map of the world.

The difference between the agrometshell and the usgs whc values are quite high, and I am trying to understand what may be causing this.

I would also appreciate to know more information on the shortcomings of the WSI, including the use of crop coefficients.

[rene]

Hello Tamuka

thanks for the questions! As I have started answering this elesewhere, I hope you won't mind my repeating some of the text here... after which I'll provide some additional views.

a. I have never believed (contrary to others) that the FAO WSI is a "model" in the physical sense. It is far too primitive for that. It is an "impressionistic model". I don't enter into details but everything is primitive, starting with ET always proceeding at the potential level, even when there is little water left. Then the concept of crop coefficient is very shaky too. Plus a couple of other things. What the WSI is very good at is converting rainfall amount AND distribution into <b>one</b> index.

b. WHC is, therefore, not the real physical WHC, but some "effective WHC", just like rainfall is "effective rainfall". My approach has always been that I define a value of WHC that is an operational value, i.e. the value that gives me the best result. This is done by comparing yield against WSI, and using WHC as a tuning variable (call it "sensitivity analysis"if you like the jargon. The best WHC is the one that gives you the best correlation. Second point: if you use a WHC that's too high, you'll never catch a water stress.

c. Crop Coefficient is a very primitive concept too, because it assumes plants are "stupid" in their constant pre-programmed phenology. Basically: they don't adapt. We have all seen millet plants that are 30 cm high, and flowering, because their are daylength sensitive (photosensitive). Sorghum, for instance, is much more "intelligent" than maize: in case of water stress, it goes semi-dormant and "waits" for the hard times to be over. After rainfall returns, so does the growth of sorghum. Clearly, the Kcs cannot mimick photosenstitivity or stress-induced dormancy: they assume constant phenology!

d. There are a number of other subtle and arbitrary ways of computer a water balance, even a simple one like the FAO WSI. Take this example. We aSSUME WE ARE TALKING OF DEKADS. Here are two ways of computing the water balance. Just in case, the variables are the following:

SW_beg: soil water at the beginning of the dekad mm
SW_end: soil water at the end of the dekad mm
Rain: precipitation mm
Crop_watreq: mm
WHC: soil water holding capacity: mm

Method 1

SW_end= SW_beg + Rain - Crop_watreq
if SW_end > WHC --> SW_end = WHC
if SW_end < 0 --> SW_end = 0

Method 2

SW= SW_beg + Rain
if SW > WHC --> SW = WHC
SW = SW - Crop_watreq
if SW < 0 --> SW_end = 0 else SW_end=SW

Method 1 assumes that all rain is well distruibuted over the dekad. Method 2 assumes that all rainfall comes down in one shower at the beginning of the dekad. You can invent other approches still, assuming that 50% of the train comes down on day 1, and 50% on day 5. Or, you will rightly assume that it is better to use daily data - which is definitely right.

I have actually experimented with the two contrasting methods above. In some semi-arid places, where rainfall is scarse but abundant for each shower, Methods 2 can yield better results...

Finally, my experience with the FAO WSI, is that it can be improved CONSIDERABLY by using a realistic inintial water storage value. I would even say that this is the single most critical variable! Some people use 0 mm (we are at the end of the dry season and there is no soil moisture storage) and others prefer WHC (farmers don't plant if there hasn't been at least one big shower) and others still use a conventinal fraction of WHC, e.g. WHC/2. I have noticed that, if I start the WB calculations 5 dekads before actual planting (as in the latest versions of WABAL I wrote) I will have a significant improvement of the correlation between yield and the WB variables, including average soil moisture.

I would add thet the FAO impressionistic "model" performs well at the regional scale BECAUSE it is simple. It does not require any inputs that are meaningless at the regional scale. All you have to do is to admit all inputs are "effective" variables.

Hope this helps!

R.

[lmjrw]

hello Mr Rene,
i am also working in this subject area. of-course your description is valuable for me too. thankx

Janaka from sri lanka

[rene]

Thanks, Janaka! If you have any specific question about the WSI, I' probably have some ideas.
Regards
R

[rahimihaji]

Hi Rene,

do you have any reference or handbook(manual), about WSI? what about software?

Thanks. Mohammad

[rene]

Dear Mohammad

there are many software(s) [*] around but most of them were developed in an institutional context, and they are probably not freely available. Also remember, as I mentioned above, that most WSIs computed by FEWS or JRC or others are actually "customizations", i.e. there is no longer only one standard WSI. Therefore, if you manage to put your hands on a software, make sure you also have a description of the algorithm that was implemented. Also, WSI is now usually computed on a grid, based on gridded inputs with resulting grid output. Look up this sites for two US implementations at IRI and FEWS: http://iridl.ldeo.columbia.edu/document ... readme.pdf and http://chg.geog.ucsb.edu/tools/geowrsi/index.html. For a European counterpart, try http://www.gsdidocs.org/gsdiconf/GSDI-9 ... 1paper.pdf at JRC. The JRC GWSI software For an example from Wageningen Agricultural University, look up https://www.researchgate.net/publicatio ... te_sensing. If you search the net for GWSI & JRC, you'll find more documents.

At FAO, you will find AgroMetShell (AMS), which implements the WSI on a station or grid basis. Unfortunately, the software is a bit old, but still perfectly functional! It is, as far as I know, one of the few software(s) that is (are) actually available to the public. It has been used in a number of national and international studies, publications, theses etc. A detailed manual is available.

http://www.fao.org/fileadmin/user_uploa ... Armeia.pdf
http://www.fao.org/nr/climpag/aw_6_en.asp
A lot of information and downloads are available from http://www.hoefsloot.com/agrometshell.htm

I have written a simple "bare bones" programme of the WSI for incorporation into a bigger exercise called MOSAICC (see http://edepot.wur.nl/218153). The programme is called "WB" and comes in three "flavours" WB1, WB2 and WB3 that differ by the way data are passed to it. The programme is available for windows and linux.

Here's my suggestion: drop a line to FEWS (e.g. Tamuka, who started this discussion) and to JRC, ans see if they have anything usable (i.e. non-web based) they can share. Also try AMS. If desperate, I can share WB1, WB2 and WB3 with you.

Best regards!

R

[*] "Software" is supposed to be singular, but then how should one say "many software(s)". Maybe the right way would be to say "there is much software around."

[tamuka]

Hello Rene,

Happy new year to you.

Thank you very much for your detailed response to my question, it was actually very useful and informative. One of your comments caught my attention. You mentioned:

You mentioned that

Crop Coefficient is a very primitive concept too, because it assumes plants are "stupid" in their constant pre-programmed phenology.

Is maize one of the more "stupid" plants that are more accurately modelled by WSI and Kc, or does it also do some clever adjustments that are not captured by Kc? If maize is well modelled by Kc, do you think maize is the only crop that is well modelled by WSI in this context?

Your comments on the assumption of whether rainfall fell at the beginning or the end of the dekad are very interesting. I will try running the model at a pentadal or daily time step and see if I get significant differences in southern Africa.

Thanks again, and regards,
Tamuka

[rene]

Hi Tamuka!

There are, of course, several ways for plants to be "stupid." Dailength dependence can be one form, with plants (e.g. bulrush millet) flowering regardless of their size, as I have mentioned. I would suggest that maize is indeed one of the more "stupid" common cereals and, as you suggest, this is why it can be more accurately modelled by the simple WSI (I almost wrote "stupid WSI!").

About which plants are well modelled by WSI, the answer also depends on the variant of WSI you use. The latest FAO WSI which allows for surface water storage behind bunds performs well if combined with actual PET (as a proxy for radiation) for rice. For the rest, I am a poor agronomist in terms of breadth of experience with different crops beyond the main cereals. My guess would be that crops with indeterminate phenology (cassava, cotton) would not work unless you are under very dry conditions. Also remember that the linearity between relative yield and relative ET only holds for moderate water stress probably not exceeding 70%. Below that, I guess all plants become "stupid".

From a modelling point of view, it is not very easy to make Kc phenology dependent...

By the way, I just discovered that there is a new version of the standard FAO publication on "Crop yield response to water". It's available here http://www.fao.org/docrep/016/i2800e/i2800e00.htm. Unfortunately, the document is no longer about "Crop yield response to water" as such but focuses on AquaCrop. One of the chapters (http://www.fao.org/docrep/016/i2800e/i2800e02.pdf) has the traditional stuff about Ky. It would be very interesting to work out a relation between Ky and Kc!

All the best tou you and family!
R

[tamuka]

Hi Rene,

Thanks again for the detailed response, and the research tips/hints! Much appreciated.

Regards,
Tamuka