working with the ISRIC 250 m African soil grids

I'm preparing to run the InVEST sediment and water yield models in Rwanda and want to use the new 250 m ISRIC African soils data ( for the needed PAWC and soil erodibility inputs.

Problem is, the inputs are available at multiple depths (2.5, 10, 22.5, 45, 80, 150) and AWC is available at different field capacities (pF 2.0, 2.3, 2.5). Which depths would you recommend using for developing the k factor layer for soil erodiblity and which depths and field capacities would you recommend for calculating the PAWC layer?

Since there are no redistribution restrictions on the data I'm happy to share any derived products for others' use elsewhere in Africa.


    It would be interesting to hear if anyone has thoughts on the choice of different field capacities, and which may be optimal in different parts of Africa.

    When calculating PAWC layers for the large Rufiji Basin in Tanzania, we calculated AWC at pF 2.5.
     Soil suction pF = log10 (cm negative head).
    pF 2.5 corresponds to -33 kPa as a standard definition of soil matric potential at field capacity (Kirkham 2014, Principles of Soil and Plant Water Relations). 

    It can be argued that a different FC value would be better, depending on prevalence of dryland soils for example.
    However, at the spatial scale of a country or large watershed, there may be extensive variation in FC, so perhaps the use of a standard value is wisest?

    Kirkham 2014: 
    "One should always try to measure FC in the field for each soil. The matric potential associated with FC can be as high as 0.0005 MPa in a highly stratified soil or as low as 0.06 MPa in a deep, dryland soil (Baver et al., 1972, p. 382). If one cannot measure FC in the field, it is often estimated to be the soil water content at a soil matric potential of 0.033 MPa
    or 33 kPA (one-third bar)."

    AWC note on AfSIS data:
    Note that in GIS processing, we converted the values for water bodies from NoData to AWC=1.0. 


    We calculated the weighted average for the soil layers down to 1 m depth, for both k factor and PAWC:
    The choice of the top 1 m of soil was made for consistency with the common units for hydrological models (e.g. AWC in mm/m), and the presence of most roots in the top 1 m of soil. No soils within the study area were estimated to have depths of <90 cm (depth to restricting layer). 

    Depth correlations and spatial patterns: 
    AWC for the top 1m was strongly correlated with AWC for deeper soil layers, where they exist, and the estimates for the top 1 m capture well the patterns in spatial variability that would be observed when calculating AWC across the full soil profile. (AWC would generally be smaller per m when including the greater depths, because they are closely correlated, but have lower mean AWC.)
  • PerrinePerrine Moderator, NatCap Staff
    Hi both,

    Erodibility: I recommend using the *surface* layer only. This is the soil layer that will be eroded, so erodibility should be computed for this layer only.

    AWC: I recommend using the weighed-average of AWC for all soil layers (weights are the depths of each soil layer). The rationale is that we are interested in the total available storage (water that plants can use), so we need to sum the storage from each soil layer (until a restricting soil layer is reached).

    Jessie, you make a good point about data sources. Usually, if you are modeling at a large scale, it's best to use one dataset (or or to make sure that values from different sources are consistent). Otherwise it's likely that differences in outputs are driven by errors in inputs.

    Also, both of you may be interested in the dataset by FutureWater (Boer, F. de, 2015. HiHydroSoil: A High Resolution Soil Map of Hydraulic Properties. Report 134; It is based on the SoilGrids1km dataset and is available upon request (to FutureWater).



  • Thanks Perrine, and yikes, I realised we didn't calculate or use our PAWC layer for InVEST, but for a simple matlab water balance that did not have a separate accounting for soil depth, hence needed to consider only where the majority of roots were.
    For InVEST and just about every other water balance, the average should indeed be taken for the full soil profile, not the first metre, so that the total available water storage will be given by AWC x depth. The difference in our case is very small, but I should have realised this before posting. Apologies!
  • hi InVEST users,
    I want to run InVEST models (sediment retention model and nutrient retention) but whenever i try to download INVEST setup ,I cant see the interface   for both models.

    apart from that my water yield model can not give me full results like (excel files ) in output folder
    it run for a long time ......without finishing!!!


  • jdouglassjdouglass Administrator, NatCap Staff
    Hello @rkundo, sediment retention and nutrient retention have been replaced by newer models SDR (sediment delivery ratio) and NDR (nutrient delivery ratio).  The latest version of InVEST should have start menu links for both SDR and NDR.

    Is is possible that you're running water yield on very large input datasets?  It's possible that the model may take a while to run with very large data.  Could you attach a screenshot of the model window?  That would help us determine where the model might be held up.

  •  thanks James, sorry I didnt read all updates before posting!

    I think the input datasets for my water yield is not large, let me attach a screenshot and take a look or you may give me your
    email so Ican  drop my inputs
  • the attached file is screenshot of my model
    160 x 84 - 3K
  • jdouglassjdouglass Administrator, NatCap Staff
    Not to worry!  My email is
Sign In or Register to comment.