Re: [ISRICWorldSoil/SoilGrids250m] depth to bedrock (DTB) surprisingly high (#44)

[Tomislav Hengl]

Hi Sophia,

Yes we are aware that our predictions of depth to bedrock come with
large errors (especially in Latin America where we have only limited
number of training points). I think this has been explained in detailed
in http://dx.doi.org/10.1002/2016MS000686

What you could do is re-callibrate our predictions using some local data
(already 200-300 points with actual DTB values could serve the purpose).
That means that you would have to fit a local model of the form:

DTB ~ DTB_soilgrids + X1 + X2 + ... + X_p

where X1 are additional (optional) local covariates. This methodology is
explained in detail in https://arxiv.org/pdf/1705.08323.pdf

Please also consider that we provide 3 variables of DTB (absolute DTB,
DTB to 200 cm and occurrence probability for R horizon) so also using
all three variables in combination can help with estimating the true DTB.

HTH,

T. Hengl


On 2018-01-26 13:30, SophiaUP wrote:
> Hello,
> At Uni Potsdam we would like to work with SoilGrids data for hydrologic
> modelling. One important parameter is depth to bedrock, which is based
> on Shangguan et al 2006.
> We were very surprised when we saw, that absolute depth to bedrock
> (BDTICM of 250 m resolution) is in our study area in Brazil always very
> high, even when we choose locations where there should be just rock,
> depth to bedrock (DTB) was at least 290 cm, sometimes even over 4000 cm.
> As we read that DTB is often overestimated, especially for shallow soils
> in the tropics, we also looked for mountainuous regions in Germany but
> once again, DTB was very high.
> Additionally, we recogniced that the limited DTB (BDRICM) also shows
> irregularities, so the indicated maximum value varies between 175 over
> 200 up to 250 cm.
> Is there any trick how we could benefit more from SoilGrids' DTB? Any
> help is very welcome.
> Best wishes,
> Sophia Dobkowitz
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub
> <https://github.com/ISRICWorldSoil/SoilGrids250m/issues/44>, or mute the
> thread
> <https://github.com/notifications/unsubscribe-auth/AAnG0mPNvVGx9FPTOep5BwEa93TqbAWqks5tOcV_gaJpZM4RuPbk>.
>