Hi Researcher,
I am currently working on my master's thesis on Digital Soil Mapping – Soil Organic Carbon. Since I am using soil profiles from a legacy soil survey as training data, I am facing some confusion. The attributes meet my needs, but the temporal dimension poses a challenge.
I have over 700 soil profile points, covering the period from 1980 to 2005. I only want to use this data for the year 2003, but after filtering, the density of the profile points is low, covering only 50+ points.
I would like to ask, how can I resolve this issue? Can I use the entire sample set as training data despite the temporal differences? If so, is there a need for special treatment of this temporal attribute, such as harmonizing or aggregating SOC values up to 2003?
Thank you.
It seems quite complicated, so it is difficult to find resolution. Anyway, I suggest to:
1. Prepare at least 2 versions of DSM, based on legacy data:
- with the whole dataset of 700 soil profiles (or about 600+ profiles up to year 2002;
- with dataset of 50+ points from the years 2003+,
and than compare their agreement.
2. Be aware, that the soil changes, and the content of SOC could be different at moment of obtaining data and this changes depend on land use (arable, pastures, forest?), fertilization, particularly manuring, and other factors. For this reason the assessment of current status of SC in the points of obtaining legacy data would be very important, if possible.
3. Be aware on some inaccuracies which could have appeared during soil mapping, such as less exact position of soil sampling points or profiles (it was difficult to record precisely geographic coordinates GPS receivers for earlier studies), map generalization, perhaps other methods for determination of SOC?
- Badges
- Science topic
- Similar topics
- Thesis Research