BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.earthmonitor.org//soil-health-now-2025//QTNE37
BEGIN:VEVENT
UID:pretalx-soil-health-now-2025-GVZ39A@pretalx.earthmonitor.org
DTSTART:20250408T100000Z
DTEND:20250408T103000Z
DESCRIPTION:Machine learning is promoted as a game-changer for soil health 
 assessment\, offering new ways to model complex relationships and generate
  high-resolution soil property maps. However\, while ML has shown promise\
 , its application in soil science is often met with overstated expectation
 s and underappreciated limitations. This keynote critically examines the r
 ole of ML in space-time soil mapping for soil health\, highlighting both i
 ts strengths and pitfalls.\n\nML has certainly advanced soil mapping in an
  unprecedented way to achieve continental and even global maps at high res
 olution for numerous soil properties. Entangled soil processes and the var
 iability of locations\, all nearly having an individual set of soil-formin
 g factors\, result in complex space-time soil patterns. In the commonly us
 ed mapping approach\, ML has to learn all this complexity fully data-drive
 n from the surveyed soil samples and the environmental predictors such as 
 remote sensing data or elevation models. For cases where no environmental 
 predictor dataset can differentiate the observed soil property patterns\, 
 ML predictions will play save and predict the average observed value for s
 imilar locations. From a soil process knowledge perspective\, the mean mig
 ht often not be the best prediction. For example\, a forest topsoil may be
  buffered by carbonates and have a pH around 8 or its pH might already hav
 e dropped to reach the aluminum buffer range of around 4. A mean pH of 5-6
  likely to be predicted by ML is not often observed within unfertilized fo
 rests and\, hence\, is rather unlikely.\n\nSimilar limitations also appear
  while quantifying prediction uncertainty at each location. ML-based predi
 ction intervals often contain value ranges that\, from a soil process view
 point\, we already know are very unlikely. While certainly more field surv
 eying is due to support unbiased mapping\, it does not resolve the challen
 ge. The marginal benefit of more data points for fully-data driven ML ofte
 n decreases rapidly\, more so in the presence of measurement errors. Soil 
 sampling will always only provide a tiny fraction of the total 3D soil con
 tinuum we are interested in. Data-hungry ML techniques such as deep learni
 ng are therefore unlikely to excel in space-time soilmapping of soil healt
 h indicators.
DTSTAMP:20260807T104600Z
LOCATION:HugoTECH
SUMMARY:Machine learning for soil health - Is the horizon the limit? Flaws\
 , potentials and future challenges - Madlene Nussbaum
URL:https://pretalx.earthmonitor.org/soil-health-now-2025/talk/GVZ39A/
END:VEVENT
END:VCALENDAR