Rationale For SoLIM (Soil Landscape Inference Model)

Rationale For SoLIM (Soil Landscape Inference Model)

Historically, the soil scientist’s mental model of how, when, why, and where soils occur on a landscape in a particular location is lost when he or she transfers or retires. This is particularly a sensitive and timely issue as over 50% of the soil survey workforce is eligible to retire within the next few years, resulting in a tremendous loss of information that has been acquired through years of study and observation. In addition, there is not a base of qualified and available soil scientists to fill these positions. SoLIM essentially transfers this carefully developed mental model to a knowledge base that can be stored, improved, and used at any time. The current method of mapping soils involves stereoscopic use and time-consuming manual cartographic work that introduces unintentional errors, depending on the soil scientist’s proficiency in these methods. Unfortunately, the science may be lost in the cartographic process. SoLIM replaces these somewhat antiquated and laborious practices through the use of modern GIS procedures and an automated inference scheme.

When field checked, SoLIM-derived soil’s maps exhibit a better quality map as compared to a conventional soil map. For example, field sites investigated by soil scientists confirmed that maps produced using SoLIM correctly identified over 80% of the soil series at these sites, while conventional maps correctly identified between 60% to 70%. Differences between the two maps, referred to as mismatches, showed that the SoLIM-derived map was correct 71% of the time, compared to 17% for the conventional map when field examined by a soil scientist (Zhu, et al. 2000).

• 
  A fuzzy inference engine is used to determine the similarity vector for the soil at each pixel position. As a result, fuzzy membership maps can be produced to exhibit the spatial gradation of soils. Because of limitations in producing conventional soil maps, known transition areas between polygons are recognized as inclusions in the map unit. Soil interpretations do not account for these areas. Fuzzy membership maps identify and recognize the intermediate nature of soils and provide for better interpretations.
  
• 
  Soil bodies on a detailed raster soil map may be as small as one pixel, which translates to a more detailed soil map compared to a conventional soil map which may be limited by scale. In addition, uncertainty maps can be produced using fuzzy memberships to validate decisions made on naming local soils.
  
• 
  Conventional soil polygon maps can be produced by “hardening” soil similarity vectors. Just as traditionally made soil maps have inclusions of unnamed soils within the polygon, so do SoLIM polygon maps. However, the composition of each individual polygon can be identified and described in detail, providing a more accurate and useful map.
  

SoLIM is a tool that has been developed to assist in producing more accurate and higher quality soil maps. It is not a system that replaces the soil scientist. Instead, it uses the soil scientist’s extensive knowledge of the soils in a particular area, combines it with the appropriate DEMs and key environmental information that determine conditions where soils form, and applies the fuzzy inference engine to produce an “inferred” map. Soil scientists verify the map. Discrepancies do not indicate problems with SoLIM, but reflect areas where the soil scientist’s concept of the soil model has not been fully captured and needs to be refined. The ability to revise and improve the model as the soil scientist increases his or her knowledge of soil model allows for an immediate update of the soil map.

This work is being carried out by Dr. A-Axing Zhu (axing@geography.wisc.edu) and Dr. Jim Burt (jburt@geography.wisc.edu) at the University of Wisconsin at Madison in cooperation with the Natural Resources Conservation Service. The project website is located at http://solim.geography.wisc.edu/.