Characterizing Sediment Source Variability

Coauthors

Maria Luna Miño
Taras Lychuk
Arnie Waddell
Alan Moulin

Funding

• Sediment fingerprinting links sources to downstream sediment
  • Using soil/sediment properties as fingerprints (tracers)
  • Provide an estimate of the relative contribution from each source
• Used to understand watershed processes and guide management practices

1. Identify potential sources of sediment
   • Land use, geology, erosion processes
2. Characterize sediment sources
   • Sampling
   • Soil properties
3. Collect downstream sediment
   • Suspended, bed, floodplain
4. Estimate relative proportion from each source
   • Mixing model

• Characterizing the sources of sediment is an important step
• Focus has been on:
  • Novel fingerprints
  • Fingerprint selection
• What about the sampling design?
  • Logistics
  • Cost
  • Judgement vs systematic
  • Prior information

Using a range of soil colour and geochemical properties across two contrasting land uses:

1. Quantify the variability

2. Characterize the spatial patterns

3. Assess the the importance of terrain attributes

• Sieved to < 63 um
• Geochemistry
  • Aqua-regia
  • 51 geochemical elements
• Spectral reflectance
  • FieldSpecPro
  • 15 colour coefficients

Based on previous work (Luna Miño et al. 2024)

• Ca, Co, Cs, Fe, Li, La, Nb, Ni, Rb, and Sr

• a*, b*, h*, and x

• Mean
• Standard deviation
• Skewness
• Coefficient of variation

Overall

• Colour properties and the agricultural land use
  • Exhibited lower variability and more symmetrical data
• Forested site has a more complex topography and geomorphic setting (floodplain)
  • Greater variability in SOM and grain size
• Colour properties make ideal fingerprints
• Differences between sites makes direct comparisons a bit tricky
  • Transformations?

• Spatial autocorrelation
  • Semivariograms
• Interpolation and mapping
  • Kriging

Nugget = 0.0
Sill = 7.2
Range = 580m
Spatial Class = Strong

• Spatial autocorrelation
  • Semivariograms
• Interpolation and mapping
  • Kriging

Nugget = 1.6
Sill = 2.7
Range = 269m
Spatial Class = Moderate

Semivariogram interpretation

• Small nugget reflects low measurement or sampling error
• Small sill indicates low overall variance
• Small range indicate spatial correlation persists over short distances

Spatial autocorrelation

• Soil properties at the agricultural site exhibited stronger spatial autocorrelation
  • 6 soil properties at the forested site exhibited no spatial autocorrelation
• Soil properties presented patterns that roughly matches the topography

• Terrain attributes
  • System for Automated Geoscientific Analyses (SAGA)
  • Random Forest Regression

1. Plan curvature
2. Profile curvature
3. SAGA wetness index
4. Catchment area
5. Relative slope position
6. Vertical channel network distance

• Elevation was ranked as the most important predictor
  • SAGA Wetness Index
  • Relative Slope Position
• Patterns linked to hydrologic properties and processes
• Terrain attributes can be used to guide sampling and interpret data

• Agricultural site:

  • Gently sloping terrain

  • Lower variability

  • Approximately normal data distributions

  • Moderate to strong spatial autocorrelation

• Forested site:

  • Complex terrain

  • Higher variability

  • Data often non-normal

  • Fewer properties with spatial autocorrelation