GIS and remote sensing approach to modeling spring soil water for Montana rangelands

1
GIS and remote sensing approach to modeling
spring soil water for Montana rangelands
Joel Sankey Land Resources and Environmental
Sciences Department Montana State University
2
Introduction

• Soil survey
• Rangelands
• Role of soil water

3
3 soil
3 precip
6 soil
6 precip
precipitation
Yield (lbs./acre)
r .76
r .84
r .74
r .80
r .72
Moisture (inches)

• Rogler and Haas, 1946

4
Spring Soil Water Stocking Rates
Acclaimimages.com
5
Project Goals

• Model spring soil water
• Ranch management tool
• Remote sensing and GIS
• Minimize field data collection

6
Spatial Distribution of Spring Soil Water
7
Spatial Distribution of Spring Soil Water
Spring soil water content
Storage Inputs (P) - Outputs (ET re-transport)

• Solve for Storage w/ GIS
• 3 data types

8
Objectives

1. Create and test a spatially explicit model to
   predict soil water
   -test with decreased soil water
   sample size
2. Evaluate suitability of soil survey data for such
   a model

9
Study Sites
10
Methods Sampling Design

• Sample locations
• Avoid bias and autocorrelation
• Random w/in soil survey map units
• Each major component sampled twice

20 km
12 km
11
Field Methods

• Field sampling
• 1 meter profiles
• 8 samples/profile

12
Methods Sample Size
Henthorne n 82
Decker/Bales n 100
Validation n 41
Calibration n 41
Calibration n 50
Validation n 50

• Decker 100 locations 8 depths 800
• Henthorne 82 locations 8 depths 656

13
Analysis Objective 1

• Raster Modeling
• Multiple regression
• Predict gravimetric water content
• Landsat
• DEM
• Soil survey

14
Analysis Objective 1

• Model Validation
• Reserved water data set
• Predict reserved data set
• Test model with smaller data sets
• Determine smallest data set

15
Initial Results Raster Modeling Examples

• Decker Model (Landsat TM 8/12/03)
• PROFILE band 5 band 6 band 7 aspect
  band 5band 6 band 5band 7 band 6band 7
  band 5band 6band 7
• Henthorne Model (Landsat TM 8/01/03)
• PROFILE band 4 band 5 aspect band 4band
  5 band 5aspect

16
Initial Results Model Validation
Model Build R2 Validation RMSE (grav.) Validation RMSE (cm water)
Decker 1 .61 .040 6.7
Decker 2 .63 .035 6.4
Henthorne 1 .75 .056 9.7
Henthorne 2 .62 .067 12.9

• Interpretation
• R squared vs. RMSE (gravimetric water content)
  vs. RMSE (cm equivalent plant available water)

17
Initial Results Summary

• Landsat and DEM
• stronger predictors than Soil Survey
• Aspect
• stronger predictor than slope
• useful as interaction term with Landsat bands
• Landsat
• previous years biomass relatively strong
  predictor?

18
(No Transcript)
19
Implications

• Ranchers
• Consultants

20
Acknowledgements
UMAC Rick Lawrence Gordon Decker Wes Henthorne