Measuring Soil Moisture Content Using Spatially Interpolated Meteorological Variables and the Soil Water Balance Equation

1
Measuring Soil Moisture Content Using Spatially
Interpolated Meteorological Variables and the
Soil Water Balance Equation

• Shanon Connelly

2
Introduction

• In situ measurements examine the phenomenon
  exactly in place where it occurs.
• The most accurate of soil moisture measurements
  are in situ, but these methods can be labor
  intensive, expensive, and destructive to the
  soil, and are only accurate at the point of
  measurement (Schmugge et al., 1980).

3
Introduction

• The first portion of this project is to find the
  best interpolation methods for spatial prediction
  of continuous surface layers from in-situ point
  measurements of weather data.
• Then, using the raster calculator function in the
  Spatial Analyst extension, the interpolated
  surface layers will serve as inputs into the soil
  water balance equation to derive soil moisture
  estimates.

4
Study Area

• The Oklahoma Mesonet
• Operated by the Oklahoma Climatological Survey

• Network of over 110 automated stations covering
  Oklahoma
• At least one Mesonet station in each of the 77
  counties
• Approximately 100 sites monitor soil moisture

5
Study Area

• 13 atmospheric and subsurface variables recorded
  every 5 minutes, producing 288 observations of
  each parameter per station per day.

• Air temperature
• Humidity
• Barometric pressure
• Wind speed

• Wind direction
• Rainfall
• Solar radiation
• Soil temperature

6
Data Availability

• Data obtained from the Oklahoma Mesonet
  encompasses 4months of observations.
• August and October,2000 - dry period
• March April 2003 - wet period
• For this project, several key dates were chosen
  based upon the climate trends occurring at that
  time.

7
Data Availability
Date Reason chosen
August 7, 2000 high 24-hour precipitation (1.88)
August 26, 2000 maximum daily temperature (111ºF)
October 19, 2000 last dry day of drought period
October 20, 2000 abundant precipitation after drought period
October 23, 2000 high 24-hour precipitation (9.15)
March 5, 2003 minimum daily temperature (11ºF)
March 19, 2003 high 24-hour precipitation (3.34)
April 16, 2003 high 24-hour precipitation (3.21)
8
Methodology

• The soil-water balance equation will be used to
  quantify soil moisture  

?w ? t
or
wi wi-1 P E S
P E S
9
Methodology

• A database was compiled containing the point data
  from each of the study areas
• Precipitation measurements
• Net radiation
• Soil physical properties
• Etc.
• Some variables were used directly, while others
  were used to derive inputs to be used in the soil
  water balance equation.

10
Methodology
Equation Component Data Parameter(s) Notes
Precipitation Rainfall 24 hour cumulative rainfall
Evaporation Station Pressure (avg) Solar Radiation (total) Relative Humidity (min, max, avg) Temperature (min, max, avg) Average Wind Speed (avg) Derived from the ACSE s standardized reference evapotranspiration equation and multiplied by an evapotranspiration coefficient.
Surplus Soil physical properties (Percent Sand/Silt/Clay, Saturation point, Field Capacity, Wilting Point, Critical Moisture Point, and Bulk Density) Derived using the soil texture triangle
? soil water content Volumetric Water Content Neutron probes in soil measure ? soil temperature and soil water content is derived.
11
Methodology

• For each layer, the data points were divided into
  two sets
• Training set with 85 of sites was used for
  developing a geospatial model
• Testing set with the remaining 15 of sites was
  used to test the performance of the model

12
Methodology

• Point data was then interpolated into continuous
  surface layers and validated.
• Various interpolation techniques were tested to
  find the most appropriate geostatistical method.
• Geostatistical (Kriging)
• Deterministic (Inverse Distance Weighted)

13
Results

• Kriging is more accurate than IDW
• Simple and Ordinary kriging methods are most
  suitable
• No data transformation was used
• Gaussian, Spherical and Exponential variogram
  models are most appropriate
• Slight modification to variogram models
• Lag size, number of lags and searching
  neighborhoods were slightly modified on each run
  to yield the best predictions

14
Results Kriging
Date 8/7/2000 8/26/2000 10/19/2000 10/20/2000 3/5/2003 3/19/2003 4/16/2003
Count 86 90 85 83 82 88 90
Minimum -165.8022 -174.4563 -210.4428 -203.9146 -197.6153 -214.0744 -217.3216
Maximum 50.1404 35.104 208.1363 199.6216 156.2219 179.8839 48.9228
Average -10.43402 -12.765687 -20.071792 -18.102 -45.082996 -55.54878 -52.40135
               
Underestimate gt 20mm 26 22 40 39 48 64 59
Underestimate gt 5mm 11 23 12 11 12 6 10
Within -5mm and 5mm 17 20 10 8 4 3 3
Overestimate gt 5mm 20 21 8 7 6 3 7
Overestimate gt 20mm 12 4 15 18 12 12 11
15
Results August 7, 2000
16
Results August 26, 2000
17
Results October 19, 2000
18
Results October 20, 2000
19
Results March 5, 2003
20
Results March 19, 2003
21
Results April 16, 2003
22
Results IDW
Date 8/7/2000 8/26/2000 10/19/2000 10/20/2000 3/5/2003 3/19/2003 4/16/2003
Count 90 90 85 83 82 88 90
Minimum -243.378 -262.7851 -292.0266 -292.8517 -314.8727 -322.8947 -398.8954
Maximum 59.895 47.5517 42.1342 46.0397 42.6735 25.3343 71.2037
Average -11.172103 -14.035441 -23.681219 -22.160365 -53.026518 -61.47354 -56.99088
               
Underestimated gt 20mm 11 13 36 31 57 70 54
Underestimated gt 5mm 43 58 26 22 11 7 9
Within -5mm and 5mm 25 15 15 19 4 8 16
Overestimated gt 5mm 8 2 4 5 8 1 7
Overestimated gt 20mm 3 2 4 6 2 2 4
23
Future Research

• Spatial join the validation layer with the master
  data layer
• Determine which stations have greatest amount of
  error, and if it is a single occurrence or
  recurring error
• Pinpoint cause of error
• Calibrate or eliminate

24
Future Research
STID Kriging 03052003 Kriging 03192003 Kriging 04162003 IDW 03052003 IDW 03192003 IDW 04162003
ACME -20.3127 -13.0814 5.7378 -64.8383 -66.6643 -19.2345
ADAX 14.2958 -13.1394 -5.6912 -27.6352 -45.6900 -15.4939
ALTU -69.0035 -89.9628 -106.8401 -21.7217 -16.4008 -27.1364
ALV2 -84.1125 -98.8147 -140.4594 -38.0241 -31.6019 -88.9651
ANTL 80.4431 40.8968 25.1486 7.8222 -4.2107 -0.7215
ARDM -53.2295 -110.6949 -111.5389 -85.3641 -88.9974 -124.9146
ARNE - -57.8628 -49.4104 - -59.8418 -38.1378
BEAV - -90.0245 -119.5648 - -80.0803 -148.6082
BESS -12.6944 -27.8246 -46.7740 3.5608 0.1717 -32.4433
BIXB 61.9558 26.7188 12.9318 15.5142 -4.5888 8.0525
BLAC -61.0129 -65.4505 -81.9027 -55.5567 -54.5864 -58.3882
BOWL -15.5319 -30.9528 8.6032 -29.3473 -46.6255 8.7024
BREC - -46.9068 14.8144 - -44.5172 29.7529
BRIS 26.1504 20.4336 37.1197 42.6735 25.3343 43.8859
25
Conclusion

• Kriging is more accurate method than IDW
• Simple and Ordinary kriging methods
• Gaussian, Spherical and Exponential variogram
  models
• Soil moisture content estimates tend to be
  greatly underestimated
• Future research to pinpoint stations with high
  errors
• Investigate further calibrate or eliminate
• Wet season yields less accurate SMC estimates
  using this methodology

26
References

• Earls, Julie, and Barnali Dixon. 2007. "Spatial
  Interpolation of Rainfall Data Using ArcGIS A
  Comparitive Study." Accessed from
  http//gis.esri.com/library/userconf/proc07/papers
  /papers/pap_1451.pdf on January 6, 2009.
• Oklahoma Climatological Survey. Estimates of soil
  moisture from the Oklahoma Mesonet. Available
  online at http//www.mesonet.org/instruments/SoilM
  oisture.pdf.
• Schmugge, T.J., Jackson, T.J., and McKim, H.L.
  1980. Survey of Methods for Soil Moisture
  Determination. Water Resources Research 16 (6)
  961-979.
• Walter, I.A., R.G. Allen, R. Elliott, M.E.
  Jensen, D. Itenfisu, B. Mecham, T.A. Howell, R.
  Snyder, P. Brown, S. Echings, T. Spofford, M.
  Hattendorf, R.H. Cuenca, J.L. Wright, and D.
  Martin. 2000. ASCEs standardized reference
  evapotranspiration equation. In Proc. of the 4th
  National Irrigation Symposium, ASAE, Nov. 14-16,
  Phoenix, AZ.