Relation between Ground-based Soil Moisture and Satellite Image-based NDVI

Relation between Ground-based Soil Moisture
and Satellite Image-based NDVI
Presenter: Xianwei Wang
April 15, 2005
Instructor: Dr. Hongjie Xie
Earth and Environmental Science Department
University of Texas at San Antonio

Introduction
• Soil moisture is very important parameters for vegetation
growth and climatic and hydrological modeling. Surface
soil moisture (<10cm) can be easily acquired, like
remotely sensing by microwave, or estimating with
satellite surface vegetation index. However, there was few
ways to obtain the deeper zone soil moisture. Ground-based
measurement of soil moisture is very expensive
and can’t satisfy the need of climatic and hydrological
modeling. The Soil Climate Analysis Network (SCAN)
provides the hourly profile soil moisture, which provides
us chance to estimate soil moisture using MODIS NDVI.

Objectivity
To find if there is any correlation
between soil moisture and MODIS
image-based NDVI.
To estimate the soil moisture using
NDVI if there is good correlation
between soil moisture and NDVI.

Soil Climate Analysis Network (SCAN) Sites
http://www.wcc.nrcs.usda.gov/scan/

Research Areas and Period
Three sites with different climate and vegetation.
Adams Ranch (2015) in New Mexico, semi-arid region,
grassland.
Walnut Gulch (2026) in Arizona, semi-arid region,
shrubland;
Prairie View (2016) in Texas, moderate humid region,
grassland, to be completed.
Period is from Feb 26, 2000 to Apr 31, 2004.

SCAN Site Information
Adams Ranch: 2015
Lincoln County in New Mexico
Latitude: 34° 15' N
Longitude: 105° 25' W
Elevation: 6175 feet
Period of Record: 10/1/1994 to Present
Walnut Gulch: 2026
Cochise County, in Arizona
Elevation: 4500 feet
Prairie View: 2016
Waller County in Texas
Elevation: 270 feet

Data Sources
Soil moisture
Soil Moisture data was downloaded from three SCAN
Sites.
Measured with neutron probe;
Including 5cm, 10cm, 20cm, 50cm, and 100cm depth.
Frequency: hourly
Period is from Feb, 2000 through Apr, 2004

1 2 3
4 5 6
7 8 9
Data Source
NDVI
NDVI was calculated using band1 and
band2 from the MODIS images
downloaded from Earth Observation
System (EOS) data gateway.
NDVI= (R2-R1) / (R2+R1)
R1, R2: Reflectance of band1, 2.
Period is from Feb, 2000 through Apr,
2004
250 by 250 meter spatial resolution
Frequency: daily
8-day average

Research Methods
Compare time-series data (8-day average)

Methods
Time series average of five-year data
SM d i
n
SM d
n
i Σ=
= 1
( , )
( )
Cross Correlation analysis
R = C i j
2 ( , )
SQRT C i j C i j
( ( , ) * ( , ))
R2 is a matrix of correlation coefficients from matrix X.
C is the covariance matrix of Matrix X.
SQRT(X) is the square root of the elements of Matrix X.
X is a matrix composed of soil moisture and NDVI

Methods
Regression Analysis
Y is an n by 1 vector of observed
soil moisture.
X is an n by 2 matrix composed
of 1 and NDVI.
b is a p by 1 constant vector
calculated from X and Y
^Y, estimated n by 1 vector.
r or e is an n by 1 vector error
between observed soil moisture
and estimated soil moisture.

Results: Plot Data
Five-year average soil moisture at NM Ranch

Correlation between NDVI and soil moisture
(through five years) at NM Ranch
Under 95% confidence level
Soil moisture has small correlation with simultaneous NDVI through five
years. But the time-lagged NDVI increases their correlation.

Correlation between average NDVI and
average soil moisture (through a year) at NM
Five years time-series average soil moisture has a better correlation with average NDVI.
Their correlation increases up to 0.7 at 100cm deep when NDVI lags SM 40 days.

Correlation between SM and NDVI at NM
(May-Aug)
NDVI has good correlation with 10 and 20cm soil moisture, but small correlation
With 50cm and 100cm deep soil moisture during May-Aug.

(April-July)
NDVI has good correlation with 100cm soil moisture, but small correlation
With 10cm, 20cm and 50cm deep soil moisture during April-July.

(Sep-Mar)
NDVI has small correlation with soil moisture during non-growing
season, while slightly increase as time lags.

Regressed 100cm SM based on growing-season
NDVI VS observed growing-season SM
at NM
Regression based on simultaneous growing season NDVI at 95% CL

at NM

Average soil moisture at AZ and NM
Arizona New Mexico

Correlation between 100cm SM and NDVI
on both sites
Arizona New Mexico

Correlation between 10cm SM and NDVI during
growing season (May-Aug)
Arizona New Mexico

Correlation between SM and NDVI during non-growing
season
Arizona New Mexico
NDVI has small correlation with soil moisture during non-growing season,
while slightly increase as time lags.

Correlation between different-depth average
SM and NDVI during growing season
Arizona New Mexico

Arizona New Mexico
Regression based on simultaneous growing season NDVI at 95% CL

Arizona New Mexico

Five-year average soil moisture at
TX Prairie View and NM Ranch
Texas New Mexico

Correlation between SM and NDVI during
growing season (May-Sep) at TX and NM
Texas New Mexico

Conclusion
The root and below-root zone soil moisture in semi-arid
climate has good correlation with simultaneous NDVI and
can be estimated using NDVI during growing season.
The root zone soil moisture in humid climate also has
moderate correlation with simultaneous NDVI and can be
estimated using NDVI during growing season.
The below-root soil moisture in humid climate has small
correlation with NDVI and can’t be effectively estimated
using NDVI.
Soil moisture has small correlation with NDVI and can’t be
effectively estimated using NDVI during non-growing
season (Oct-March).

Relation between Ground-based Soil Moisture and Satellite Image-based NDVI

More Related Content

What's hot (18)

Similar to Relation between Ground-based Soil Moisture and Satellite Image-based NDVI (20)

More from Sumant Diwakar (20)

Recently uploaded (20)

Relation between Ground-based Soil Moisture and Satellite Image-based NDVI