Estimating Soil Moisture Using Satellite Observations in Puerto Rico
By
Harold Cruzado
Advisor: Dr. Ramón Vásquez
University of Puerto Rico - Mayagüez Campus
1. Introduction
2. Study area characteristics
3. Ground weather stations4. Instrumentation 5. Algorithm to estimate volumetric soil
moisture6. Preliminary results
Contents
Soil moisture is a key component in the land surface schemes in regional climate models in the tropics. An application of an algorithm for a selected area of Puerto Rico is presented. NOAA satellite observations produce the remote sensing data, which supply the input parameters for the algorithm. Satellite images with one (1) km resolution were used to implement the algorithm using Matlab software.
Introduction
Characteristics of Selected Region and Vegetation Types
Detailed vegetation types information
Topographic Map
Combining vegetation, soil types and topographic maps using ERDAS software
Soil Types and Profiles
The polygon arrays of the soil maps were digitalized, resulting in a complex soil surface. Each of these polygons represents a soil profile, some with more than one soil textural class and others with a single one. The depth of a complete profile is more than 2 meters for all the polygons.
Detailed and Generalized Soil Type Information
South-West map of Puerto Rico and its weather stations, visualized by Arcmap software
An aerial photo showing locations of ground weather stations
Ground weather stations
Theta prove ML2x
This device is a sensor to estimate volumetric soil moisture
with ±1% accuracy
Instrumentation
Data logger HH2
This device is used to store information from the theta probe
2
2
2 cos1
1sin
R
R
Soil texture
Soil Temperature
Surface temperature
Apparent emissivity
Roughness correction
Effective Temperature
Inversion of
Fresnel Equation
Vegetation correction
eReff
B
TT 1)( dsfdeff TTCTT
22 2
4
)cosexp()()(
h
hRR rs
Brightness temperature
Brightness temperature
Vegetation Type (ndvi)
Surface roughtness
Compute
Soil moisture
Algorithm to estimate volumetric soil moisture
Brightness Temperature
The radiating (or brightness) temperature is the apparent temperature of a blackbody. It can be measured by a remote sensing device such as a radiometer.
The possible data sources used are Band 3, 4 or 5 from NOAA satellite or L-band of SAR.
Brightness Temperature
Brightness temperature from channel 3, NOAA satellite, using Matlab software
Surface Temperature
) 5 4 ( *3. 3 4ch ch ch Ts This parameter can be approximated from air temperature near the soil surface and may also be obtained from satellite images from NOAA, using channels 4 and 5
Surface Temperature
Surface temperature image from channel 3,
NOAA satellite, using Matlab software.
The blue color indicates cloud presence.
7.2827
30.8720
Classified Soil Surface Temperature
Classified images (unsupervised, ERDAS software) of a thermal band
of a NOAA satellite showing levels of land surface temperature.
Soil Temperature
• The algorithm requires soil temperature for 10 to 15 cm of depth. This is provided by experimental stations such as Maricao, Adjuntas, Guanica, and Cabo Rojo in the study area.
• Because of insufficient data from the stations other methods need to be considered.
Soil Temperature• Method 1:
– Assuming some degrees less than surface temperature
– In presence of dense vegetation the surface and deep temperature are almost the same.
• Method 2:– By training an artificial neural network, whose inputs are the
following variables:
• Vegetation type
• Soil type
• Elevation levels
• Satellite observations on thermal frequency range
The second method is preferred for research.
Apparent Emissitivity
eR
eeff
B
TT
1
e : apparent emissitivity
R: apparent reflectivity
Due to signal attenuation, the emissivity isn’t real before making the correction
Effective Soil Temperature
)( dsfdeff TTCTT
2.8 0.802±0.006
6.0 0.667±0.008
11.0 0.480±0.010
21.0 0.246±0.009
Wavelength (cm) C
49.0 0.084±0.005
• For remote sensing applications there are a simple form to obtain this effective soil temperature, mean look up table for C constant for the wavelength being used
• The net intensity (called the effective temperature) at the soil surface is a superposition of intensities emitted at various
depths within the soil.
Effective soil temperature
This image (effective soil surface temperature) is generated in Matlab software using surface temperature and depth soil temperature (depth temperature is estimated by method 1 as mentioned before); actual colors do not represent the real value.
17.5357
28.8586
Vegetation Correction
)secexp(* VWCb
This process is required to determine the initial radiation emitted by the soil surface which depends on transmisivity. There are more than two ways to determine the transmisivity. The simplest and practical way is mentioned here.
• The first way to determine the transmisivity is:
Vegetation Correction
• Another way, used for this work, more directly to obtain transsmisivity through vegetation is by considering NDVI too:
)(6141.07049.0 NDVI
5429.1)(2857.4:5.0
)(3215.0)(9134.1:5.0 2
NDVIVWCNDVIif
NDVINDVIVWCNDVIif
To get an estimation of VWC, there was considered a function piecewise defined depending of vegetation index (NDVI):
Vegetation Correction
Then, when the transmissivity is already estimate, the reflectivity is corrected by
2/RRv
Vegetation Correction
This image (NDVI) is generated in Matlab software using channels 1 and 2 of NOAA satellite. Actual colors do not represent the real value.
-0.5426
0.6230
0
Apparent Emissitivity
eReeff
B
TT 1
where e is the apparent emissitivity, and R is apparent reflectivity
Due to signal attenuation, the emissivity isn’t real before making the correction, the following estimations for emissitivity and reflectivity are apparent, because its not considering the losses through signal trajectory:
Roughness Correction
)cosexp()()(2
42
2 hRRh rs
Where respectively Rs and Rr are reflectance of smooth and rough surface
For this preliminary work, this parameter is estimate y considering the class of soil only, in each region with same soil characteristics.
Computing soil moisture
ClaySandwp 0047.000064.006774.0
• The relationship between volumetric soil moisture and dielectric constant was comprised in two distinct parts separated at a transition soil moisture value wt,
where the wp is an empirical approximation of the wilting point moisture given by:
wpwt 49.0165.0
Compute the soil moisture
wtwpa
acbbwp
wp
and
PPcbwt
a
a
acbbwp
riw
effriiw
for ,2
4
0.57-0.481 and
porosity, soil theis P ly,respective
rock and ice, for water, constants dielectric theare,where
))1((,1,)(
2
41
2
2
For soil moisture less than wt:
Compute the soil moisture
)(
where
2for ,1
)1()(2
nit iwi
w
rwiniteff wtwpPPwt
wp
For soil moisture greater than wt:
Preliminary Results
• The algorithm was performed in Matlab software.• Soil moisture readings from satellites need to be
validated with more experimental work.• Point measurements using the soil probe are lower
than the satellite readings, which is not unexpected.
• The term “soil moisture” may need to be refined. The term “surface moisture” seems to describe the conditions better from a remote sensing point of view.
loacation town depth Sand clay Bulk density
Monte del Estado maricao 8-25 31.4 42 1.5
Monte Guillarte adjuntas 0-10 10.3 57.7 1.09
Bosque Seco Guanica 0-10 25 55 1.5
combate Cabo rojo 0-12 81.8 11.9 1.59
Table below shows the quantitative characteristics of different places where the stations provide the data
station % moisture(from station)
%moisture (from algorithm)
Monte del Estado
Monte Guillarte
Bosque Seco 2.4 0.540
Combate 2.3 0.2537
The values of soil moisture for different locations, given by the station and algorithm are as follows: