Transcript Slide 1
Remote sensing and statistical techniques
to estimate soil moisture over tropical areas
1
Nazario D. Ramirez Beltran, 2Christian H. Calderón Arteaga, 3Ramon Vasquez, 4Eric Harmsen, 3Carlos R. Rios Mora
1Industrial Engineering, University of Puerto Rico – Mayagüez
2Mechanical Engineering, University of Puerto Rico – Mayagüez
3Electrical and Computer Engineering, University of Puerto Rico – Mayagüez
4Agricultural and Biological Engineering, University of Puerto Rico – Mayagüez
Abstract.
Estimation of soil moisture is a complex process since it is the result of the interactions of many
variables. In this study, a small number of the most relevant variables, which control soil moisture
dynamics, were selected based on results from several field experiments. The selected empirical
model used in this study was a piecewise regression model that expresses the monthly soil
moisture at a 1 km resolution. The proposed algorithm was successfully implemented and
validated for Puerto Rico climate conditions. The model was used to estimate the surface soil
moisture content over the 8,701 km2 area of Puerto Rico for each month between January and
June, 2005. Results from the model could potentially be useful to create soil moisture initial
conditions required by atmospheric and hydrological models, for instance, the Regional
Atmospheric Modeling System (RAMS), the Mesoscale Model (MM5) and the WASH123D
hydrological model.
Spatial Variability
Introduction.
Soil moisture affects numerous climatologically phenomena and is a critical parameter to model
the water cycle. It is well established that performances of atmospheric numerical models are
very sensitive to initial and boundary conditions. The soil moisture over land is a fundamental
component of the surface water and energy budget. The soil moisture regulates the partition of
latent and sensible heat fluxes at the surface, affecting the boundary layer. Several studies have
investigated the influence of soil moisture on the atmospheric boundary layer and have provided
insights into the importance of soil moisture in controlling the feedbacks between land surface
and atmosphere that influence climate.
Therefore the main purpose of this paper is to develop a reliable algorithm for estimating at 1 km
resolution the soil moisture over tropical areas. The temporal and spatial variability are modeled
by an empirical equation that relates the major factors that affect soil moisture.
The remote sensing data are download from
MODIS and NEXRAD, this data has been
interpolated for 1Km of resolution.
The topography and textural data was
provided for NRCS and USGS. All data has
been interpolated for 1Km of spatial
resolution.
Cross Validation
Leave-one-out cross-validation technique was used to asses the performance of the proposed
empirical models. The empirical models were validated using six months of the available
observations from January to June 2005. The cross-validation was implemented by each month
and consists of dropping one observation at a time and comparing the estimated versus the
observed value. Each time that an observation was dropped the parameters of the model were
estimated. The deviation between the estimated soil moisture and the actual observation is called
the validation error. The validation error was computed for each observation and for each month.
Methodology.
The soil moisture is a climatologically variable and is affected by atmospheric phenomena like the
rain or temperature. And also is affected by physics phenomena like the water transport in the soil
and the soil textural characteristics. Having in account these, three main points are the wilting
point, the field capacity and porosity. This three points are estimated if the texture of the soil are
known. The water contained from the field capacity up to the saturation point is gravitational water
and exhibits relatively rapid drainage. Thus, after the rain stops the soil moisture content decays in
an exponential form until it reaches the field capacity, which may occur within a couple of days of
the rainfall event. The amount of water contained from the wilting point up to the field capacity is
the available water and the water contained below the permanent wilting point is known as the
hydroscopic water.
An area of 250 x 750 meters was selected to perform an analysis of soil moisture variability. The
objective is to know moisture changes in a given area. The average was 37.7% water content.
The number of observations was 70. In 97.7% of the measurements there was a variability of +7% in moisture content.
Conclusions.
Normalized difference vegetation index (NDVI)
for June 2005.
Percentage of Clay for Puerto Rico.
A diagnostic soil moisture equation is derived from the piecewise linear multiple regression. The
soil moisture average is mainly influenced by the rain phenomena, and the soil texture is a factor
very important during the dry months. The multiple regression analysis is a good tool for
preliminary studies and to know the potential behavior of the soil moisture.
Experimental results show that the critical variables to estimate soil moisture are: rainfall, soil
texture, and vegetation index. These variables were included in all derived models. In addition the
secondary variables were also included in the model to complement the empirical relation. The
secondary variables are: the gradient air temperature, elevation, and average slope, which
contribute in a smaller proportion to explain the soil moisture.
Future Work.
The Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) instrument
on the NASA EOS Aqua satellite provides global passive microwave measurements of terrestrial,
oceanic, and atmospheric variables for the investigation of water and energy cycles. Ancillary data
include time, geolocation, and quality assessment. Input brightness temperature data,
corresponding to a 56 km mean spatial resolution, are resample to a global cylindrical 25 km
Equal-Area Scalable Earth Grid (EASE-Grid) cell spacing. Data are stored in HDF-EOS format.
This equations related the three points
described above with the textural
characteristics of the soil
We are proposing a piecewise regression model to represent the monthly soil moisture
behavior over densely vegetated areas:
We are proposing a piecewise regression model to
represents the monthly soil moisture behavior over
densely vegetated areas.
The first three rows in equation control the estimates
of soil moisture from the regression equation
assuring that the estimated soil moisture content will
fall within the interval defined by the wilting point
and the field capacity. The fourth row of equation
estimates the soil moisture that falls between the
field capacity and the saturation point. The last row
of equation 5 estimates the soil moisture for the
cases where the material over the surface is not a
vegetable.
Data from seventeen stations were studied to
identify the linear relationship between the
dependent variable (soil moisture) and the
independent variables which includes rainfall,
texture, vegetation index, air temperature, elevation
and surface slope.
The data used for building a model includes a
mixture of in-situ observations and remote sensing
data.
Advanced
Microwave
January
1,
2005;
Brightness temperature.
Accumulated Rain for June 2005
Scanning
6.9
GHz;
Radiometer
footprint
(AMSR-E)
56
km.
Percentage of Sand for Puerto Rico.
Acknowledgments
Difference between day temperature and night
temperature for June 2005.
Soil Moisture estimation for June 2005
The research has been supported by NASA-EPSCOR program with grant: NCC5-595, NOAA –
CREST grant NA17AE1625, and also by the University of Puerto Rico at Mayagüez. Authors
want to appreciate and recognize the funding support from these institutions.