Realtime Remote Sensing Monitoring of DrippingIrrigationCotton - PowerPoint PPT Presentation

Title: Realtime Remote Sensing Monitoring of DrippingIrrigationCotton


1
Real-time Remote Sensing Monitoring of
Dripping-Irrigation-Cotton
Ofer Beeri and Ammatzia Peled. The Remote Sensing
and GIS Laboratory, Department of Geography,
University of Haifa, 31905, Israel ofer_at_geo.haifa.
ac.il

• INTRODUCTION
• The water amounts given to cotton plants effect
  its development and can change the quality and
  the quantity of the yield. These water amounts
  differ from one week to the other, and are
  determined by plants-height in 3-4 sites in each
  field. In general, mapping the field with remote
  sensing enables to describe the plant development
  and reduce the ground measurements. In order to
  achieve this aim and to lower damage made to the
  plants, the goal of this research was set
• Creating an overall model, for monitoring cotton
  by using real-time remote sensing
• The intermediate goals were
• to determine the amounts of water according to
  the relationship between crop parameters
• (height, wetness) and spectral crop features
• (2) to develop a Remote Sensing Formula for
  Mapping the Cotton Plants Height.
• RESEARCH METHOD
• Two major factors played a role in the
  experiment 2 cotton types (Pima PF-15 PIMA
  EP-1) and 6 water supply treatments.

RESULTS developing a Formula for Mapping the
Cotton Plants Height Collection of spectral data
and plant height enabled calculating statistical
correlation between those parameters. Based on
this, formulae for mapping plant-height in cotton
can be developed. However, it is not possible to
create a general formula since the study results
indicate major differences between cotton types.
The formula developed in the framework of this
study is designed for the PF-15 type since it
will be the main cotton type crop in Israel. Two
of the indices produced through regression
analysis, were found to map plant height
distribution better than the rest GRndvi and
TCHVI. The first index, GRndvi, had already
emerged in previous experiments as the best index
correlated to height. In order to describe
plant-height distribution using these indices,
two regression formulae were calculated
RESULTS Determining the amounts of water For
each photography session, maps of plant-height
and moistness percentage were produced. In
addition, a map depicting changes in the plants
development (plant height on one epoch minus
plant height on the previous epoch) was also
produced. Decisions regarding water quantities to
be applied to the blocks receiving remotely
sensed treatment, were based on the map of
development changes (plant-height differences).
The decision was made according to the monitoring
method growth of more than 2 cm/day requires
reducing water quantities and growth of less
than 1.5 cm/ day requires increasing quantities.
Formulae coefficients grow as time passes,
because the field observations measure the exact
plant height, while remote-sensing measurements
include the entire surface in the calculations.
In order to develop a regression formula for each
growing season, the Day of Season (DOS) figure
was incorporated into each of the formulae. These
formulae are Predicted Height (DOS0.88)
((DOS0.003) GRndvi) (1)
Predicted Height
(DOS1.2) ((DOS/360) GRndvi) ((DOS/400)
TCI) (2) Maps produced by these formulae were
compared to the ground truth (reality) derived
from the 1999-2000 experiments as presented in
table below.
The yield from the remote sensing treatment
blocks was not greater than that of the control
treatment blocks. However, the amount of water
used was 5 less than in these control blocks (as
shown in the graph below).
Each week, from 15.5.00 until the 26.8.00, the
field was aerially-photographed and the plant
measurements were taken (height, leaves moisture)
from each block. The photographs were
geometrically and radiometrically corrected, and
vegetation indices (such as NDVI) were produced.
The spectral values corresponding to the exact
locations of the ground measurements were
recorded for each photography session.
Note all figures are in cm.
SUMMARY AND CONCLUSION (1) A newly remotely
sensed method enables to produce maps and
recommendation within 45 hours. The study shows
that the blocks monitored with remote sensing did
not produce greater yield, but the method did
succeed in reducing water use by 5. The fact
that the remotely sensed monitoring included
both cotton types was found to be problematic,
since it turned out there were significant
differences between them, in terms of reaction to
changes in water quantities, spectral
reflectance, etc. This influenced the quality of
monitoring as evident from the results. (2)
Taking the example of cotton plant-height, a
method is presented for developing spectral
formulae for mapping a vegetation feature
required for precise monitoring. The formulae
displayed here, are not corresponding to reality
by a 100. Yet, the difference between mapping
with the formulae compared to the true plant
height is not more than a few millimeters. This
is insignificant for the purpose of monitoring.
Therefore, it may be postulated that the formulae
given here for plant-height are a promising
solution for precise agriculture monitoring.