1
Early detection of drought stress in
potato (Solanum tuberosum L.) and grapevine
(Vitis vinifera L.) crops through multifractal
analysis applied to remotely sensed data
Chávez P.1,2 Ribas-Carbó M.1 Medrano H.1 Mares
V.2 Posadas A.2 Yarlequé C.2 Quiroz R.2 Flexas
J.1
Abstract The importance of timely detection of
drought stress in agricultural crops is
increasing due to the imminent climate change.
Several methodologies are being developed for
assessing, monitoring, and managing water
availability, to supply the accurate water amount
to crops, while maintaining the highest WUE
feasible. The objective of this work was to
determine the suitability of remote sensing as a
monitoring tool to detect drought stress in
plants. Continuous measurements of multispectral
reflectance and derived vegetation indices, taken
from potato and grapevine crops, were analyzed,
and compared with simultaneous measurements of
photosynthesis, stomatal conductance and
sap-flow. Multifractal analysis of reflectance
data did discriminate between the well irrigated
and drought treatments around 2 - 6 days earlier
than physiological measurements. Vegetation
indices of discrete regions also provided early
detection of drought. Results evidenced that
remotely sensed data might be useful as early
detectors of drought stress and that the use of
multifractal analysis of multispectral data might
provide a more robust discrimination between
turgent and stressed plants
The reflectance spectrum was divided into blue,
green, red and near infrared sectors for
calculating the percentage of reflectance against
time. The differences were determined using a
repeated measurements statistical analyses. The
reflectance data were pre-processed applying a
background correction, and then submitted to the
Continuous Wavelet Transform (CWT) and the
wavelet transform modulus maxima (WTMM) method
(2) using as mother wavelet analyzer the second
derivative of the Gaussian function (Mexican
hat).
   





































































   
1Research Group on Plants under Mediterranean
Conditions. University of Balearic Islands.
Crrtra. Valldemossa km.7.5, 07122, Palma de
Mallorca, Balearic Islands, Spain 2Production
Systems the Environment Division.
International Potato Center. P.O. Box 1558, Lima
12, Peru
Figure 3. Passive reflectance of potato plants
(left) and their corresponding multifractal
singularity spectra (centre). Daily
photosynthesis (right,1a) and stomatal
conductance (right,1b).
Introduction Availability of water is the most
limiting factor in crop production. This problem
will be exacerbated with the imminent climate
change. Even if the rainfall levels are held
constants, the risks of severe dryness increases
due to the rise of the evaporative atmospheric
demand caused by the global warming. Direct plant
based measurements are mainly limited to leaf
water potential by pressure chamber, stomatal
conductance by gas-exchange, and porometry. These
are time-consuming and require a number of
observations to characterize a whole field.
Non-destructive non-invasive remote sensing
methods emerge as effective alternatives for
assessing the status of crops through reflectance
and imagery (1). The aim of this work was to test
remotely sensed reflectance as a revealing
technique for retrieving invisible changes caused
by drought in live plants.
Results and Discussion The raw reflectance of
plants did not discriminate among treatments
(Figure 2) a situation corrected with
pre-processing and multifractal analysis of data
(Figure 3). In potato, discrimination was
perceived 7 dpt (i.e. 2-4 days prior to the
gas-exchange and RWC measurements). In grapevine,
it occurred 6 dpt, i.e. 2 days earlier than the
gas-exchange and sap flow measurements (Figures 4
and 5). Divided reflectance evidenced stressed
plants at around 7 dpt for both potato and
grapevine plants. The main bands for detecting
drought stress were, from the best to the worst,
the blue, followed by NIR, red and finally the
green region.

• Bibliography
• Chávez P., Yarlequé C., Piro O., Posadas. A.,
  Mares V., Loayza H., Chuquillanqui C., Zorogastúa
  P., Flexas J., Quiroz R. 2009(b). Applying
  multifractal analysis to remotely sensed data for
  assessing PYVV infection in potato (Solanum
  tuberosum L.) crops. Remote Sensing Journal.
  Under revision.
• McAteer J.R.T., Young C.A., Ireland J., and
  Gallagher P.T. 2007. The bursty nature of solar
  flare x-ray emission. The American Astronomical
  Society..The Astrophysical Journal, 662691-700.

Figure 4. Reflectance (left), multifractal
singularity spectra (centre) and daily stomatal
conductance (right) of plants from the grapevine
experiment (Gray bars indicate days of
irrigation).
Figure 1. (A) Measuring physiological parameters
in a potato plant (Solanum tuberosum L.) using an
infrared gas analyzer (IRGA) LI-6400. (B) Detail
of the IRGAs chamber.
The spectral vegetation indexes (SVI) tested did
show an inconsistent response, even those indexes
specifically developed to assess water content in
plants.
Materials and Methods Experiments were carried
out under outdoor conditions in Mallorca, Spain,
2006. In potato, the treatments were Control
(Ctrl), 100 of the daily measured
evapo-transpiration (dme), moderate-drought
(D75), 75 of dme, and severe-drought (D50), 50
of dme. Pots were weighted every day to determine
the amount of water available with respect to the
Control. Their relative water content (RWC) was
determined. Drought treatments in grapevine were
induced by stopping irrigation during 5 days per
week, as 1) severe-drought (D1), irrigation
re-initiated the evening of day 6 2)
moderate-drought (D2), irrigation re-initiated
the morning of day 6 and Control (Ctrl),
irrigated at field capacity. Light reflectance
measurements were taken at noon, through a
multispectral spectrometer. Gas exchange
measurements were performed using a portable
infrared gas analyzer (Figure 1). Sap flow
measurements in grapevine were made through the
thermo heat balance method (THB).
Figure 5. Continuous records of sap flow in
grapevine. At the beginning of the experiment, a
and b indicate that there were no differences
among treatments. Differences among control and
drought treatments are observable in c 8 days
post treatment and onwards.

• Conclusions
• Conventional methods such as gas-exchange,
  relative water content and sap flow measurements
  (considering both potato and grapevine) evidenced
  drought stress 9 11 dpt. Multifractal analysis
  of reflectance data did it 2 -4 days before. The
  challenge now is to replicate the findings in
  commercial fields.
• Although splitting the reflectance into bands
  showed comparable results, multifractal analysis
  seems more reliable since results were more
  stable along the trial.
• For a more reliable estimation of drought in
  potato through gas exchange, AN and g must be
  considered simultaneously.
• SVI are not a reliable method to retrieve water
  stress in plants.

Figure 2. Pre-processing of reflectance data.
Notice that the observable differences in the raw
reflectance spectra do not allow a logical
treatments discrimination. Neither the observable
differences in the second order derivative nor
the calculation of light absorbance, do
demonstrate the drought levels as the
multifractal singularity spectra does.