Spectral entropy as a mean to quantify water stress history for natural
vegetation and irrigated agriculture in a water-stressed tropical
Spectral entropy (Hs) is an index which can be used to measure the
structural complexity of time series data. When a time series is made up
of one periodic function, the Hs value becomes smaller, while Hs becomes
larger when a time series is composed of several periodic functions. We
hypothesized that this characteristic of the Hs could be used to
quantify the water stress history of vegetation. For the ideal condition
for which sufficient water is supplied to an agricultural crop or
natural vegetation, there should be a single distinct phenological cycle
represented in a vegetation index time series (e.g., NDVI and EVI).
However, time series data for a vegetation area that repeatedly
experiences water stress may include several fluctuations that can be
observed in addition to the predominant phenological cycle. This is
because the process of experiencing water stress and recovering from it
generates small fluctuations in phenological characteristics.
Consequently, the value of Hs increases when vegetation experiences
several water shortages. Therefore, the Hs could be used as an indicator
for water stress history. To test this hypothesis, we analyzed Moderate
Resolution Imaging Spectroradiometer (MODIS) Normalized Difference
Vegetation Index (NDVI) data for a natural area in comparison to a
nearby sugarcane area in seasonally-dry western Costa Rica. In this
presentation, we will illustrate the use of spectral entropy to evaluate
the vegetative responses of natural vegetation (dry tropical forest) and
sugarcane under three different irrigation techniques (center pivot
irrigation, drip irrigation and flood irrigation). Through this
comparative analysis, the utility of Hs as an indicator will be tested.
Furthermore, crop response to the different irrigation methods will be
discussed in terms of Hs, NDVI and yield.