Crop identification for the delineation of irrigated regions under scarce data conditions: a new approach based on chaos theory

Abstract:

Groundwater is one of the main water reservoirs used for irrigation in regions of scarce water resources. For this reason, crop irrigation is expected to have a direct influence on this reservoir. To understand the time evolution of the groundwater table and its storage changes, it is important to delineate irrigated crops, whose evaporative demand is relatively higher. Such delineation may be performed based on classical classification approaches using optical remote sensing. However, it remains a difficult problem in regions where plots do not exceed a few hectares and exhibit a very heterogeneous pattern with multiple crops. This difficulty is emphasized in South India where two or three months of cloudy conditions during the monsoon period can hide crop growth during the year.

An alternative approach is introduced here that takes advantage of such scarce signal. Ten different crops are considered in the present study. A bank of crop models is first established based on the global modeling technique [1]. These models are then tested using original time series (from which models were obtained) in order to evaluate the information that can be deduced from these models in an inverse approach. The approach is then tested on an independent data set and is finally applied to a large ensemble of 10,000 time series of plot data extracted from the Berambadi catchment (AMBHAS site) part of the Kabini River basin CZO, South India.

Results show that despite the important two-month gap in satellite observations in the visible band, interpolated vegetation index remains an interesting indicator for identification of crops in South India.

[1] S. Mangiarotti, R. Coudret, L. Drapeau, & L. Jarlan, Polynomial search and global modeling: Two algorithms for modeling chaos, Phys. Rev. E, 86(4), 046205 (2012).