Remote Sensing Application in Mapping Agricultural Crop Areas and Monitoring Rice Maturity

Nguyen Quoc Hiep,Nguyen Anh Hung,Nguyen Duy Khanh,Pham Thi Thu Hien,Pham Quang Loi

doi:10.11648/j.stpp.20200401.16

Abstract

Climate change has evolved in an unpredictable trend and droughts have occurred more and more severely in the central provinces of Vietnam. Determining the irrigated area and water requirement for various crops and the growth stage of each crop is an urgent need as water resources for irrigation are getting scarce year by year. This research examines the application of Sentinel-2 and Sentinel-1 images to map crop areas and identify the current development stage of paddy rice areas. The images are collected and pre-processed from 2017 to 2018 for Ha Tinh Province in Vietnam. The Maximus Likelihood method is used to interpret Sentinel-2 imagery for mapping agricultural crop distribution status. The research presents a new approach for identifying rice maturity using the Sentinel-1 image series. The Overall Accuracy (OA) and Kappa coefficient methods are used to evaluate the generated maps of the agricultural crop’s distribution status. This study shows the relationship between the Sentinel-1 VH band and the growth of rice. From the image bands, we could calculate the slope of the line correlating between the VH backscattering value and the growth time of rice. Along with the local planting schedule, rice life cycle, and simple deduction, we could determine the rice growth stage at each time of image acquisition. The results identifying the rice maturity progression are illustrated for Cam Hoa commune in Cam Xuyen district and Thach Hoi commune in Thach Ha district, Ha Tinh Province.

Highlights

In recent years, climate change has evolved in an unpredictable trend and droughts have occurred more and more severely in the central provinces of Vietnam
After being downloaded from European Space Agency (ESA)'s database, the image was converted from level 1C to 2A which is a level where errors due to atmospheric, topographic, haze effects were removed and preliminary classification of the land cover was performed with sen2cor tool. 2A-level Sentinel-2 products were resampled to have a uniform resolution for image bands
The study area of Ha Tinh province was created as a subset and the reference system was converted to the EPSG: 3405 reference system of Vietnam using SNAP Desktop, an open-source code software provided by ESA

Summary

Introduction

Climate change has evolved in an unpredictable trend and droughts have occurred more and more severely in the central provinces of Vietnam. A crop classification map for a province of Valencia (Spain) was obtained from the Sentinel-1 and Sentinel-2 data using the decision tree method with an accuracy of 93.96% [1]; applying the Random Forest (RF) algorithm on Sentinel-2 and Landsat-8 data in semi-arid environments in the Eastern Mediterranean [2]; Using the Maximum Likelihood (MLC), Support Vector Machine (SVM), RF method to produce crop distribution maps from Sentinel-2, Landsat-8 [3,4,5,6]. Using Monte Carlo simulation with RADARSAT data to predict rice maturity [7]; Using ENVISAT/ASAR data to establish a rice map for the Mekong Delta in Vietnam, piloted in An Giang province with an overall accuracy of 85.3% and the kappa coefficient of 0.74 [8, 9] and a series of studies using radar satellite images to map rice distribution [10,11,12,13,14]. Height and biomass of rice were calculated based on Sentinel-1 data trained by machine learning algorithms, Multiple Linear

Methods

Results

Conclusion