Application of Sentinel-1 satellite to identify oil palm plantations in Balikpapan Bay

M Lazecky,T Penaz,S Lhota,D Klushina

doi:10.1088/1755-1315/169/1/012064

Abstract

Satellite remote sensing has proved to be efficient for monitoring of canopy changes. In tropical areas, optical or multispectral satellite images are very often negatively affected by cloud cover, on the other hand satellites with polarimetric radars have a great advantage given their ability to penetrate clouds, smoke and atmospheric haze. Copernicus Sentinel-1 radar constellation offers both vertically co-polarized and cross-polarized imagery in a relatively high revisit time and resolution. This work describes an approach to identify selected palm oil plantations in Balikpapan Bay, East Kalimantan (Borneo). It gives an overview about advantages for monitoring temporal changes in the tropic environment using radar imagery but also constraints due to ambiguity of canopy type identification. The paper shows a brief comparison with application of multispectral Copernicus Sentinel-2 data and points a roadmap towards a practical application of the technologies.

Highlights

IntroductionIt is one of the major reasons of the loss of 100,000 Bornean orangutans over past 16 years [5], Indonesia covers only 1 % of the total land surface, yet it hosts approximately one tenth of all flowering plant species, one eight of all mammalian species and one sixth of all amphibians, reptiles and birds [6]
The plantations managed by the RSPO (Roundtable on Sustainable Palm Oil) member companies experience reduced rate of deforestation, the difference is not great compared to the conventional plantations [4]
Sentinel-1 data proved its cloud penetrating abilities but generally its only two bands failed in detailed distinguishing between an oil palm and acacia, deforested areas or other main land use classes

Summary

Introduction

It is one of the major reasons of the loss of 100,000 Bornean orangutans over past 16 years [5], Indonesia covers only 1 % of the total land surface, yet it hosts approximately one tenth of all flowering plant species, one eight of all mammalian species and one sixth of all amphibians, reptiles and birds [6]. Oil palm trees counting applications based on the processing of very high resolution multispectral QuickBird images are described for example in [17] This is a detailed mapping of individual trees, which is not suitable for assessing the occurrence of oil palm plantations and evaluating the changes of their extension in large areas. Semoi and Lawe-Lawe Rivers are planned as future reservoirs whilst others such as Riko and Tempadung Rivers, provide major sources of underground water

Input data and processing methodology

Findings

Discussion and conclusions