Agricultural Information Management System (AGIS): Mapping System of Sago Potential in Southern Papua, Indonesia

Sri Murniani Angelina Letsoin

Disertační práce

Abstract:

The sago palm (Metroxylon Sagu Rottb) is one of the ecological tree species that may grow wildly in the forest, primarily in Southeast Asian countries and Papua New Guinea (PNG). Sago palm trees could potentially be found in a variety of environments in Indonesia, particularly in South Papua. About 85% of the world’s sago production is in Indonesia, of which 90% is located in Papua and West Papua [1]. A number of past studies [2–5] revealed the food and non-food industry features of sago. It is possible to use the bark, leaves, starch, and sago waste. The bark can be used for traditional flooring, walls, or craft paper. Further, the leaves are for roofing, and the waste is for animal feed or compost. Sago palms contain a lot of starch which is used as a food product in traditional cakes, as well as by the food and beverage industries, and also a raw material for the agro-industry, biopesticides, and the bioethanol industry[2,5–7]. In our study location, which is Merauke Regency, the easternmost city in Indonesia, sago palm trees normally grow in wild stands with a height of between 7-15 meters. These trees associate with different types of ecosystems, such as peatland areas or swampy forests. During the harvesting season, which may be distinguished manually by the blossoming of white flowers between the leaves on the tops of palm trees, sago plays an important role as a staple food in the area. However, some previous studies highlighted the effect of land use changes, for example, the conversion of sago areas to other crops [8,9] or the inefficient utilization of resources [10]. The sago area was also not investigated yet in the statistical report of local stakeholders, besides grains crops, due to the manual inspection since 2016 [11]. Conversely, the advancement of technology nowadays has been determined to investigate sago palm condition, for instance, by extracting satellite imageries combined with different methods, such as support vector machine (SVM), object-based image analysis (OBIA), and image processing [12]. Nevertheless, the study pointed out that morphology and similarity with other palms could affect the classification result. Another sago palm detection model uses the convolutional neural network (CNN) architecture, namely Alex Net, Xception, ResNet and CraunNet, to identify the maturity of sago acquired from Unmanned Aerial Vehicle (UAV) images [13]. Moreover, maximum likelihood as a classifier in sago palm distribution from the satellite was studied in the Philippines [14]. Sago palm terrain based on sago environments, such as climate, proximity from a river or lake, altitude, gradient derived from spatial satellite data which integrated with field data, other geographical information software, i.e., ArcGIS and supervised classifier, have also been studied in Jayapura, a different region of Papua Province [15]. Nonetheless, the previously related works were not applicable to our fieldwork settings, for example, the challenge of harvesting time prediction. As previously mentioned, local communities employed a visual eye inspection to estimate the harvest time. However, the human evaluation might be influenced by the palms’ height as well as their ecosystems, such as a long river or lake. Second, the statistical report was not provided by the stakeholders; therefore, the existing condition of sago palm in this area is still unresearched. Third, the impact of land use changes on the sago environment in this area is not examined. In fact, a previous study found that this factor might have contributed to the gradual decreasing pattern of sago habitat. Fourth, the sago palm in our area is predominantly wild sago, which overlaps with other vegetation, and the sago canopy is unclearly defined at a certain altitude. A previous study used satellite data combined with machine learning methods and image processing in sago palm mapping as one of the approaches; however, the wild stand and the

Abstract:

similarity of palm life with other vegetation are challenging. To address this, sago palm study in view of sago detection is urged to be researched. Despite the fact that sago palm has multiple advantages, there is substantiation of a declining trend in the ecosystem, as well as a lack of visual eye inspection to determine the harvest time and to recognize the palm; the expected study to support the development and sustainability of the sago palm must be taken into account.

Keywords

sago, detection, classification, model

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 14. 6. 2023

Obhajoba závěrečné práce

• Vedoucí: prof. Ing. David Herák, Ph.D.
• Oponent: Jitka Kumhálová, doc. Mgr., Ph.D., Michal Petrů, externi, Gürkan Alp Kagan Gürdil, externi

Česká zemědělská univerzita v Praze

Technická fakulta

Doktorský studijní program:
Engineering of Agricultural Technological Systems