Mapping Dindéresso Forest Landscapes with Sentinel-2 and Machine Learning | InformativeBD

Boalidioa Tankoano,  Dramane Ouedraogo,  Zézouma Sanon, Jérôme T. Yameogo, and Mipro Hien, from the different institute of Burkina Faso. wrote a Reseach Article about, Mapping Dindéresso Forest Landscapes with Sentinel-2 and Machine Learning. Entitled, Mapping heterogeneous landscapes using sentinel-2 imagery and machine learning algorithms: A case of the Dindéresso classified forest. This research paper published by the International Journal of Biosciences (IJB). an open access scholarly research journal on Biosciences. under the affiliation of the International Network For Natural Sciences| INNSpub. an open access multidisciplinary research journal publisher.

Abstract

The anthropization of natural ecosystems has not excluded the domain classified by the State. As a result, the landscape of protected areas such as the Dinderesso Classified Forest is highly heterogeneous. The overall objective was to assess the performance of machine learning algorithms in better mapping the land use classes of the Dinderesso Classified Forest. To do this, a Sentinel-2 image and information collected in the field were used. The Sentinel-2 image was classified using Random Forest and Support Vector Machine algorithms. 850 regions of interest were selected for model training and validation. Random Forest performed best, with a Kappa coefficient of 91.49% compared with 90.17% for Support Vector Machine. The F-score for the Bare land and Agroforestry parks class was the highest (0.98) and the Gallery and Dense Vegetation class had the lowest F-score (0.82). Both algorithms showed high levels of performance, so they are suitable for classifying heterogeneous landscapes. The proportion of the Bare land and Agroforestry parks class was 29.29% compared with 70.71% for the natural formation classes (shrub savannahs, tree savannahs, Gallery, and Dense Vegetation). Given the level of anthropization of the Classified Forest, measures need to be taken to limit this process to conserve biodiversity.

Read more : White Spot Syndrome Virus: A Major Threat to Shrimp Farming in Asia | InformativeBD

Introduction

Burkina Faso, a Sahelian country, is home to major reservoirs of biodiversity in West Africa (Ouoba, 2006; Tankaono et al., 2017; Tiendrebeogo et al., 2019). The State's classified domain, which covers around 14% of the national territory, is the foundation of the national biodiversity conservation policy (Tankaono et al., 2016; Zida et al., 2015). However, human activities such as inappropriate agricultural practices, overpopulation, exploitation, and urban sprawl, combined with the poverty of rural populations, constitute serious threats to this classified State domain (Tankoano et al., 2015; Sanon et al., 2019). According to the latest report on Burkina Faso's forests, around 60% of the country's protected areas are under human occupation (DIFOR, 2007). Between 1990 and 2015, the surface area of plant cover was reduced by around 1% per year (FAO, 2015). One of the main causes of this deforestation of protected areas is agriculture and gold panning (Ouedraogo et al., 2010; Dimobe et al., 2015; Soulama et al., 2015; Zoungrana et al., 2015; Semeki Ngabinzeke et al., 2016). These two main activities lead to the fragmentation of the forest ecosystems in these protected areas (Kabulu et al., 2008; Kpedenou et al., 2016; Tankoano et al., 2016; Sanon et al., 2019). Faced with this situation, monitoring the country's last vestiges of biodiversity is becoming crucial, even imperative, at the risk of witnessing an erosion of national biodiversity. Unfortunately, financial and human resources are lacking.

Most studies concerning vegetation cover mapping in Burkina Faso are based on Landsat satellite images, but very few have used Sentinel-2 images. Nowadays, remote sensing has become a powerful tool for monitoring protected areas. Satellite imagery is commonly used to study the dynamics of land-use units, mutations between land-use units, and the impacts of agricultural activities and logging (JofackSokeng et al., 2016; Gansaonré et al., 2020; Tankoano et al., 2023). These various activities within protected areas lead to a certain het erogeneity in the landscape, which makes it difficult to classify land-use units with a high level of precision.

More and more satellites and classification algorithms are being developed for this purpose. Machine learning algorithms are also being used to classify satellite images. Sentinel-2 images, with their high resolution (10m), make it easier to detect the smallest units in the landscape. Machine learning algorithms enable accurate cartographic results, facilitating timely decision-making by protected area managers. 

However, the application of machine learning algorithms in classifying heterogeneous ecosystems has been explored little. Their contribution to improved accuracy, hence the reduction of interclass confusion, therefore needs to be explored in highly heterogeneous savannah ecosystems.

This study aims to evaluate the ability of machine learning algorithms to classify a heterogeneous landscape using a sentinel-2 image with high accuracy. Specifically, the aim was to (i) map the Dinderesso Classified Forest using a Sentinel-2 image and machine learning ; (ii) assess the ability of each two machine learning algorithms (RF and SVM) to better classify the land use/land cover within Dinderesso classified forest.

Reference

Breiman L. 2001. Random forests. Machine Learning 45, 5-32. https://doi.org/10.1023/A:1010933404324

Chowdhury MS. 2024. Comparison of accuracy and reliability of random forest, support vector machine, artificial neural network and maximum likelihood met hod in land use/cover classification of urban set ting. Environmental Challenges 14. https://doi.org/10.1016/j.envc.2023.100800

Congalton R. 1991. A review of assessing the accuracy of classification of remotely sensed data. Remote Sens. Environ. 37, 35–46. https://doi.org/10.1016/0034-4257(91)90048-B

Cracknell MJ, Reading AM. 2014. Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information. Comput. Geosci. 63, 22–33. https://doi.org/10.1016/j.cageo.2013.10.008

Dagne SS, Hirpha HH, Tekoye AT, Dessie YB, Endeshaw AA. 2023. Fusion of sentinel-1 SAR and sentinel-2 MSI data for accurate urban land use-land cover classification in Gondar City, Ethiopia. Environmental Systems Research 12(1), 40. https://doi.org/10.1186/s40068-023-00324-5

Diallo H, Bamba I, Barima YSS, Visser M, Ballo A, Mama A, Vranken I, Maïga M, Bogaert J. 2011. Effet s combinés du climat et  des pressions anthropiques sur la dynamique évolutive de la végétation d’une zone protégée du Mali (Réserve de Fina, Boucle du Baoulé). Sécheresse 22(3), 97-107. DOI: 10.1684/sec.2011.0306

Dimobe K, Ouédraogo A, Soma S, Goet ze D, Porembski S, Thiombiano A. 2015. Identification of driving factors of land degradation and deforestation in the Wildlife Reserve of Bontioli (Burkina Faso, West Africa). Global Ecology and Conservation 4, 559-571. https://doi.org/10.1016/j.gecco.2015.10.006

Foody G. 2002. Status of land cover classification accuracy assessment. Remote Sens. Environ. 80, 185–201. https://doi.org/10.1016/S0034-4257(01)00295-4

Geymen A, Baz I. 2008. The potential of remote sensing for monitoring land cover changes and effects on physical geography in the area of Kayisdagi mountain and its surroundings (Istanbul). Environmental Monitoring and Assessment 140(3), 33-42. https://link.springer.com/article/10.1007/s10661-007-9844-6

Gholamy A, Kreinovich V, Kosheleva O. 2018. Why 70/30 or 80/20 relation between training and testing sets: A pedagogical explanation. Dep. Tech. Rep. 1209, 1–6.

Inoussa MM, Mahamane A, Mbow C, Saâdou M, Yvonne B. 2011. Dynamique spatio-temporelle des forêts claires dans le Parc national du W du Niger (Afrique de l’Ouest). Sécheresse 22(3), 97-107. DOI: 10.1684/sec.2011.0305

Islami FA, Tarigan SD, Wahjunie ED, Dasanto BD. 2022. Accuracy assessment of land use change analysis using Google Earth in Sadar Watershed Mojokerto Regency. IOP Conf. Series: Earth and Environmental Science 950, 012091. https://iopscience.iop.org/article/10.1088/1755-1315/950/1/012091

Kabba STV, Li J. 2011. Analysis of land use and land cover changes, and their ecological implication in Wuhan, China. Journal of Geography and Geology 3, 104-118.

Liu C, Frazier P, Kumar L. 2007. Comparative assessment of the measures of thematic classification accuracy. Remote Sens. Environ. 107, 606–616.

Mbow C. 2009. Potentiel et  dynamique des stocks de carbone des savanes soudaniennes et  soudano-guinéennes du Sénégal. Thèse de Doctorat d’Et at, Université Cheikh Anta Diop, Dakar, Sénégal, 319p.

N’Da DH, N’Guessan EK, Wadja ME, Affian K. 2008. Apport de la télédétection au suivi de la déforestation dans le parc national de la Marahoué (Côte d’Ivoire). Télédétection 8(1), 17-34.

Nery T, Sadler R, Solis-Aulestia M, White B, Polyakov M, Chalak M. 2016. Comparing supervised algorithms in land use and land cover classification of a Landsat time-series. Int. Geosci. Remote Sens. Symp, 5165–5168.

Ouédraogo I, Tigabu M, Savadogo P, Compaoré H, Oden PC, Ouadba JM. 2010. Land cover change and its relation with population dynamics in Burkina Faso, West Africa. Land Degradation and Development 21, 453-462.

Pointius RG Jr. 2000. Quantification error versus location in comparison of categorical maps. Photogrammetric Engineering and Remote Sensing 66(8), 1011-1016.

Rahman A, Abdullah HM, Tanzir MT, Hossain MJ, Khan BM, Miah MG, Islam I. 2020. Performance of different machine learning algorithms on satellite image classification in rural and urban set up. Remote Sensing Applications: Society and Environment 20.  https://doi.org/10.1016/j.rsase.2020.100410

Smits P, Dellepaine S, Schowengerdt R. 1999. Quality assessment of image classification algorithms for land cover mapping: a review and a proposal for a cost-based approach. Int. J. Remote Sen. 20, 1461–1486.

Soulama S, Kadeba A, Nacoulma BMI, Traoré S, Bachmann Y, Thiombiano A. 2015. Impact des activités anthropiques sur la dynamique de la végétation de la réserve partielle de faune de Pama et  de ses périphéries (sud-est du Burkina Faso) dans un contexte de variabilité climatique. Journal of Applied Biosciences 87, 8047-8064.

Tabopda WG, Huynh F. 2009. Caractérisation et  suivi du recul des ligneux dans les aires protégées au Nord du Cameroun: analyse par télédétection spatiale dans la réserve forestière de Kalfou. Journées d’animation scientifique (JAS09) de l’AUF, Alger, 11p.

Tankoano B, Hien M, N’Da DH, Sanon Z, Akpa YL, Jofack Sokeng V-C, Somda I. 2016. Cartographie de la dynamique du couvert végétal du Parc National des Deux Balé à l’Ouest du Burkina Faso. International Journal of Innovation and Applied Studies 16, 837-846.

Tankoano B, Hien M, Sanon Z, Dibi NH, Yameogo TJ, Somda I. 2015. Dynamique spatio-temporelle des savanes boisées de la Forêt Classée de Tiogo au Burkina Faso. Int. J. Biol. Chem. Sci. 9(4), 1983-2000.

Tiendrebeogo M, Bamna D, Pedabga A, Goungounga J. 2019. Fiche descriptive Ramsar, Burkina Faso, Complexe d’Aires Protégées Pô-Nazinga-Sissili. Ramsar. Available at: https://rsis.ramsar.org/fr/ris/2366?language=fr

Source : Mapping heterogeneous landscapes using sentinel-2 imagery and machine learning algorithms: A case ofthe Dindéresso classified forest