Forest Cover Change Assessment in the Sangu River Basin Using Sentinel-2 Satellite Data

doi:10.21203/rs.3.rs-3224118/v1

Download PDF

Research Article

Forest Cover Change Assessment in the Sangu River Basin Using Sentinel-2 Satellite Data

https://doi.org/10.21203/rs.3.rs-3224118/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Forest covers are changing at an alarming rate, in particular to the tropics, emphasizing the need for continuous forest monitoring using satellite technology. Forest cover change in the Sangu River Basin in the Chittagong Hill Tracts (CHT), Bangladesh is typically missing and this study aims to fill the gap by applying a suitable methodology for mapping forest cover change in the area using Sentinel-2 satellite images. Maximum Likelihood, Minimum Distance, Support Vector Machine, and Random Forest classification methods were examined and Minimum Distance Classification Algorithm was chosen for mapping forest cover. Forest cover maps of 2016 and 2020 and forest cover change map of 2016–2020 were prepared. The accuracy of image classification was assessed with 300 random points; overall accuracy of classification was 91.7% and 94.7% for 2016 and 2020 satellite images, respectively. The study revealed that the major Land Use and Land Cover (LULC) classes remained stable over the study period (2016–2020) with a total area of 98.2 thousand ha. Among these classes, forest area is extended over 46.4 thousand ha (37.5%) and shrub classes cover around 45.7 thousand ha (37.0%). The change classes comprise 23.2 thousand ha (20.6%), with the largest conversion class being forest to shrubs or forest to bare land representing deforestation, which is around 17.7 thousand ha (around 14.3%). Shifting cultivation is identified as the major driver of forest cover change in the region. This process has converted many areas of CHT into shrub-dominated areas. The result of the study can assist local forest managers in combating deforestation and conserving the remaining forest in the region.

Forest cover change

Sangu Reserve Forest

Sentinel-2

Classification

Change detection

Forest cover is changing in several places of the world because of the severe human interference. The globe has lost 178 million ha of forest since 1990 though the pace of net forest loss significantly slowed between 1990 and 2020 (FAO, 2020). Over the decades 1990–2000, 2000–2010, and 2010–2020, the rate of net forest loss has decreased from 7.8 million ha to 5.2 million ha and finally to 4.7 million hectares per year, respectively (FAO, 2020). In order to understand the dynamic changes of forest cover in different parts of the globe, there is a need for continuous forest monitoring and generating accurate statistics on forest and forest cover change.

Monitoring changes in forest cover has been notably progressed on global, regional, and local scales by the advancement of new imaging sensors and geospatial technologies. Many research efforts have been undertaken using various approaches and data-sets to detect forest cover changes throughout the globe. Hansen & DeFries (2004) utilized Advanced Very High-Resolution Radiometer (AVHRR) data and a supervised algorithm technique to assess the changes in global forest cover. Hansen et al. (2008) integrated Landsat and MODIS data to produce a per-pixel change map of forest cover in the Congo Basin. Using Landsat and IRS LISS III data, Panigrahy et al. (2010) detected changes in forest cover in Western Ghats of Maharashtra by employing a visual interpretation approach.

Spatial changes of Romania’s montane forest cover were mapped using a supervised maximum likelihood classification algorithm to Landsat and Sentinel- 2 images (Mihai et al. 2017). Multi-sensor image data from Landsat 1–8, Sentinel- 1 and 2, CBERS, ASTER, ISS have been utilized to detect changes in Brazilian forest cover (Fortin et al. 2020). Nguyen-Trong & Tran-Xuan (2022) estimated changes of coastal forest cover in Vietnam using Sentinel- 2 images and a convolutional neural networks approach. Machine learning algorithms have recently gained popularity in recent time as a technique for processing and analysing satellite images due to their ability for large amount of data handling and automation procedures. Roberts et al. (2022) generate a Python package that enables the detection of near-real-time change in forest cover in Kenya using Sentinel and Landsat data.

Forests in Bangladesh are also undergoing changes due to extreme human interference like many other tropical countries (i.e. Emch & Peterson, 2006; Zaman & Katoh, 2011; Redowan et al., 2014; Abdullah et al., 2015). Studies related to forest cover change in Bangladesh have primarily focused on the Sundarbans (Islam et al., 1997; Emch & Peterson, 2006; Giri et al., 2007; Hasan et al., 2020; Chowdhury & Hafsa, 2022). However, a few studies have also been done on the forest regions of the south-eastern part (Nath & Acharjee, 2013; Samrat et al., 2022), north-eastern part (Redowan et al., 2014; Masum & Hasan, 2020) and Sal forest located in the central and the north-western part of the country (Zaman & Katoh, 2011; Abdullah et al., 2015). The majority of these studies used Landsat images and performed a post-classification comparison strategy due to the availability of historical time-series data-set, after incorporating a supervised maximum likelihood classification algorithm.

Studies are limited to the geographical region of the Sangu River Basin located in the South-eastern corner of Bangladesh, while the current study will fill the gap. This study aims to apply a suitable methodology for forest cover and forest cover change mapping in the upper Sangu River Basin and to generate forest cover maps and forest cover change maps for the last few years.

Study Area

The study area is located in the south-eastern corner of Bangladesh and belongs to the Thanchi Upazila of Bandarban District, Chittagong Hill Tracts (CHT) (Fig. 1). It covers an approximate area of 1,021 sq. km. The terrain is undulating with elevation ranging from 20 m at the river and valley to 1,053 m at the mountain peak. The hill ranges are distributed in the north-south direction while the Sangu River is passed from the northern to the southern direction. The upazila is bounded by Myanmar on the east and south, Ruma and Belaichhari Upazila on the north, and Alikadam and Lama Upazila on the west.

The area still holds untouched patches of forest in Bangladesh and is within the boundary of the Indo-Burma Biodiversity hotspot (Sourav & Anwar, 2020, Sourav et al., 2020). The north of the upazila belongs to the Unclassified State Forest which is under the supervision of the Deputy Commissioner (DC) and the southern section belongs to Sangu Reserve Forest (SRF) controlled by the Bangladesh Forest Department. SRF was declared a reserved forest in 1881 (Chaudhury & Chaudhury, 1970). It is currently administered by the Bandarban Forest Division of CHT.

The study area is characterized by the tropical climate of CHT. Extremely hot and dry weather last from mid-March to mid-June, and the temperature falls with monsoon rains while humidity remains high. Heavy rainfall occurs between May to October, and winter lasts from November to mid-March (Chaudhury & Chaudhury, 1970). Tropical evergreen and deciduous species coexist in the study area's forest, with the former predominating. The stand composition contains about 50 different species, even though the actual floral diversity is yet unknown. Civit (Swintonia floribunda) and Garjan (Dipterocarpus spp.) are the prevalent species; Chapalish (Artocarpus chaplasha) is occasionally mixed in the forest composition.

Dataset

Forest cover change in the Thanchi Upazila has been mapped with Sentinel-2 satellite images. Sentinel-2 satellite images of Level-1C were downloaded from the United States Geological Survey (USGS) website (Table 1). The Level-1C product offers Top-Of-Atmosphere (TOA) multispectral reflectance which is orthorectified and at sub-pixel level registration. The investigation utilized the spectral bands with a spatial resolution of 10–20 m. WorldView-2 satellite image of 25 sq. km size was procured from the designated vendor to get help for satellite image interpretation and change process analysis (Table 1). Five different land use and land cover (LULC) types have been identified in the area: forest, shrubs, bare land, shadow, and water (detailed description in Table 2).

Table 1

Satellite image used in the analysis
Satellite sensor	Date	No. of bands	Spatial resolution (m)	Remarks
Sentinel-2	30 November 2016	10 bands	10m, 20m	-
Sentinel-2	29 November 2020	10 bands	10m, 20m	-
WorldView-2	17 January 2017	8 bands	0.5m	Pan sharpened, Selected sample area
WorldView-2	29 November 2020	8 bands	0.5m	Pan sharpened, selected sample area

Table 2

The description of Land Use and Land Cover (LULC) classes
Serial	Class	Description
1	Forest	The class belongs to tropical wet evergreen and semi-evergreen forest. The forest is affected by shifting cultivation and timber felling. Forest is defined by FAO as “land spanning more than 0.5 hectares with trees higher than 5 meters and a canopy cover of more than 10 per cent, or trees able to reach these thresholds in situ. It does not include land that is predominantly under agricultural or urban land use”.
2	Shrubs	The class is dominated by the shrubs and scattered trees, which is often the result of shifting cultivation and logging activity to the earlier forest class. In some places, vegetation cover remained unchanged for a longer period of time where the site productivity is low. Small scale farms belong to this class.
3	Bare Land	Bare land is the result of land clearance primarily by the shifting cultivation. No appreciable amount of vegetation is present in this class.
4	Shadow	This class may represent forest or shrubs, while the original LULC class is obscured by the presence of hill shadow during the satellite image acquisition. This class appears in a dark tone on the satellite images.
5	Water	The class is represented by the course of the Sangu River. The area of this class is covered by the water extent.

Satellite image collection

The geometric accuracy of Sentinel-2 images was checked with the available reference images and maps and found adequate for further processing. The study area is covered by the two frames, qdj and qdk, and these two scenes were joined together. Satellite images were clipped by the polygon administrative boundary of Thanchi Upazila obtained from the Survey of Bangladesh (SOB).

Satellite image interpretation and classification

Satellite images were interpreted visually on the computer screen. To build the training set for supervised classification, an extensive Area of Interest (AOI) training pixel sets were manually created by visual interpretation of the cloud-free observation of Sentinel-2 satellite images. Additional datasets that include freely available VHSR images from GoogleEarth and expert judgment will be used as reference materials to aid image interpretation and select training data sets. Training sets were chosen for the designated LULC classes. Efforts were given only the homogenous pixel group to be chosen as the training pixel sets and a minimum number of training pixel in each AOI as 100 pixels. A minimum number of 5–10 training pixel sets were selected to be used as the input of image classification.

The vast majority of image interpretation effort will be focused on distinguishing tree canopy from other vegetation like shrubs and scattered trees. With the initial set of training, the classification results will be assessed; misclassified pixels will be identified and add them to the training set. The process was repeated until the quality of the forest cover map was considered sufficient. In this process forest cover maps of 2016 and 2020 and forest cover change maps of 2016–2020 have been generated.

Different image classification methods have been examined: (i) Maximum Likelihood (ii) Minimum Distance (iii) Support Vector Machine and (iv) Random Forest and finally, Minimum Distance Classification Algorithm was chosen for map preparation. In addition, the river class was visually interpreted and digitized. In these hilly regions, the river course was relatively stable and there was no frequent change in the course. Therefore, the river class was digitized and checked with the images of 2016 and 2020 and finally, the river polygon was converted to the raster image and appended to the river class.

Accuracy assessment

Accuracy assessment of the classified satellite images was implemented by generating 300 random points on the classified satellite images by using stratified random selection criteria where each land use and the land cover class was considered as a stratum. The number of minimum points was 10 considered in the random point generation process. Random points were visualized on the unclassified satellite image and the VHSR images available for the selected sampling frame. The image was interpreted, and a reference image class was assigned for accuracy assessment.

Forest cover change assessment

The forest cover maps (Fig. 2 and Fig. 3) and forest cover change map (Fig. 4) show an overall declining trend of forest cover over the study period (Table 3). The major parts of LULC classes are stable over the study period (2016–2020) and the total area of the stable class is 98.2 thousand ha which is around 79.4%. Among these classes, forest and shrubs occupy the largest spatial extent; forest area is extended over 46.4 thousand ha (37.5%) and shrub classes cover around 45.7 thousand ha (37.0%). It should also be noted that the extent of the shadow class covers around 4.1 thousand ha which is around 3.3%. The overpass of Sentinel-2 satellite images was around 10:30 AM. A shadow appeared on the satellite images due to the sun elevation angle and the topographic condition of the study area.

Table 3

The statistics representing Land Use and Land Cover (LULC) and LULC conversion classes in Thanchi Upazila during 2016–2020
LULC class		Area (ha)	Percentage (%)
Unchanged / stable class	Forest	46,363	37.51
	Shrubs	45,698	36.97
	Bare land	820	0.66
	Shadow	4,094	3.31
	River	1,199	0.97
	Sub-total	98,174	79.42
LULC conversion type		Area (ha)	Percentage (%)
Changed class	Forest to Shrubs or Forest to Bare land	17,664	14.29
	Shrubs to forest or Bare land to Forest	3,758	3.04
	Shrubs to Bare land	1,012	0.82
	Bare land to Shrubs	806	0.65
	Other changes	2,199	1.78
	Subtotal	23,240	20.58
Total		1,23,613	100.00

The amount of change classes in Thanchi Upazila is 23.2 thousand ha (20.6%). In these change classes, the largest conversion class is the forest to shrubs or forest to bare land representing deforestation with the extent of 17.7 thousand ha (around 14.3%). The second largest change class is shrubs to forest or bare land which is around 3.8 thousand ha (3.0%). The class represents forest succession, the abandoned shifting cultivation class, and the fallow sites left after logging activity. Since forest loss is larger (14.3%) than forest gain (3.0%), the area represents the decreasing trend of forest cover in the region during the study period, 2016–2020. The process should be halted; otherwise, the remaining forest will vanish from the area in the future.

There are also other changes of 2.2 thousand ha (1.8%), which are usually uninteresting and often do not contain any useful information related to LULC conversion since these changes have originated from the satellite image classification procedure. These classes include forest to shadow or shadow to forest and shrubs to bare land or bare land to shrubs etc. The class representing forest to shadow and vice-versa is not interesting at all from the LULC change perspective while bare land to shrubs or the opposite has some implications in the LULC change process.

The location of forest cover change hotspots is scattered and distributed in different locations of Thanchi Upazila. Many change hotspots are located within the proximity of the Sangu River. This is understandable since the Sangu River is the only means of accessibility to this area as well as the way of log transportation. Within the study period, it was noticed that many change hotspots are located in the southern part of Thanchi Upazila belonging to SRF. Bangladesh Forest Department (BFD) should monitor the forest cover in the region by satellite monitoring since the accessibility is restricted to the area. The organization should revisit its policy permits to transport logs in the vicinity of SRF.

Forest cover change processes

By analysing the forest clearing patterns on the Sentinel-2 and VHSR satellite images, shifting cultivation has been identified as the major driver of forest cover change (Fig. 5). Through this process, land clearings can take place in forest land or shrub land or even in abandoned shifting cultivation places. However, it should be noted that shifting cultivation places (bare land) are usually recovered within a short period like 1–2 years. But in the recovered places only smaller vegetation is noticed after this period (Fig. 6).

Shifting cultivation is a prominent land use in the Chittagong Hill Tracts of Bangladesh, practiced by indigenous communities. It entails using slash-and-burn methods to remove a patch of vegetation, producing various crops there for one or two seasons, and then shifting to another location (Biswas et al., 2012). Traditional shifting farming comprises a longer fallow season, but it has been replaced by shorter fallow periods as a result of rising population pressure. This process is vivid in the VHSR satellite images (Figs. 5 and 6). Shifting cultivation was identified to be the primary cause of the conversion of a considerable forest area in the Chittagong Hill tracts into either a degraded forest or agricultural land (Giri & Shrestha, 1996). Rahman et al. (2011) have explored the main reason for shifting farming in the Chittagong Hill Tracts area. Other factors contributing to the region's changing forest cover include excessive wood extraction, encroachment, settlement, urbanization, and poor management techniques (Mamnun and Hossen, 2020).

Accuracy assessment

The accuracy of Sentinel-2 image classification both for the historical and recent images is reasonable; the overall accuracy of 2016 and 2020 image classification are 91.67% and 94.67%, respectively. The forest class of 2016 shows the producers and users accuracies are 89% and 97%, respectively, while these accuracy statistics for the 2020 classification are 96% and 97%, respectively (Table 4). Overall, Kappa Statistics for 2016 and 2020 satellite images are 0.8642 and 0.9148, respectively. The Kappa Statistics of forest class for 2016 and 2020 satellite image classification are 0.9277 and 0.9328, respectively.

Table 4

Accuracy statistics of the classified satellite images (2016 and 2020)
Class name	Classified Sentinel-2 satellite image (2016)		Classified Sentinel-2 satellite image (2020)
Class name	Producers Accuracy	Users Accuracy	Producers Accuracy	Users Accuracy
Forest	89.17%	96.55%	95.89%	96.55%
Shrubs	97.98%	84.35%	100.00%	90.35%
Bare Land	64.71%	100.00%	57.14%	100.00%
Shadow	100.00%	89.47%	95.00%	100.00%
Water	100.00%	100.00%	100.00%	100.00%

Conserving the remaining forest

To conserve the remaining forest, BFD should consider the following recommendations: (i) the place is concerned to devote adequate staff in the remote SRF area due to insurgency and availability of officials; the organization should carefully revisit its policy to issue passes and allowances of permit to transport forest products to the downstream area (ii) The SRF should be brought under a continuous satellite monitoring system by BFD and (iii) Steps should also be taken to accelerate the artificial regeneration in the study area in particular to the vacant spaces and shrubs dominated areas. Finally, the Bangladesh Government should take the initiative to settle down the shifting cultivators so that there would be minimum disturbance to the forest by this practice. The above measures will be helpful to conserve the remaining forest resources in the region.

Limitations of the study

Several limitations have been identified in the study; the limitation was related to the spatial resolution of satellite images, in particular to the fragmented landscape. Sentinel-2 satellite data have a spatial resolution of up to 10m for the selected bands. In the fragmented landscape, this spatial resolution creates complicacy in the classification procedure because of the presence of many mixed pixels. Another limitation of classification is related to the topographic condition as the area contains undulating terrain. Sun-facing slopes show brighter reflectance and the slopes opposite to the sun have darker reflectance. The pixels in the complete shadow zones appearing in a dark tone contain no LULC information.

Field verification and validation could not be implemented in the region due to security reasons. The study area was located in the south-eastern corner of the country with the border of Myanmar where access is limited, and the region is inhabited by the tribal people. Therefore, validation was done using Sentinel-2 unclassified images, VHSR images of Google Earth, and the WorldView-2 satellite images of the selected sampled area.

In this study, forest cover maps of 2016 and 2020 and a forest cover change map of 2016–2020, as well as the relevant statistics, have been generated. The dominant stable class in LULC change analysis are forest (46.4 thousand ha 37.5%) and shrubs (45.7 thousand ha, 37%). The largest conversion class is a forest to shrubs or bare land (17.6 thousand ha, 14.3%) from 2016 to 2020. In contrast, the area representing shrubs to forest or bare land to the forest is only about 3.8 thousand ha (3%). The main driver of the forest cover change process is shifting cultivation and by this process, many areas of Chittagong Hill Tracts are converted to the class dominated by the shrubs. BFD should keep the forest in this region under a continuous satellite monitoring system.

The limitations of the study are related to the availability of suitable satellite images with finer spatial resolution and the presence of mixed pixels on satellite images that hinder the accurate mapping of forest cover. The complication in classification also originated because of the topographic condition of the study area. Further studies should be conducted to overcome these limitations.

Acknowledgement

The Sentinel-2 satellite image was downloaded from the United States Geological Survey (USGS) website and the data was available through the European Commission’s Copernicus program. The funding of this study was obtained from the National River Conservation Commission, a statutory institution of the Bangladesh Government. The contributions of Zebunnesa Khatoon and Sultana Razia of Bangladesh Space Research and Remote Sensing Organization (SPARRSO) are acknowledged for manually identifying the land clearing spots on Sentinel-2 satellite image and digitizing Sangu River in the study area, respectively.

Conflict of interest

The authors declare that they have no conflict of interest.

Abdullah, H. M., Mahboob, M. G., Rahman, M. M., & Ahmed, T. (2015). Monitoring natural Sal forest cover in Modhupur, Bangladesh using temporal Landsat imagery during 1972–2015. Int. J. Environ, 5(1), 1–7.
Biswas, A., Alamgir, M., Haque, S. M. S., & Osman, K. T. (2012). Study on soils under shifting cultivation and other land use categories in Chittagong Hill Tracts, Bangladesh. Journal of Forestry Research, 23, 261–265.
Chaudhury, B., & Chaudhury, M. U. (1970). Working Plan of Sangu-Matamuhuri Reserved Forests for the period from 1967-68 to 1986-87.
Chowdhury, M. S., & Hafsa, B. (2022). Multi-decadal land cover change analysis over Sundarbans Mangrove Forest of Bangladesh: A GIS and remote sensing based approach. Global Ecology and Conservation, 37, e02151.
Emch, M., & Peterson, M. (2006). Mangrove forest cover change in the Bangladesh Sundarbans from 1989-2000: A remote sensing approach. Geocarto International, 21(1), 5–12.
FAO. (2020). Global Forest Resources Assessment 2020—Key Findings. Int J Marine Coast Law.
Fortin, J. A., Cardille, J. A., & Perez, E. (2020). Multi-sensor detection of forest-cover change across 45 years in Mato Grosso, Brazil. Remote Sensing of Environment, 238, 111266. https://doi.org/https://doi.org/10.1016/j.rse.2019.111266
Giri, C., Pengra, B., Zhu, Z., Singh, A., & Tieszen, L. L. (2007). Monitoring mangrove forest dynamics of the Sundarbans in Bangladesh and India using multi-temporal satellite data from 1973 to 2000. Estuarine, Coastal and Shelf Science, 73(1–2), 91–100.
Giri, C., & Shrestha, S. (1996). Land cover mapping and monitoring from NOAA AVHRR data in Bangladesh. International Journal of Remote Sensing, 17(14), 2749–2759.
Hansen, M. C., & DeFries, R. S. (2004). Detecting long-term global forest change using continuous fields of tree-cover maps from 8-km advanced very high-resolution radiometer (AVHRR) data for the years 1982–99. Ecosystems, 7, 695–716.
Hansen, M. C., Roy, D. P., Lindquist, E., Adusei, B., Justice, C. O., & Altstatt, A. (2008). A method for integrating MODIS and Landsat data for systematic monitoring of forest cover and change in the Congo Basin. Remote Sensing of Environment, 112(5), 2495–2513. https://doi.org/10.1016/j.rse.2007.11.012
Hasan, M. E., Nath, B., Sarker, A. H. M., Wang, Z., Zhang, L., Yang, X., Nobi, M. N., Røskaft, E., Chivers, D. J., & Suza, M. (2020). Applying multi-temporal Landsat satellite data and markov-cellular automata to predict forest cover change and forest degradation of Sundarban reserve forest, Bangladesh. Forests, 11(9), 1016.
Islam, M. J., Alam, M. S., & Elahi, K. M. (1997). Remote sensing for change detection in the Sunderbands, Bangladesh. Geocarto International, 12(3), 91–100.
Khalile, L., Rhinane, H., Kaoukaya, A., & Lahlaoi, H. (2018). Forest cover monitoring and change detection in nfifikh forest (Morocco). Journal of Geographic Information System, 10(02), 219.
Mamnun, M., & Hossen, S. (2020). Spatio-temporal analysis of land cover changes in the evergreen and semi-evergreen rainforests: A case study in Chittagong Hill Tracts, Bangladesh. Int. J. For. Ecol. Environ, 2, 87–99.
Masum, K. M., & Hasan, M. M. (2020). Assessment of Land Cover Changes from Protected Forest Areas of Satchari National Park in Bangladesh and Implications for Conservation. Journal of Forest and Environmental Science, 36(3), 199. https://doi.org/10.7747/JFES.2020.36.3.199
Mihai, B., Săvulescu, I., Rujoiu-Mare, M., & Nistor, C. (2017). Recent forest cover changes (2002–2015) in the Southern Carpathians: A case study of the Iezer Mountains, Romania. Science of the Total Environment, 599, 2166–2174.
Nath, B., & Acharjee, S. (2013). Forest cover change detection using normalized difference vegetation index (NDVI): a study of Reingkhyongkine lake’s adjoining areas, Rangamati, Bangladesh. Indian Cartogr, 33(2), 348–403.
Nguyen-Trong, K., & Tran-Xuan, H. (2022). Coastal forest cover change detection using satellite images and convolutional neural networks in Vietnam. IAES International Journal of Artificial Intelligence, 11(3), 930.
Panigrahy, R. K., Kale, M. P., Dutta, U., Mishra, A., Banerjee, B., & Singh, S. (2010). Forest cover change detection of Western Ghats of Maharashtra using satellite remote sensing based visual interpretation technique. Current Science, 657–664.
Rahman, S. A., Rahman, M. F., & Sunderland, T. (2012). Causes and consequences of shifting cultivation and its alternative in the hill tracts of eastern Bangladesh. Agroforestry Systems, 84, 141–155.
Redowan, M., Akter, S., & Islam, N. (2014). Analysis of forest cover change at Khadimnagar National Park, Sylhet, Bangladesh, using Landsat TM and GIS data. Journal of Forestry Research, 25(2), 393–400. https://doi.org/10.1007/s11676-014-0467-9
Roberts, J. F., Mwangi, R., Mukabi, F., Njui, J., Nzioka, K., Ndambiri, J. K., Bispo, P. C., Espirito-Santo, F. D. B., Gou, Y., & Johnson, S. C. M. (2022). Pyeo: A Python package for near-real-time forest cover change detection from Earth observation using machine learning. Computers & Geosciences, 167, 105192.
Samrat, N. H., Islam, N., & Haque, A. (2022). Forest Land Cover Changes and Its Socio-economic Consequences on South-eastern Part of Bangladesh. Journal of Sustainable Forestry, 1–19.
Sourav, M. S. H., & Anwar, K. Bin. (2020). Arenga westerhoutii Griff. and Livistona jenkinsiana Griff., two new species of wild palms for Bangladesh. Trop Plant Res, 7(2), 374–378.
Sourav, M. S. H., Anwar, K. Bin, & Sultana, M. (2020). Rediscovery of Dendrobium angulatum Lindl.(Orchidaceae), from Bangladesh after 198 years. Tropical Plant Research, 7(2), 326–330.
Zaman, S., & Katoh, M. (2011). Assessment of Forest Covers Change in Tropical Moist Deciduous Forest in Thakurgaon, Bangladesh Using ALOS Data (Special Issue, Multipurpose Forest Management). Journal of Forest Planning, 16(Special_Issue), 285–292. https://doi.org/10.20659/jfp.16.special_issue_285

Download PDF

Editorial decision: Rejection
20 Mar, 2024
Reviewers agreed at journal
23 Jan, 2024
Reviewers invited by journal
25 Aug, 2023
Editor assigned by journal
08 Aug, 2023
First submitted to journal
04 Aug, 2023

You are reading this latest preprint version

Forest Cover Change Assessment in the Sangu River Basin Using Sentinel-2 Satellite Data

Status:

Version 1

Abstract

Figures

Introduction

Material and methods

Study Area

Dataset

Satellite image collection

Satellite image interpretation and classification

Accuracy assessment

Results and discussion

Forest cover change assessment

Forest cover change processes

Accuracy assessment

Conserving the remaining forest

Limitations of the study

Conclusion

Declarations

References

Status:

Version 1