Urban growth assessment in the Northeastern region of Bangladesh for sustainable landscape management and conservation

ABSTRACT Land use and land cover change are regarded as vital ingredients to evaluate suitable and sustainable forecasting tools for ecological balance on earth to planners and policymakers. This paper investigated land-use changes in the Northeastern region of Bangladesh, a region with diverse landscapes and bio-cultural heritage, using Landsat images from 1987 to 2017 with 10 years interval. Our results showed that 242.79% expansion of built-up lands and 39.65% reduction of tree cover, which was alarming and quite dominant between 1987 and 2017. Over the last three decades, 18.38% of non-tree vegetation increased while water, seasonal cropland, and bare land were decreased by 14.67%, 11.46%, and 35.75%, respectively, which indicates that the loss of cropland is linked with expansion of urban growth or built-up area. For the sustainability of concurrent natural ecosystem and proper conservation of its remnant biodiversity, a long-term land management plan should be undertaken demarking clear land zoning with full involvement of the government (i.e., Ministry of Environment, Forest, and Climate Change) and allied non-government agencies.


Introduction
Land resources are one of the vital resources for human lives.For the rapid expansion of urban areas, it is important to sustainably govern the land resources to conserve biodiversity (Huang et al., 2018) as well as mitigate local and regional climate change (Hirsch et al., 2018).Effective policy implementation such as rationally use land resources is important to control urban expansion (Huang et al., 2018;Long et al., 2014).Besides, to keep the international commitment of UNFCCC (United Nations Framework Convention on Climate Change, 2015) to limit global warming to 2°C, land-based mitigation strategy should be implemented through reforestation (Popp et al., 2017).But now a days, land is facing a variety of demands due to immense agricultural, rapid settlement, and other anthropogenic activities by increased population (Hossain Bhuiyan et al., 2019;Zdruli, 2014).Therefore, to understand the land cover and land use processes is a very significant topic to be studied.Land use and land cover (LULC) are two different phenomena, but they often substitute each other (Islam et al., 2018;Dimyati et al., 1996).Land use or land cover can be changed both naturally and by humans.Land use change, on the other hand, is the human-induced change for their different interests.Due to huge augmentation in world population, urban and commercial development, quick progression in industrial and agricultural activities all over the world, land use and land cover had been changed rapidly over the past half century; wetlands, forest agricultural lands, arable, and developed lands had undergone a huge change (Chen et al., 2014).The high growth rate of the population is one of the primary factors of this land conversion process (Hossain Bhuiyan et al., 2019;Roy et al., 2015).It is immensely significant to have continuous, historical, and proper up to date information related to land use land cover change (LULCC) to develop a plan of an area.Precise knowledge on LULCC and its change upon a time are highly important for any kind of sustainable development program where LULCC can be a major input criterion (Abd El-Kawy et al., 2011).
The Land Use and Land Cover Change (LULCC) is a progressive process by nature (Mondal et al., 2016).So, an extensive study on LULCC and its pattern, as well as their social and environmental impact, are necessary at different temporal and spatial scales (López et al., 2001).There are several articles that emphasise studies on different LULCC are important to monitor natural resources like forests or the overall environment (Humayun-Bin-Akram & Masum, 2020;Lal & Anouncia, 2015).According to Iqbal and Khan (2014) and Kantakumar and Neelamsetti (2015) LULCC has very significant impacts in natural resource management.Islam et al. (2018) also highlighted that land use and land cover change are a very important issue which is often considered as a major factor for environmental change.As it is an important topic worldwide, some international organizations always keep their eye on it.IGBP (The International Geosphere-Biosphere Program) and IHDP (International Human Dimension Program) initiated an international program together to detect land use and land cover change and their impacts in various dimensions (Irwin & Geoghegan, 2001).
As land use and land cover change (LULCC) is an ever changing and important process (Mondal et al., 2016), many studies have been conducted worldwide over the last few decades in global or regional stages focusing land conversion.According to FAO (2015), largest amount of forest land has been converted into other land classes in the tropical region over the last fifteen years.From the liberation war in 1971, land use pattern in Bangladesh is changing in a significant pace (Al-Amin, 2018).A huge population lives in a small area in Bangladesh (1063 people per square km; BBS.2016) which triggers to destroy natural resources; having only 12.5 decimal lands per capita (Quasem, 2011).Approximately 1% of total agricultural lands in Bangladesh is converting to different types of infrastructures (Sayed & Haruyama, 2015).Due to shrimp firm and salt bed, some mangrove forest area had lost in Chakaria Sundarbans during 1989-1999 (Musa, 2008).Evidence was found for reduction in agriculture and increment of wetland nearly 50% and 500% respectively (G.T. Islam et al., 2015).
Northeastern region of Bangladesh is persistently important area as it is endowed with subtropical moist forest, wetland of international importance and mostly all tea gardens of the country.However, humaninduced load and economic development in last three decades are posing threat to its home garden, forest, tea gardens and associated biodiversity.Gupta et al. (2018) indicated that unplanned urbanization produces pressure on the natural functioning of land and resources and develops rapidly after 2011.In addition, transformation of land cover into anthropogenetic land uses has an enormous consequence on the ecological environments of plant and animal species of this area which is a warning for the sustaining of this brittle ecosystem (Islam et al., 2016).Studies on the forest covers change in Khadimnagar National Park, by Redowan et al. (2014) reported that the dense forest cover declined from 526 hector (ha) to 417 ha whilst medium dense forest increased from 155 ha to 317 ha and bare land decreased to 8 ha from 104 ha during 1987 to 2010.Hasan et al. (2013) also reported a decreasing trend in cropping land during 1976-2010 in Sylhet division at rate of 0.33% per year.Though several researchers have demonstrated in their study towards forest monitoring but still it has remained an interesting task to know about LULCC in the Northeastern region in details for last three decades.Thus, the study has been conducted in northeastern region of Bangladesh for assessing the trend of land use and land cover since 1987 as well as examining trend of temporal changes in different land cover types (tree, non-tree vegetation, seasonal cropland, built up land, bare land, and water body).Through this study the overall scenario of this region has been evaluated and also future scenario has been forecasted if any immediate actions are not undertaken.Finally, the study will aid in making a sustainable plan for protecting this bio-cultural heritage landscape by implementing land zoning rules of land use policy.
Moreover, it will be helpful for the policy maker in developing a practical management plan for the remnant forest and tea garden.

Study area
Northeastern regions in Bangladesh have been selected for this study for its rich bio-cultural inheritance and landscape diversity.The region consists of Sylhet and Moulvibazar districts in Sylhet division (Figure 1) .Study area is flanked by the Indian states of Meghalaya in the north, Assam in the east, Tripura in the south and the Bangladesh district of Sunamgonj and Habiganj in the other sides (Rashid, 1991).The land of the study area is covered by different types of land such as forest, homegarden vegetation, waterbody, low land, developed area, bare soil, etc.There are some hilly regions in Sylhet and Moulvibazar district which are covered by tea estate and thin forest.In Sylhet, it is predominantly hot and humid in summer, tropical monsoon and relatively cool in winter with common maximum temperature 30.7°C and average minimal temperature 18.9°C (BBS, 2020;Rashid, 1991), though it is unpredictable in different times of the year.May to September is the prime time for precipitation in Sylhet, in this period almost 80% of annual average rainfall occurs.(M.Islam et al., 2019).The annual average precipitation of this region is 4195.9mm.

LULCC analysis for 3 decades
Landsat satellite data were acquired for four different years of the study area the years are 1987, 1997, 2007, and 2017 from GloVis.Shapefile of the Sylhet division was collected from Bangladesh Space Research and Remote Sensing Organization (SPARRSO) (Table 1).All the districts were separated by boundaries on that shapefile.By using the ArcMap toolbar, the two districts were separated and merged to form a complete shapefile.The shapefile was previously geo-rectified, so further geo-rectification was not applied.As ERDAS Imagine requires Area of Interest (AOI) file instead of shapefile, the shapefile (.shp) was converted to the AOI (.aoi) file for further use.As the percentage of cloud and haze was very little, a simple radiometric correction and haze reduction tool was applied.Shortwave Infrared-2 (SWIR-2) which is the 7th band of Landsat 5, 7 and 8 was used to extract the water body.The band combination was 7(SWIR-2), 4 (NIR), 2 (Green) for the 1987, 1997, and 2007 images (Landsat-5, Landsat-5, and Landsat 7) and 7(SWIR-2), 5 (NIR), 3 (Green) for the 2017 image (Landsat-8).After combining the image an unsupervised ISO Data classification was performed on this image to extract water.In this ISO Data classification, 55 classes were set which was suitable for the intended number of classes (Lo & Choi, 2004).The water bodies were extracted from it by the mask extraction method (Land cover 1).Other pixels rather than water were left as it was previously as well.Another image was prepared for extracting other land cover classes (rather than water) by using False Color Composite (FCC).Some clusters of ISO data clustering were pure, and some were mixed.All the homogenous clusters that only coincided with a specific land cover type out of six were retained and labeled (Land cover 2).Some clusters were showing the characteristics of mixed pixels, they had been masked out from the image by the mask extraction method.So only pure pixels made an incomplete land use land cover map (Land Cover 2).The masked-out clusters from the previous stages were belonging twi or more classes at the same time.To deal with this problem, class grouping tools were used which makes the image ready to be recorded with the fuzzy classification method.In this stage, all the mixed clusters were assigned into 6 per-defined classes though, there was no cluster belongs to the water class (because all water pixels were masked out already) (Table 2).The fuzzy recoding method was performed on the previously grouped image to assign pixels of mixed characteristics.The neighbor weight option was set to 0 while performing the process.Though all clusters were used to computation, this process was only applied to ambiguous classes.This process recorded all the mixed pixels through a sophisticated algorithm to 6 predefined classes and thus prepare another land cover map (Land cover 3).The overlapping (combination of land cover 1, 2, and 3) was performed by using the union overlay process to create a complete land use land cover map.In this overlapping process, the waterbody, and the pure clusters and the fuzzy recoded clusters were overlapped accordingly.Area calculation is done by using this formula: Histogram*0.09/100 in Erdas imagine 2014 which provides values for each category class in square kilometer unit.Stratified random sampling is used to evaluate the accuracy of the classified image of the years 1987, 1997, 2007 and 2017 as per the recommendation for such kind of accuracy assessment by (Congalton, 1991).

Future trend analysis for LULCC
Values of future change for coming two decades (tenyear interval) of 2027 and 2037 have been forecasted.A simple projection was made based on the assumption that the rate of LULCC will continue at the average rate seen between 1987 and 2017.Trend lines have been made using EXCEL forecasting function where a simple statistical relationship between the dependent variable, Y (area) and the independent variable, X (decade) were established.It follows the following linear equation, y = a + bx for forecasting; where -a ¼ � y À b� x and b ¼

LULCC analysis with accuracy assessment
Image accuracy was found at 81.66% (Average Accuracy) which was quite satisfactory.The accuracy validates the visual interpretation as well as the validity of this study overall.The classification accuracy for different years image was 81.33%, 84%, 83%, and 78.33% (1987, 1997, 2007, and 2017).Change detection encompasses the practice of multi-data images or aerial photos to assess alterations in land cover due to environmental conditions and human actions between the acquisition dates of images.Based on the analysis of satellite imagery, all the pixels in the image were classified into six groups including Trees, Non-tree vegetation, Seasonal croplands, water bodies, built-up areas, and bare lands.The classified LULCC map for the years 1987, 1997, 2007, and 2017 are shown in Figure 2 and their respective occupied areas in Supplementary Table 13.The result derived from the classified satellite image of four different periods indicates that there is rapid urban growth took place, especially the built-up areas that had a huge expansion over 29 years.The built-up areas have grown more rapidly in recent decades comparing with the previous two.Supplementary Table 13 shows that as of 2017, built-up areas occupy a land area of 564.196Km 2 which is as same as 6.51% of the total study area.In 1987, the total built-up land was 164.59 Km 2 it became 219.63 Km 2 after having an expansion of about 55 square kilometers (5.50 Km 2 /year) within a decade (Supplementary Table 13, Table 3).The expansion of built-up areas was more rapid with time.Within the next decade (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007), the built-up land expanded by 120.2 square kilometers (15.16 Km 2 /year) and occupied a total 339.83 km 2 land area.An expansion of 224.366Km 2 has been noted over the period 2007 to 2017 (22.43 Km 2 /year).During 1987 and 2017, built up grown at a pace of 13.32 Km 2 /year (Table 3).

Trend analysis
According to the projection, tree cover will be in real threat (Figure 3) .By 2027 total tree cover will be only 576.25 Km 2 and by 2037 the cover will be only 422.46 Km 2 .Unlike tree cover, the non-tree vegetation would be 2659.95Km 2 and 2794.3Km 2 by the years 2027 and 2037, respectively.Forecasted values also revealed that water and bare land will be decreased by 353.125 Km 2 and 68.48 Km 2 by 2027 and by 2037, they will be 326.746Km 2 and 60.44, respectively.If the decreasing trend continues, the seasonal cropland will be decreased to 1825.73 Km 2 and by 2027 and 2037 respectively as well.Built-up areas have a huge expansion since 1987.As per the increasing trend, by 2027 and by 2037 the amount of area acquired by built up area will be 651.81Km 2 and 783.71Km 2 as well.

Discussion
The important aspect of this study is to assess what is essentially changing to what, i.e., which land-use class is changing and expanding.From the initial recognition of the analysis, the land cover in the study area is dynamic and changes over time.According to the findings, a huge expansion (399.61Km 2 or 2.36% per year) of built-up areas has been identified (Supplementary Table 14, Table 3).This result supports the earlier findings that the Sylhet district has been expanded extensively due to accommodate the growing population (Masum and Hasan, 2020, Rai et al., 2017and Haque et al., 2008b).According to Masum and Hasan (2020) Rai et al. (2017), the Sylhet division had a huge built-up growth from 1987 to 2007 replacing forest areas, agricultural land as well as water bodies.The Urban growth rate in Sylhet city was very high among all the major cities in Bangladesh (Hassan, 2017).In the study area, tree cover was reduced by 467.37 over the period 1987 to 2017 (Supplementary Table 14, Figure 4) .These changes probably due to the increasing anthropogenic pressure on both home garden and natural forests.In this time interval, a decrement of 15,899.04hector forest cover had been also reported by Haque et al. (2008b) in Sylhet.A study by Hasan et al. (2013a) demonstrated that total forest cover in Bangladesh had a significant loss between 1976 and 2010 whereas FAO argued that the forest cover did not change significantly in Bangladesh between 1990 and 2000.Again, Halim et al. (2008) found about 133 hectares increment in other vegetation types (other than the tree) between 1996 and 2006 in Bhanugach reserve forest.On the other hand, this study finds that non-tree vegetation occupied a total of about 35% in 1987 of the total study areas which occupied 42% in 2017.So, it has increased its size by a total of 397 Km2 (between 1987 and2017).This expansion may be caused by forest degradation or conversion to agriculture/non-tree vegetation or settlement.The rate of waterbody declined in the study In another assessment carried out in the Sylhet division exhibited that waterbody has decreased significantly, especially the haors had decreased by 14% between 1989 and 2010 due to conversion to mudflat and cropland (Ali, 2006;Salauddin & Islam, 2011).However, the report exposed an increment of   et al. (2016).This could be the fact that a small increment of cropland in some areas might have not a significant effect on the total area.Another outcome stated that the cropland area in the Sylhet division had decreased at a rate of 0.39% each year from 1976 to 2010 (Hasan et al., 2013a).In contrast, to support the growing population, their increasing food demand, and shifting cultivation, some small portions of land was converted to cropland area somewhere in Sylhet (Rahman et al., 2016;Roy et al., 2015;Shrestha et al., 2016).Regarding Bare lands, it has been decreased by 35.141 Km2 (0.06% each year) within the past 3 decades.Conversely, there was an abnormal increment of this land type is seen in the year 2007.These discrepancies might have resulted from the changes in vegetation cover as well as the shrinkage of waterbody at that time.This phenomenon supports the view of Halim et al., (2008)  Otherwise, increasing demand for tea production will be hampered, ecological balance of the forest will be vulnerable, home garden will lose its productivity, and sustainable land use will become a jingle only.The most dangerous and grave situation is the conversion of hilly land for urban development which leads to frequent earthquake experienced by the dwellers in this region.Bare and unproductive land can be taken under consideration for future urban expansion.Now a day's vertical expansion is also an effective way of reducing spatial load urban development for the increasing populations.

Figure 1 .
Figure 1.Map of the study area.

Figure 3 .
Figure 3. Projected net change between 1987 and 2037 for each of the major land use.

Figure 4 .
Figure 4. Area occupied by different land use types.

Table 1 .
Details of acquired satellite data.
ConclusionUsing Landsat images, the dynamics of land use land cover changes are addressed.Analyzing the data, reduction in the tree, waterbody, seasonal cropland, and bare land have found.The tree cover and seasonal croplands are significantly decreasing compared to other land use land cover classes which is posing threat to the remnant forest, home garden and tea garden.Increasing Built-up areas are mainly occupying those areas.Results show that the areas near the city, town, and suburbs are converting rapidly.Therefore, it can be inferred that the dynamic factors such as population density and proximity to urban areas influence land use land cover changes in this area.Another interesting feature from this study is that the change is very high in recent past years comparing to the beginning years of the study.It will be very tough for us to regain our ecological balance if policymakers do not make their plan now against this rapid change or to avoid environmental problems like sedimentation, flooding, and declining to carrying capacity of the study area.So, it is vital for the government to take obligatory steps to acquire optimal land use that can eliminate land use conflict.According to the National Land Use Policy a proper land zoning should be finalized immediately by the Land Ministry to halt any unplanned urban expansion in this region forming a body including Forest Department, Environment Department, Industrial Department, and Land Department along law enforcing agencies.
* no sign -increase in area | minus sign -decrease in area equation.3:year2 eg:1997

Table 2 .
Selection criteria for specific classes.