Spatiotemporal changes of vegetation and land surface temperature in the refugee camps and its surrounding areas of Bangladesh after the Rohingya influx from Myanmar

Abstract

In August 24, 2017, a massive outbreak took place in the Rakhine state of Myanmar which triggered a huge refugee influx to the Teknaf Peninsula, Bangladesh. To settle the refugees, makeshift camps were built in large numbers destroying huge amount of forest areas near the existing Kutupalong and Nayapara camps. Refugees have been encroaching the nearby forest covers to collect fuelwood and other purposes. These forest destructions have put the wildlife and biodiversity of the system in a substantial pressure as well as altering the land surface temperature (LST). This paper has examined the extent of vegetation change and the changes of LST from 2017 to 2019 throughout Kutupalong and Balukhali camp and adjacent areas using Landsat 8 images. Random forest algorithm and Plank equation were applied on images to identify vegetation change and LST, respectively. The overall and kappa accuracies for the maps of 2017 are 96% and 92%, respectively, while it stands at 94% and 88% for the 2019 image. Results derived from the analysis suggest that an estimated 1876 hectares of forested lands have been decreased in the study area. LST of the study area increased spatially throughout the whole region with a maximum value of 34 °C which is significantly higher than the pre-influx period. If this trend of forest-clearing activities continues, the place will become barren land soon and the LST will also increase. All these factors will ultimately trigger the climate change impacts and biodiversity loss of the area.

Introduction

High density of population in an area undergoing through uncontrolled situations often causes over exploitation of resources (Dabelko et al. 2002; Berry 2008; Bremner et al. 2010; Mohammed et al. 2018). Following the massive violence in Rakhine province, Myanmar, nearly 700,000 Rohingyas (an ethnic minority of Myanmar) have fled their homes since August 2017 risking their lives by sea or on foot with horrific memories of violence, demolition of homes, burning of villages and loss of loved ones (BBC News 2018). The Rohingyas have continued to come across the border in the Teknaf region of Bangladesh since the early 1990s, mostly seeking relative safety. As of now, the Rohingya refugees and asylum seekers have outnumbered the local people of Ukhia and Teknaf regions of Cox’s Bazar district (Imtiaz 2018) of Bangladesh. New makeshift camps were built clearing forested lands and vegetations in large number near the existing refugee settlements of Kutupalong and Nayapara to inhabit most of the people during the influx of August 2017 (UNHCR 2019). The protected forests and wildlife habitat have marked endangered considering the extreme pace and scale of refugee inflow in that particular area (Hassan et al. 2018). The high density of population living in such close proximity has a huge toll on the environment and biodiversity of the surrounding regions as well. The region where the Rohingya refugees are mostly concentrated in makeshift camp is a protected forest for wild animals which are considered endangered. It is an extremely delicate ecological area due to the massive pressure that has been created on the natural resources which eventually altered the local environment (Hassan et al. 2018). The Forest Department of Cox’s Bazar reported that an estimated 5013 acres of forestland has been encroached by the Rohingyas while this number is increasing day by day (Hussain 2018). Rohingya families use 6800 tonnes of fuelwood each month along with 60 culms of bamboo for construction purpose, which are collected directly from the nearby forests posing a great risk of landslides as the terrain of the hills losing their natural settings (UNDP Bangladesh and UN Women Bangladesh 2018). Besides, deforestation can negatively affect the water balance and composition of soil, resulting in increased soil erosion (Glade 2003; Ghimire et al. 2013). If this current trend of forest-clearing continues, soon the vegetated areas will likely to turn into barren land and the wildlife habitats will be in grave danger. Hence, there will be serious consequences of forest clearing in the camp areas within a 15-km radius, while ground water pollution, shortage of firewood, overdraft of groundwater and desertification can be some of them (Ghimire 1994). Furthermore, vegetation loss at this unprecedented rate can significantly alter regional climate. Wang and Myint (2016) found that deforestation can reduce the evapotranspiration rate and increase land surface temperature (LST). Tropical deforestation dramatically alters temperature, relative humidity, wind and precipitation levels of local areas (Sud et al. 1996).

Remote sensing is a powerful and most useful technique for identification of spatial and temporal changes of vegetation cover and local temperature with historical images. Numerous assessments of vegetation loss, crop stress detection and studies of other environmental phenomena have been performed with the use of satellite images (Skole and Tucker 1993) with less time, low cost and with better accuracy (Kachhwala 1985). For instance, Langer et al. (2015) used pre-processing technique of satellite time series images of different Landsat sensors covering the whole lifespan of the Camp Lukole, Tanzania, and observed long-term degradation of natural vegetation around the camps. Braun et al. (2016) used ERS-2 and Sentinel-1 for an assessment of landscape changes in Semi-arid Savannah, Western Kenya, and noticed an overall decrease in natural resources by 11.8% between 1997 and 2014 along with an increase in bare rocks and soil areas around the camps. Mantey et al. (2014) revealed a correlation between Normalized Difference Vegetation Index and LST of Accra city, Kenya, using Landsat images and found that LST values were relatively high in built-up areas where natural vegetation is absent and vice versa. Classification of satellite images has been a crucial task during land cover monitoring and change detection. Algorithms with better precisions are consistently being developed in the field of remote sensing. Random forest algorithm is quite popular these days for providing more accurate results at faster speed. Many studies have applied random forest for classifying multispectral satellite imageries (Gounaridis et al. 2016; Jhonnerie et al. 2015; Karlson et al. 2015; Immitzer et al. 2012; Sesnie et al. 2008). It is more desirable due to less user-defined parameters, outliers’ detection and not overfitting (Gislason et al. 2006; Pal 2005). Comparing with other classification algorithms, RF always performed well (Thanh Noi and Kappas 2017; Jhonnerie et al. 2015; Gislason et al. 2006). Rodriguez-Galiano and Chica-Olmo (2012) used RF in land cover analysis over Landsat data along with land surface temperature, digital terrain model and textural images as auxiliary data which led to more accurate results.

LST is an important phenomenon of regional climate, and excess heat might have negative consequences on the local inhabitants and on the environment, increasing various health risks and energy consumptions (Adiguzel et al. 2020; Cetin 2015, 2016; Poumadere et al. 2005). People’s mental and physical health is closely connected with climatic and environmental conditions; therefore, their well-being and happiness is related to it (Cetin et al. 2018; Li et al. 2012). Increased LST might also alter the biotic habitat of that particular region (White et al. 2002). Therefore, to know the change of LST with the alternation of vegetation loss is important for the study area. Although a few attempts have been made to assess the extent of vegetation loss in Teknaf Peninsula due to Rohingya influx (Braun et al. 2019; Imtiaz 2018; Hassan et al. 2018), no single study exists which tried to identify the land surface temperature (LST) change in response to vegetation loss. To fulfill this research gap, this paper has attempted to estimate the vegetation loss till summer 2019 and to detect the land surface temperature (LST) change with the response to vegetation loss. Therefore, the aim of this paper is to find out the vegetation loss and changes of LST after the recent influx. The research also answers the questions of suitability assessment of RS techniques to determine the environmental impacts due to refugee influx. It is also expected that the findings will contribute in raising global voice to stop this sort of displacement of people from their origin, destroying forest and natural habitats and exaggerating the climate change impacts in the least developed countries.

Materials and methodology

Study area

The location of the study area is situated in the southernmost part of Bangladesh bordering Myanmar. It is located between latitudes 21° 13′ N and 21° 70′ N and longitudes 92° 05′ E and 92° 11′ E. Map of the study area is shown in Fig. 1. Ukhia and Inani forest range of Cox’s Bazar district is within the study area. The topography of these forested areas is diverse with medium to almost flat hillocks. The entire study area has low hills of not more than 100-m-height from the mean sea level. Soils are mostly sandy to sandy loam in high elevated areas, whereas it is clay to silty clay in the lower areas. The climatic condition is monsoonal with an average annual rainfall of 4411 mm. The temperature ranges from 14.9 to 27 °C with an average relative humidity of 75% (BBS 2011). Kutupalong refugee camp that exists in this region since 1991 became hugely populated with the recent influx of 2017. The earlier population of the study area was around 100,000 to 440,000, and an additional 713,000 people were added following the recent breakdown to make it world’s largest refugee camp. Although there are several refugee camps across the whole Ukhia and Teknaf regions, Kutupalong–Balukhali holds majority of them. To inhabit these vast arrays of refugees, makeshift camps and other infrastructures were built, which led to destruction of forest covers in the adjacent areas

Fig. 1
figure1

Location of the study area including camps inside

Data

Data for this study were collected from the Earth Explorer website of the United States Geological Survey (USGS). Multispectral Landsat 08 Level 1TP data were collected of the year 2017 and 2019. Cloud free images were obtained from similar season in order to avoid variation in vegetation phenology. Landsat 08 collects 11 band images, including two thermal bands. For validating LST, air temperature data were collected from Bangladesh Meteorological Department (BMD) for 2017 and 2019. Table 1 illustrates the characteristics of satellite images and camp shape file.

Table 1 Characteristics of collected data

Methods

Image pre-processing

Landsat L1TP is geometrically corrected using ground control points and suitable for pixel-based change detection (USGS 2016). Radiometric calibrations and atmospheric corrections were done using Semi-automatic Classification Plugin in QGIS 3.4. Landsat images are provided as DN (digital numbers) values that have no physical significance (Yankovich et al. 2019). Bands must be converted into reflectance or radiance for digital processing. Thus, the OLI and TIRS bands have been transformed into top of atmosphere (TOA) reflectance and radiance, respectively. Following this, atmospheric correction using Dark Object Subtraction (DOS) method was applied to prepare images for further analysis.

Calculation of vegetation change

Procedures for image classification were carried out using QGIS 3.4. All the bands of each year images were clipped according to the study area. Bands 02 to 07 of two preceding year were stacked and false color composite (SWIR1–NIR–red) were applied to better visualize land features. Training samples of three distinctive classes were collected from both images. Classification plugin dzetsaka in QGIS contains machine learning classifiers like random forest, support vector machine, etc. Random forest was applied over both images for land use/land cover classification.

  1. (a)

    Random forest classifier

To date, multiple classification algorithms have been developed and implemented to classify images. Machine learning algorithms (e.g., random forest, boosting, etc.) are being popular for their better classification precision and robustness. Random forest, a nonparametric machine learning algorithm, can model a high-degree nonlinear relationship between the targeted and predicted variables (Breiman 2001; Ghosh et al. 2014). It is a combination of decision trees, and each tree provides a classification label, out of which mostly voted class is taken. It uses bagging or bootstrap aggregation to create multiple formations of subsets for obtaining variety of trees (Breiman 2001). In RF, one-third of the training samples are left out which is called out-of-bag (OOB) data, and these data are used to run an unbiased estimate of the classification error (Belgiu and Drăguţ 2016; Breiman 2001).

  1. (b)

    Post-classification comparison

This is one of the simplest methods for change detection by overlaying both classified image (Almutairi and Warner 2010). This computes “pixel to pixel” changes between classes of two images (Almutairi and Warner 2010; Raja et al. 2013). Both independently classified images were compared using SCP plugin in QGIS 3.4 to produce a change detection image. Produced images were then reclassified into five classes for convenience of this study.

Calculation of LST

Different authors have measured land surface temperature (LST) in a variety of ways (Zhang et al. 2006; Amiri et al. 2009; Li et al. 2013; Sahana et al. 2016; Wang et al. 2019). Many algorithms have been developed to calculate LST. The following procedures for measuring LST are done in RStudio with the help of raster (Hijmans 2019) and RStoolbox (Benjamin et al. 2019) packages. In this study, Planck Eq. (1) is used to perform LST calculation of both TIRS bands. Then, the averages were produced for final mapping.

$$T_{{\text{s}}} = \frac{{{\text{BT}}}}{{\left\{ {1 + \left[ {\frac{{\lambda .{\text{BT}}}}{\rho }} \right].\ln \varepsilon } \right\}}} - 273.15$$
(1)

where Ts is the land surface temperature (°C); BT is the at-sensor brightness temperature (K); λ is the wavelength of the emitted radiance; ρ is the (h * c/σ) = 1.438 * 10−2 mK; and ε is the land surface emissivity (LSE).

Planck equation is widely used for the derivation of LST from different thermal sensors (Avdan and Jovanovska 2016; Bharath et al. 2013; Liu and Zhang 2011; Qin et al. 2001). Unlike other algorithms (e.g., SCA, SWA), this method is easy to use and does not require atmospheric parameters (Avdan and Jovanovska 2016; Ndossi and Avdan 2016b). Ndossi and Avdan (2016a) calculated LST for different Landsat sensors and found that Planck function produces best results from Landsat 8 TIRS. In order to derive the land surface temperature, it is important to correct brightness temperature against LSE. Emissivity of a surface is the capacity of radiation of that surface with reference to an ideal blackbody and is determined by the terrestrial composition (Gondwe et al. 2018; Ndossi and Avdan 2016a). In that case, NDVI is of great use for the calculation of land surface emissivity (LSE). It is the difference between NIR and red bands divided by their sum. NDVI is used to get the vegetation proportion (Pv) from Eq. (2) (Carlson and Ripley, 1997).

$$P_{{\text{v}}} = \left[ {\frac{{\left( {{\text{NDVI}} - {\text{NDVI}}_{{\min}} } \right)}}{{\left( {{\text{NDVI}}_{{\max}} - {\text{NDVI}}_{{\min}} } \right)}}} \right]^{2}$$
(2)

where Pv is then applied in Eq. (3) by Sobrino et al. (2004) for LSE.

$$\varepsilon \, = 0.004P_{{\text{v}}} + 0.986$$
(3)

Results

To measure the extent of vegetations change and difference of LST, land cover maps of 2017 and 2019 have been studied. Figure 3 indicates land cover of both the images with 3 major classes, and Fig. 4 illustrates the changes of these land covers between both the years. A total of 5 classes are considered for the change detection map (vegetation to vegetation, vegetation to non-vegetation, non-vegetation to non-vegetation, non-vegetation to vegetation and wetlands to wetlands between 2017 and 2019. Figures 5 and 6 represent the comparison between LST of pre- and post-influx (from summer of 2017 to summer of 2019). The accuracy assessment and findings of the study have been discussed below.

Accuracy assessment and validation

It is known that random forest classification has a self-estimated accuracy assessment during the bagging process. RF classifier produced OOB error matrix which was used to calculate the accuracy of the land cover classes. For the years, we calculated overall and kappa assessment and achieved the percentage of correctly classified pixels between 97 and 99%. In addition to that, we also collected ground truth samples from Google Earth’s high-resolution imagery for closest dates available. Then, ground points were cross-validated with the land cover map using 90 stratified random sample points for each year. Estimated overall and kappa accuracies using the data were 96.1–97.5% and 94–96.1%, respectively. For further validation of LST, we collected air temperature data from the Bangladesh Meteorological Department (BMD) in Teknaf station, as there is no weather data collection station within the study area. The comparative data of both air temperature and LST of 2017 and 2019 are given in Table 2.

Table 2 Comparison of LST data with air temperature

As Table 2 suggests, the LST of both years of the study area has a close proximation with the air temperature of the neighboring Teknaf region. The difference in minimum value is 1.3 °C, and the maximum value is the same in LST and air temperature of 2017. In 2019, the difference in minimum temperature is 0.9 °C and 1.9 °C in maximum temperature.

Changes of vegetation between 2017 and 2019

The land cover areas with vegetation loss and gain between the years 2017 and 2019 are summarized in Table 3. Area is measured in hectares. The overall user accuracy for both 2017 and 2019 image classification is 96% and 92%, respectively. Kappa accuracy for both of the year’s classification stands at 92% and 88%. Result shows that camp increased dramatically in size following the massive outbreak of violence in Myanmar which acted as a driving force of the huge inflow of refugees in the Teknaf Peninsula. This is consistent with the findings of Hassan et al. (2018) and Braun et al. (2019). The new makeshift camps have been built cutting down forest in large numbers of vegetated areas. Satellite image of the study area is presented in Fig. 2 using false color composite.

Table 3 Change in vegetation extent: wetlands that changed into non-vegetation area was considered as non-vegetation remained, and non-vegetation area that changed into wetlands was considered as wetlands remained
Fig. 2
figure2

The satellite images of the study area of 2017 (a) and 2019 (b) using false color composite (SWIR1–NIR–red) for better visualization of land features

Study suggests that an estimated 1876 hectares of vegetated land have been decreased in Kutupalong–Balukhali and the surrounding region between 2017 and 2019. A major portion of these lands have been used for camp extension, and the rest of the forest land has been encroached by Rohingya families for fuelwood and other purposes. Figure 3 depicts the land use changes of the study area. It is evident that non-forest activities increased intensely in size, resulting in alternation of dense vegetation in the northeast part of the image. Along with the vegetation loss, 5420 hectares of vegetated areas remained to date and another 1039 hectares became vegetated from non-vegetated areas during this period. The reason behind this gain of vegetation might be community afforestation program or other initiatives taken in the study area. So, net vegetation loss is 836 hectares in the given period. The predicted reason for this huge vegetation loss might involve the anthropogenic activities by Rohingya families involving wood collection for fuel or other subsistence needs from the forested areas along with the expansion of the camps. A change detection map is produced using the two dated images, shown in Fig. 4.

Fig. 3
figure3

Classified images of the study area; a represents the land cover map of the study area before the influx (May, 2017), and b shows the same area after 20 months of the influx (April, 2019). Three classes are used (vegetated, wetlands and non-vegetated)

Fig. 4
figure4

Comparison of the both images for land cover change detection. Three classes are considered where red color marks the area where loss of vegetation occurred and light green shows the newly vegetated area

Changes of LST between 2017 and 2019

Figure 5 indicates the changes in LST between pre- influx and post-influx. Higher temperature (more than 31 °C) was present only at the northeastern side of the map where the refugee camp has existed before the recent influx. The surrounded regions were relatively low in terms of LST. But the post-influx LST map reveals that temperature has increased spatially throughout the study area, particularly where the vegetation loss occurred due to the camp extension. Built environment decreasing surface albedo, huge population density as well as increased usage of tarpaulin as materials for building the shelters also play vital roles in increasing LST. In 2017, the highest values of temperature were noticed in pre-influx camp area (Kutupalong Rohingya Camp). But the 2019 map indicates the areal extent of LST around the camps saw a noticeable rise as a result of cutting down trees regularly for household purposes and expanding the camps in an unprecedented manner. However, temperature has decreased slightly in the eastern region of wetlands where vegetation gain occurred in the surrounding areas. Figure 6 represents a density plot of LST of 2017 and 2019 images. It suggests that in 2017, LST was mainly concentrated between 28 and 30 °C. But in 2019, the peak lowered its density and stretched with a peak around 29 °C. It is also clear that the peak steadily declined to 31 °C and then gradually descended toward 36 °C. Change detection classification was compared with 2019’s LST, and class-wise statistics of temperatures were computed. Highest mean temperature was found over areas of vegetation loss, whereas the lowest over wetlands. The lowest minimum temperature was over vegetated area and the highest maximum temperature over areas of vegetation loss. A detailed picture of LST of the derived land use types is shown in Table 4.

Fig. 5
figure5

Land surface temperature (LST) of pre- and post-influx images where a, b represent the year 2017 and 2019, respectively. A significant change can be identified as the red color which represents higher LST spread throughout the region

Fig. 6
figure6

Density plot comparing two LST images of 2017 and 2019

Table 4 Class-wise statistics of temperature (°C) values in 2019 using zonal statistics as tables

Discussion

Over the years, the use of machine learning algorithms in satellite image classifications is being the torchlight of spatial change detection. In this study, RF is used on two different images to extract spatiotemporal switching of land cover, vegetation in particular. Planck equation was used for LST estimation as it gives less error than other formulas in Landsat 8 images (Ndossi and Avdan 2016b). The main objective of this paper is to measure the extent of vegetation cover change with the change of LST for the Kutupalong–Balukhali refugee camps and its neighborhood areas. Datasets are obtained for the months of April and May of respective years.

The Teknaf region has been experiencing the loss of vegetation cover since the 1980s as fuelwood collection and forest encroachment activities by the Rohingya started in that period. This trend has continued throughout the years as the families had very few livelihood options (Tani and Rahman 2018). Tani and Rahman (2018) showed vegetated forest lands are being encroached by Rohingya people either for selling wood in the local market for economic hardship as well as for using as fuelwood. Requirements of wood as fuel for domestic purposes act as a catalyst for hill forest degradation (Chowdhury et al. 2014; Uddin and Mukul 2007). However, this situation has been exacerbated by the recent influx of 2017. Most of the parts of Ukhia, especially in Kutupalong and Balukhali land use, have changed drastically within a very short period of time. Surrounding hill forests have degraded and deforested to create shelters along with the old makeshift camps where over a million refugees now reside combining before and after the influx period (UNDP Bangladesh and UN Women Bangladesh 2018). Results showed an indiscriminate vegetation loss surrounding the camp areas. Forestlands were cut down in an unprecedented manner to make new makeshift camps to settle down the inflow of refugees. However, the loss of vegetation continued in a drastic manner even after 2 years of the influx period mainly because of forest encroachment by the inhabitants. Reports say there is an increasing need for fuelwood or timber by the Rohingya families which they collect from the nearby forests. There is a current demand of 750,000 kg of fuelwood by the Rohingya families every day (Haque 2017) which mounts a huge amount of pressure on the nearby forest lands. A report claimed that 5013 acres of forested land has been cleansed by the refugees till now, while this number is on the rise (Hussain 2018). The present research result showed that a total of 1876 hectares of vegetated lands was decreased from 2017 to 2019 in the Kutupalong–Balukhali regions. The overall and kappa accuracies of the analysis were between 97 and 99%. Imtiaz (2018) carried out an analysis using remote sensing technique in the Teknaf sub-district and found vegetation loss of 1284.48 hectares after 4 months only of the recent influx. Hassan et al. (2018) used RF classification technique and found 5650 acres of forested lands was decreased between December 2016 and February 2017 for the whole camp area. These vegetation losses and forest clearing put excessive pressure on nearby protected forest lands such as Teknaf Wildlife Sanctuary, Himchari National Park and Inani National Park. These protected forest lands are home to a rich array of biodiversity and plants which are marked as critical and endangered (Tani and Rahman 2018; IUCN Bangladesh 2000). These critical habitats are further disturbed by anthropogenic activities by the refugees. Specially, the Asian elephants are marked endangered by Rahman (2019) as their corridors and routes have been blocked by the refugee settlements as well as their shortage of food due to deforestation. Forest encroachment also increases the risks of man-wildlife conflicts. As of now, since the recent influx, 13 human deaths and more than 50 injuries have been reported due to the man–elephant conflicts (Rahman 2019). Furthermore, degradation of forest covers and hill cutting increase the risk of potential landslides as the hills lose their natural settings (Mahmud 2017). Moreover, refugees preferred to build their shelters on top of hills and slopes rather than on low-elevated lands in order to avoid flooding which makes them more vulnerable toward landslide risks (Reuters 2017).

The presence of vegetation in an area has close relation with the local climatic condition along with LST. Vegetation cover cools down the environment to a great extent by providing shade to the surface (Sailor 1998), avoiding direct sunlight and reflecting most of it (Akbari et al. 2001) as well as a mature tree can lose up to 450 L of water per day to the environment through evapotranspiration (Johnston and Newton 2004). The presence of vegetation in an area increases the heat flux of radiation of the earth surface by absorbing the radiation energy during evapotranspiration (Sinha et al. 2014). On the other hand, built-up areas are non-transpiring and non-evaporative surfaces replacing natural vegetation which increases the LST (Weng 2001; Weng and Yang 2004). The analysis showed that LST in 2017 was highest in the pre-existed refugee camp. In the post-influx year (2019), LST of camp areas saw a noticeable rise; then the natural vegetation covers in those areas were replaced by shelters made of tarpaulin, mud, bamboo which are non-evaporative in nature. The highest temperature occurrence expanded spatially over camp areas except areas of vegetation gain near wetlands.

Along with the continuous forest degradation, LST increased considerably in the surrounding camp areas. If this trend continues, soon the place will become barren land without any trees and the LST will be much higher. Although repartition process of the Rohingya refugees is under consideration, it seems they are going to stay here for a while (IOM 2018). So, the Government of Bangladesh must need to come with a plan that suggests a long-term strategy to deal with the aforesaid problems.

Conclusion

After the recent massive Rohingya outbreak in Myanmar, an enormous humanitarian crisis started that triggered a huge inflow of refugees to the Teknaf Peninsula. The Government of Bangladesh and the people of the host communities were not prepared for this situation. As a result, this huge amount of refugees has put substantial pressure on the host communities and to the local environment. Based on remote sensing data and RF classification technique, it was found that an estimated 1876 hectares of forested lands turned to non-forest areas throughout Kutupalong–Balukhali Rohingya neighborhoods. It is clear from the findings that Rohingya families are collecting fuelwood from the nearby forests which eventually destroying the forest covers of that area. The pace of this destruction is at such an alarming rate that this trend continuation could be a significant driver of deforestation. Critical habitats which are already marked as endangered will face tremendous problems associated with the degradation. LST has increased substantially along with this vegetation loss, particularly in the camps and surrounding areas where most of the vegetation losses took place. Densely populated regions with high temperature could be an influence for spreading heat-induced diseases. Further studies are strongly suggested to investigate the impacts of this massive outbreak on the public health of both the refugees and the local people of Bangladesh. The whole environment of the area will be threatened in a great scale if no effective measures are taken soon. Therefore, the concerned authorities must take an efficient approach to control the forest degradation with afforestation programs and provide an alternative source of fuel such as LPG to the refugees.

References

  1. Adiguzel, F., Cetin, M., Kaya, E., Simsek, M., Gungor, S., & Sert, E. B. (2020). Defining suitable areas for bioclimatic comfort for landscape planning and landscape management in Hatay, Turkey. Theoretical and Applied Climatology, 139(3–4), 1493–1503.

    Article  Google Scholar 

  2. Almutairi, A., & Warner, T. A. (2010). Change detection accuracy and image properties: A study using simulated data. Remote Sensing, 2(6), 1508–1529.

    Article  Google Scholar 

  3. Amiri, R., Weng, Q., Alimohammadi, A., & Alavipanah, S. K. (2009). Spatial–temporal dynamics of land surface temperature in relation to fractional vegetation cover and land use/cover in the Tabriz urban area, Iran. Remote Sensing of Environment, 113(12), 2606–2617.

    Article  Google Scholar 

  4. Akbari, H., Pomerantz, M., & Taha, H. (2001). Cool surfaces and shade trees to reduce energy use and improve air quality in urban areas. Solar Energy, 70(3), 295–310.

    Article  Google Scholar 

  5. Avdan, U., & Jovanovska, G. (2016). Algorithm for automated mapping of land surface temperature using LANDSAT 8 satellite data. Remote Sensing, 2016(3), 16. https://doi.org/10.1155/2016/1480307.

    Article  Google Scholar 

  6. BBC News. (2018). Myanmar Rohingya: What you need to know about the crisis. Retrieved July 13, 2019, from https://www.bbc.com/news/world-asia-41566561.

  7. BBS. (2011). Bangladesh population and housing census 2011.

  8. Belgiu, M., & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. ISPRS Journal of Photogrammetry and Remote Sensing, 114, 24–31.

    Article  Google Scholar 

  9. Benjamin, L., Ned, H., & Jakob, S.-W. (2019). RStoolbox: tools for remote sensing data analysis. R package version 0.2.6. Retrieved August 8, 2019 from https://CRAN.R-project.org/package=RStoolbox.

  10. Berry, L. (2008). The impacts of environmental degradation on refugee—Host relationships. African Security Studies, 17(3), 125–131.

    Article  Google Scholar 

  11. Bharath, S., Rajan, K., & Ramachandra, T. (2013). Geostatistics: An overview land surface temperature responses to land use land cover dynamics. A SciTechnol Journal, 1(4), 1–10. https://doi.org/10.4172/2327-4581.1000112.

    Article  Google Scholar 

  12. Braun, A., Fakhri, F., & Hochschild, V. (2019). Refugee camp monitoring and environmental change assessment of Kutupalong, Bangladesh, based on radar imagery of sentinel-1 and ALOS-2. Remote Sensing, 11(17), 2047.

    Article  Google Scholar 

  13. Braun, A., Lang, S., & Hochschild, V. (2016). Impact of refugee camps on their environment a case study using multi-temporal SAR data. Journal of Geography, Environment and Earth Science International, 4(2), 1–17.

    Article  Google Scholar 

  14. Breiman, L. (2001). Random forests. Machine learning, 45(1), 5–32.

    Article  Google Scholar 

  15. Bremner, J., López-Carr, D., Suter, L., & Davis, J. (2010). Population, poverty, environment, and climate dynamics in the developing world. Interdisciplinary Environmental Review, 11(2), 112–126.

    Article  Google Scholar 

  16. Carlson, T. N., & Ripley, D. A. (1997). On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sensing of Environment, 62(3), 241–252.

    Article  Google Scholar 

  17. Cetin, M. (2015). Evaluation of the sustainable tourism potential of a protected area for landscape planning: A case study of the ancient city of Pompeipolis in Kastamonu. International Journal of Sustainable Development & World Ecology, 22(6), 490–495.

    Article  Google Scholar 

  18. Cetin, M. (2016). Sustainability of urban coastal area management: A case study on Cide. Journal of Sustainable Forestry, 35(7), 527–554. https://doi.org/10.1080/10549811.2016.1228072.

    Article  Google Scholar 

  19. Cetin, M., Zeren, I., Sevik, H., Cakir, C., & Akpinar, H. (2018). A study on the determination of the natural park's sustainable tourism potential. Environmental Monitoring and Assessment, 190(3), 167.

    Article  Google Scholar 

  20. Chowdhury, M. S. H., Nazia, N., Izumiyama, S., Muhammed, N., & Koike, M. (2014). Patterns and extent of threats to the protected areas of Bangladesh: The need for a relook at conservation strategies. Parks, 20(1), 91–104.

    Article  Google Scholar 

  21. Dabelko, G. D., Lalasz, R., Thomas, R. E., Hildebrandt, T., Kaczor, J., Méndez, A., et al. (2002). Environmental change and security project report. Washington: The Woodrow Wilson Center.

    Google Scholar 

  22. Ghimire, K. (1994). Refugees and deforestation1. International Migration, 32(4), 561–570.

    Article  Google Scholar 

  23. Ghimire, S., Higaki, D., & Bhattarai, T. (2013). Estimation of soil erosion rates and eroded sediment in a degraded catchment of the Siwalik Hills, Nepal. Land, 2(3), 370–391.

    Article  Google Scholar 

  24. Ghosh, A., Sharma, R., & Joshi, P. K. (2014). Random forest classification of urban landscape using Landsat archive and ancillary data: Combining seasonal maps with decision level fusion. Applied Geography, 48, 31–41.

    Article  Google Scholar 

  25. Gislason, P. O., Benediktsson, J. A., & Sveinsson, J. R. (2006). Random forests for land cover classification. Pattern Recognition Letters, 27(4), 294–300. https://doi.org/10.1016/j.patrec.2005.08.011.

    Article  Google Scholar 

  26. Glade, T. (2003). Landslide occurrence as a response to land use change: A review of evidence from New Zealand. CATENA, 51(3–4), 297–314.

    Article  Google Scholar 

  27. Gondwe, S. V., Muchena, R., & Boys, J. (2018). Detecting land use and land cover and land surface temperature change in Lilongwe City, Malawi. Journal of Remote Sensing & GIS, 9(2), 17–26.

    Google Scholar 

  28. Gounaridis, D., Apostolou, A., & Koukoulas, S. (2016). Land cover of Greece, 2010: A semi-automated classification using random forests. Journal of Maps, 12(5), 1055–1062. https://doi.org/10.1080/17445647.2015.1123656.

    Article  Google Scholar 

  29. Hassan, M., Smith, A., Walker, K., Rahman, M., & Southworth, J. (2018). Rohingya refugee crisis and forest cover change in Teknaf, Bangladesh. Remote Sensing, 10(5), 689.

    Article  Google Scholar 

  30. Haque, A. R. (2017). Influx of Rohingya refugees entails a serious burden on the economy. The Financial Express. Retrieved August 21, 2019, from https://thefinancialexpress.com.bd/views/views/influx-of-rohingya-refugees-entails-a-serious-burden-on-the-economy-1513519560.

  31. Hijmans, R. J. (2019). raster: geographic data analysis and modeling. R package version 3.0–7. Retrieved August 8, 2019 from https://CRAN.R-project.org/package=raster.

  32. Hussain, A. (2018). Rohingya influx, a threat to forest resources. Dhaka Tribune. Retrieved July 13, 2019, from https://www.dhakatribune.com/bangladesh/2018/03/20/rohingya-influx-a-threat-to-forest-resources.

  33. Immitzer, M., Atzberger, C., & Koukal, T. (2012). Tree species classification with Random forest using very high spatial resolution 8-band worldview-2 satellite data. Remote Sensing, 4(9), 2661–2693. https://doi.org/10.3390/rs4092661.

    Article  Google Scholar 

  34. Imtiaz, S. (2018). Ecological impact of Rohingya refugees on forest resources: Remote sensing analysis of vegetation cover change in Teknaf Peninsula in Bangladesh. Ecocycles, 4(1), 16–19.

    Article  Google Scholar 

  35. International Organization for Migration (IOM). (2018). Rohingya refugee crisis response. Retrieved September 3, 2019 from https://www.iom.int/sitreps/bangladesh-iom-bangladesh-rohingya-refugee-crisis-response-external-update-december-2018.

  36. IUCN Bangladesh. (2000). Red book of threatened mammals of Bangladesh. Dhaka: IUCN-The World Conservation Union.

    Google Scholar 

  37. Jhonnerie, R., Siregar, V. P., Nababan, B., & Budi, L. (2015). Random forest classification for mangrove land cover mapping using Landsat 5 TM and ALOS PALSAR imageries. Procedia Environmental Sciences, 24, 215–221. https://doi.org/10.1016/j.proenv.2015.03.028.

    Article  Google Scholar 

  38. Johnston, J., & Newton, J. (2004). Building green: A guide to using plants on roofs, walls and pavements (p. 95). London: Ecology Unit.

    Google Scholar 

  39. Kachhwala, T. S. (1985). Temporal monitoring of forest land for change detection and forest cover mapping through satellite remote sensing. In Proceedings of the 6th Asian conference on remote sensing (pp. 77–83). Hyderabad: National Remote Sensing Agency.

  40. Karlson, M., Ostwald, M., Reese, H., Sanou, J., Tankoano, B., & Mattsson, E. (2015). Mapping tree canopy cover and aboveground biomass in Sudano-Sahelian woodlands using Landsat 8 and random forest. Remote Sensing, 7(8), 10017–10041. https://doi.org/10.3390/rs70810017.

    Article  Google Scholar 

  41. Langer, S., Tiede, D., & Lüthje, F. (2015). Long-term monitoring of the environmental impact of a refugee camp based on landsat time series: The example of deforestation and reforestation during the whole lifespan of the camp Lukole, Tanzania. GI_Forum Journal Geographic Information Science, 1, 434–437. https://doi.org/10.1553/giscience2015s434.

    Article  Google Scholar 

  42. Li, Y. Y., Zhang, H., & Kainz, W. (2012). Monitoring patterns of urban heat islands of the fast-growing Shanghai metropolis, China: Using time-series of Landsat TM/ETM+ data. International Journal of Applied Earth Observation and Geoinformation, 19, 127–138.

    Article  Google Scholar 

  43. Li, Z. L., Tang, B. H., Wu, H., Ren, H., Yan, G., Wan, Z., et al. (2013). Satellite-derived land surface temperature: Current status and perspectives. Remote Sensing of Environment, 131, 14–37.

    Article  Google Scholar 

  44. Liu, L., & Zhang, Y. (2011). Urban heat island analysis using the landsat TM data and ASTER Data: A case study in Hong Kong. Remote Sensing, 3(7), 1535–1552. https://doi.org/10.3390/rs3071535.

    Article  Google Scholar 

  45. Mahmud, F. (2017). For the Rohingya in Bangladesh’s refugee camps: Living is surviving. The Wire. Retrieved July 17, 2019 from https://thewire.in/189522/rohingya-bangladesh-refugee-camps/.

  46. Mantey, S., Tagoe, N. D. & Abaidoo, C. A. (2014), Estimation of land surface temperature and vegetation abundance relationship – a case study, 3rd UMaT biennial international mining & mineral conference, 30 July–2 August, 2014.

  47. Mohammed, E. A., Hani, Z. Y., & Kadhim, G. Q. (2018). Assessing land cover/use changes in Karbala city (Iraq) using GIS techniques and remote sensing data. Journal of Physics: Conference Series, 1032(1), 012047.

    Google Scholar 

  48. Ndossi, M. I., & Avdan, U. (2016a). Application of open source coding technologies in the production of Land Surface Temperature (LST) maps from Landsat: A PyQGIS plugin. Remote Sensing. https://doi.org/10.3390/rs8050413.

    Article  Google Scholar 

  49. Ndossi, M. I., & Avdan, U. (2016b). Inversion of land surface temperature (lst) using terra aster data: A comparison of three algorithms. Remote Sensing, 8(12), 1–19. https://doi.org/10.3390/rs8120993.

    Article  Google Scholar 

  50. Pal, M. (2005). Random forest classifier for remote sensing classification. International Journal of Remote Sensing, 26(1), 217–222. https://doi.org/10.1080/01431160412331269698.

    Article  Google Scholar 

  51. Poumadere, M., Mays, C., Le Mer, S., & Blong, R. (2005). The 2003 heat wave in France: Dangerous climate change here and now. Risk Analysis: An International Journal, 25(6), 1483–1494.

    Article  Google Scholar 

  52. Qin, Z., Dall’Olmo, G., Karnieli, A., & Berliner, P. (2001). Derivation of split window algorithm and its sensitivity analysis for retrieving land surface temperature from NOAA-advanced very high resolution radiometer data. Journal of Geophysical Research Atmospheres, 106(D19), 22655–22670. https://doi.org/10.1029/2000JD900452.

    Article  Google Scholar 

  53. Raja, R. A., Anand, V., Kumar, A. S., Maithani, S., & Kumar, V. A. (2013). Wavelet based post classification change detection technique for urban growth monitoring. Journal of the Indian Society of Remote Sensing, 41(1), 35–43.

    Article  Google Scholar 

  54. Rahman, M. H. (2019). Rohingya refugee crisis and human vs. elephant (Elephas maximus) conflicts in Cox’s Bazar district of Bangladesh. Journal of Wildlife and Biodiversity, 3(3), 10–21. https://doi.org/10.22120/jwb.2019.104762.1057.

    Article  Google Scholar 

  55. Reuters. (2017). Life in the camps. Retrieved August 25, 2019 from https://fingfx.thomsonreuters.com/gfx/rngs/MYANMARROHINGYA/010051VB46G/index.html.

  56. Rodriguez-Galiano, V., & Chica-Olmo, M. (2012). Land cover change analysis of a Mediterranean area in Spain using different sources of data: Multi-seasonal Landsat images, land surface temperature, digital terrain models and texture. Applied Geography, 35(1–2), 208–218. https://doi.org/10.1016/j.apgeog.2012.06.014.

    Article  Google Scholar 

  57. Sahana, M., Ahmed, R., & Sajjad, H. (2016). Analyzing land surface temperature distribution in response to land use/land cover change using split window algorithm and spectral radiance model in Sundarban Biosphere Reserve, India. Modeling Earth Systems and Environment, 2(2), 81.

    Article  Google Scholar 

  58. Sailor, D. J. (1998). Simulations of annual degree day impacts of urban vegetative augmentation. Atmospheric Environment, 32(1), 43–52.

    CAS  Article  Google Scholar 

  59. Sesnie, S., Gessler, P., Finegan, B., & Thessler, S. (2008). Integrating Landsat TM and SRTM-DEM derived variables with decision trees for habitat classification and change detection in complex neotropical environments. Remote Sensing of Environment, 112(5), 2145–2159.

    Article  Google Scholar 

  60. Sinha, S., Pandey, P. C., Sharma, L. K., Nathawat, M. S., Kumar, P., & Kanga, S. (2014). Remote estimation of land surface temperature for different LULC features of a moist deciduous tropical forest region. In Remote sensing applications in environmental research (pp. 57–68).

  61. Skole, D., & Tucker, C. (1993). Tropical deforestation and habitat fragmentation in the Amazon: Satellite data from 1978 to 1988. Science, 260(5116), 1905–1910.

    CAS  Article  Google Scholar 

  62. Sobrino, J. A., Jimenez-Munoz, J. C., & Paolini, L. (2004). Land surface temperature retrieval from LANDSAT TM 5. Remote Sensing of environment, 90(4), 434–440.

    Article  Google Scholar 

  63. Sud, Y. C., Lau, W. K., Walker, G. K., Kim, J. H., Liston, G. E., & Sellers, P. J. (1996). Biogeophysical consequences of a tropical deforestation scenario: A GCM simulation study. Journal of Climate, 9(12), 3225–3247.

    Article  Google Scholar 

  64. Tani, M., & Rahman, M. A. (2018). Deforestation in the Teknaf Peninsula of Bangladesh. Berlin: Springer.

    Google Scholar 

  65. Thanh Noi, P., & Kappas, M. (2017). Comparison of random forest, k-nearest neighbor, and support vector machine classifiers for land cover classification using Sentinel-2 imagery. Sensors (Basel, Switzerland). https://doi.org/10.3390/s18010018.

    Article  Google Scholar 

  66. Uddin, M. B., & Mukul, S. A. (2007). Improving forest dependent livelihoods through NTFPs and home gardens: a case study from Satchari National Park. In Making conservation work: Lining rural livelihoods and protected areas in Bangladesh (pp. 13–35). East-West Center, Honolulu, and Nishorgo Program of the Bangladesh Forest Department Dhaka, Bangladesh.

  67. UNDP Bangladesh and UN WOMEN Bangladesh. (2018). Report on environmental impact of Rohingya influx. Dhaka, Bangladesh.

  68. UNHCR. (2019). Rohingya emergency. Retrieved July 17, 2019 from https://www.unhcr.org/en-us/rohingya-emergency.html?query=rohingya%20crisis.

  69. USGS. (2016). Landsat collections—2016.

  70. Wang, C., & Myint, S. (2016). Environmental concerns of deforestation in Myanmar 2001–2010. Remote Sensing, 8(9), 728.

    Article  Google Scholar 

  71. Wang, M., He, G., Zhang, Z., Wang, G., Wang, Z., Yin, R., et al. (2019). A radiance-based split-window algorithm for land surface temperature retrieval: Theory and application to MODIS data. International Journal of Applied Earth Observation and Geoinformation, 76, 204–217.

    Article  Google Scholar 

  72. Weng, Q. (2001). A remote sensing-GIS evaluation of urban expansion and its impact on surface temperature in the Zhujiang Delta, China. International Journal of Remote Sensing, 22(10), 1999–2014.

    Google Scholar 

  73. Weng, Q., & Yang, S. (2004). Managing the adverse thermal effects of urban development in a densely populated Chinese city. Journal of Environmental Management, 70(2), 145–156.

    Article  Google Scholar 

  74. White, M. A., Nemani, R. R., Thornton, P. E., & Running, S. W. (2002). Satellite evidence of phenological differences between urbanized and rural areas of the eastern United States deciduous broadleaf forest. Ecosystems, 5(3), 260–273.

    Article  Google Scholar 

  75. Yankovich, K. S., Yankovich, E. P., & Baranovskiy, N. V. (2019). Classification of vegetation to estimate forest fire danger using landsat 8 images: Case study. Mathematical Problems in Engineering, 2019, 1–14. https://doi.org/10.1155/2019/6296417.

    Article  Google Scholar 

  76. Zhang, J., Wang, Y., & Li, Y. (2006). A C++ program for retrieving land surface temperature from the data of Landsat TM/ETM+ band6. Computers & Geosciences, 32(10), 1796–1805.

    CAS  Article  Google Scholar 

Download references

Acknowledgements

The authors are thankful to the Department of Geography and Environmental Studies, University of Chittagong, Bangladesh, for different forms of supports during the study.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Md. Atiqur Rahman.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Rashid, K.J., Hoque, M.A., Esha, T.A. et al. Spatiotemporal changes of vegetation and land surface temperature in the refugee camps and its surrounding areas of Bangladesh after the Rohingya influx from Myanmar. Environ Dev Sustain 23, 3562–3577 (2021). https://doi.org/10.1007/s10668-020-00733-x

Download citation

Keywords

  • LST
  • Random forest classification
  • Remote sensing
  • Rohingya influx
  • Vegetation loss