Responses of Spatiotemporal Vegetative Land Cover To Meteorological Changes in Bangladesh

Quantifying the response of vegetation cover change (VCC) to climatic variables is a gap that is mandatory for the conservation and rehabilitation of natural landscape to ensure sustainability. This study aims to assess the response of VCC to temperature and rainfall change in Bangladesh. We used (i) Landsat images to analyze VCC using image classication method, Normalized Difference Vegetation Index (NDVI) (ii) temperature and rainfall statistics to investigate the spatiotemporal variations (SV) of meteorological factors, urban lands, VCC in all the 64 districts of Bangladesh during 1990-2018 and examined their correlation. To quantify the impact of urbanization on VCC, two regression models were built between growing-season NDVI (GNDVI) and urban land proportion (PLU). Results show that the SV of precipitation, temperature, GNDVI, and PUL varied greatly among the districts. GNDVI was found closely related to climatic variables and less sensitive to climatic factor changes. There has been found a signicant correlation between the trend of GNDVI and GP while the negative correlation between GNDVI trend and GT, ΔPUL. Strong sensitivity of GNDVI change to GP was calculated in the range of precipitation 2200-3000mm and GNDVI to GT change in the range of temperature 30 0 C-31 0 C. Besides, urban expansion was found mostly responsible for VCC in the study area.


Introduction
Forests and vegetation provide biodiversity protection that preserves the soil cover and balances the hydrological cycle, ecosystem, air temperatures, humidity, and rainfall which helps to mitigate climate change impacts (Zheng et al. 2019). But, due to population growth and urban expansion, human beings are destructing the forest-cover falling under the greater dominion of land surface vegetation-cover (Amera and Tefera, 2013). Due to massive Land Use/Land Cover (LULC) transformation, carbon emissions are increasing and carbon sinks are decreasing and accelerating climate change. In this sense, urbanization is one of the key factors impacting forest cover, which has profound effects on both national and global climate (Hunt et al. 2017). Land cover around the world is transforming, in uenced by both natural and arti cial factors (Hassan, 2017). Vegetation, one of the most important land covers of the earth's ecosystem, offers an extensive variety of social, ecological, and environmental services and bene ts habitats for sustainability (Wu, 2010;Zheng et al. 2019; Robinson & Lundholm, 2012). It helps to maintain sustainability between environment and ecology through carbon sequestration (Gratani & Bonito, 2016) improving air quality (Escobedo & Nowak, 2009), conserving soil and water, preserving biodiversity (Dana & Mota, 2002), regulating microclimates (Jonsson, 2004) and mitigating disasters (Jenerette et al. 2011). However, palaeoenvironmental and palaeoecological records show that human activities and climate change have been disrupting ecosystems as well as vegetation (Marignani et al. 2016;Mercuri et al. 2015;Li & Zhou, 2010). Over the past few decades, vegetation coverage has undergone a remarkable transformation which is greatly affecting environmental and habitat sustainability (Jin et al. 2018; Grimm et al. 2008). Understanding the relationship between climate change, dynamics of vegetation, and the services offered to humanity by ecosystems is one of the main research challenges in the 21st century.
Environmental concern has contracted as one of the major worldwide attention that distinctly as well as jointly affect all countries (Amera and Tefera, 2013). With the expansion of population and spontaneous extension of cities, land use patterns and biological systems have changed, prompting the arrangement of metropolitan situated natural di culties around the globe (Arsanjani et al. 2013). Land use and land cover (LULC) type has been changing quickly because of the many main impetuses. Thus, carbon emissions, environmental change, the shift of biological systems, ecological corruption, and the random condition have been expanding which making the climate of any region inadmissible for human home (Ameen and Mourshed, 2017;Zheng et al. 2017). Therefore, Bangladesh is one of the world's exceptionally populated nations, where the metropolitan population has developed after some time because of relocation from rustic to mechanical or support areas looking for work openings and driving quali ed everyday environments (Zaman et al. 2010). Rapid population growth may have bene cial economic development in uences but have detrimental effects on LULC change and sustainable development (Kafy et al. 2021). The urban areas of Bangladesh have been confronting the issue of spontaneous metropolitan extension like other creating urban communities (Zaman et al. 2010). The LULC management strategy is now one of the most important obstacles to the mitigation of the adverse environmental and climatic effects by limiting unplanned urban growth and promoting green coverage.
Climate is considered as the most perilous environmental factor which not only affects the ecosystem and the environment but also LULC and humans (Hunt et al. 2017). Generally, researchers use remote-sensing techniques to illustrate land use/ land cover (LUCC) changes and vegetation variations at the city level or regional level. LULC classi cation, net primary productivity (NPP), and normalized difference vegetation index (NDVI) are utilized for illustrating vegetation activities (Wang et al. 2008;Kafy et al. 2020). The NDVI derived from satellite data is an important indicator for evaluating the state of living green vegetation and showing vegetation dynamics reacting to climate change (Kalisa et al. 2019;Philippon et al. 2007;Zhang et al. 2013). Using the supervised image classi cation method, Rai et al. (2017) showed that Bangladesh has undergone a total loss of 6.2% (9054 sq km) of vegetated areas during 1976-2014. Hasan et al. (2013) have shown that total vegetation was 12.11% in 1976, 9.02% in 2000, and 9.84% in 2010. Fu et al. (2013) have shown the NDVI changing trend exhibited spatial differences at a signi cant level over land surfaces around the world. Long-term climate change or interannual climatic variations can affect photosynthesis of vegetation, respiration, and organic carbon decomposition (Fu et al. 2013 This study detailed the interannual growing-season NDVI (GNDVI), temperature (GT), precipitation (GP), and proportion of urban land (PUL) in Bangladesh over the past three decades. And then performed a sensitivity analysis, analyzed the spatiotemporal changing trend of GNDVI, GP, GT, and PUL, examined the correlation between GNDVI and GT, GP, and PUL change to assess the response of VCC to climatic variables. Finally, the impacts of urban expansion on GNDVI dynamics are quanti ed through the regression model. In general, the long-term VCC should be studied in light of their history, but used recent observation records  to perform studies in an important 'window' of human history. The changes in vegetation, the landscape, and the climate could also represent long-term environmental changes in Bangladesh.

Study area selection and Quanti cation of Urban Land Expansion
Bangladesh is a small country of 1,47,570 sq km with complicated climatic and geographical conditions and a relatively rich ecosystem diversity. However, the rapid urban expansion has enormously impacted and altered the social, economic, and environmental conditions of Bangladesh (Hasan et al. 2017). The study area of this study encompasses entire Bangladesh ( Fig. 1). It is a South Asian country located between the latitudes 20 • 340 and 26 • 380 north and the longitudes 88 • 010 and 92 • 410 east (Xu et al. 2020). The climate of the country is tropical in the south and subtropical in the center-north, the winter from November to February is pleasantly warm and sunny, a short hot spring between March and May, rainy season from June to October. Bangladesh is at and occupied by the huge Ganges-Brahmaputra Delta, and vulnerable to oods (Yu et al. 2018).
There are 64 districts in Bangladesh under eight divisions. All the districts were eventually selected in this study. The geometric center of the urban area of each district was determined by using the Feature to Point tool in ArcGIS. For each district, a radius was determined based on the built-up area and drawn a circle which was de ned urban areas as the inner zone (IZ). Exterior to this circle, a buffer area with a 10 km radius, the suburban areas were de ned as outer zone (OZ). ArcGIS was used to count the areas of extracted urban lands during the study year in IZ (AUL IZ ) and OZ (AUL OZ ). Using Eq. 1, the proportion of urban lands (PUL) under IZ and OZ was calculated. year 1991 to 2018 of each district was calculated to illustrate the urban expansion rate during the study period.

Analysis of Spatiotemporal Variations
Within a district, the climate background is considered to be the same. The mean GNDVI values in IZ (GNDVI IZ ) and in OZ (GNDVI OZ ) indicate the vegetative land covers in the inner zone (urban areas) and outer zone (suburban areas) respectively.
The GT and GP were determined by weather data at a nearby weather station to re ect the climate for each district in the growing season.
At rst, the variations GT, GP, PUL IZ and PUL OZ during the year 1991 to 2018 were analyzed for the 64 districts of Bangladesh.
Then, spatiotemporal variations of GNDVI IZ and GNDVI OZ were analyzed with respect to GT, and GP. To analyze the inter-annual variations in GT, GP and GNDVI throughout 1991-2018, linear patterns were analyzed using the ordinary lowest-quarters regression method. The GNDVI trend rates were used to illustrate the direction of change in vegetation cover (

Quanti cation of the Impact of Urban Expansion on GNDVI Dynamics
As climate change in the zone IZ and OZ are the same within a district, the different effects of the urban expansion between IZ and OZ could be attributed to GNDVI IZ and GNDVI OZ trend rates (Jin et al. 2018). In this regard, the differences between the trends rate of GNDVI IZ and GNDVI OZ were calculated. The differences in urban land expansion rates within zone IZ and OZ were calculated by calculating the differences between ΔPUL IZ and ΔPUL OZ . Finally, the impact of urban expansion on the vegetative cover was examined by analyzing the correlations between the trend rates of GNDVI IZ and ΔPUL IZ; and between GNDVI OZ and ΔPUL OZ .

Climate Factors
The climatic condition of all the districts of Bangladesh from season to season with the longitudinal change (Fig. 3, 4). The mean annual average growing-season temperature of Bangladesh has been calculated 30.24 o C, highest temperature during the study time recorded 38.95 o C in Rajshahi districts in 1992 and the lowest 2.6 o C in Tetulia in 2018. The average temperature in Bangladesh was observed from 16 o C to 24 o C from January to March. During April to July, the average temperature varies between 32 o C to 34 o C and the temperature on the west side of Bangladesh is comparatively higher than in the east (Fig. 2).
The mean GP has been calculated 196.22mm and 152.00mm for 1994 and 2018 respectively which implied that the mean GP has been reduced with a rate of 44.218mm/year or reduced by total annual precipitation 530.62mm. Figure 3B shows that the precipitation in the west side of Bangladesh has been reduced mostly than in the east. Table 1 shows that the trend of GT and GP ranged from 0.10 to 0.30 o C and 38.09 to 35.43mm per decade respectively. The GT for 46 districts had ascending trends and 18 has descending trends and the mean change trend of GT has been recorded to be increased 0.733 o C per decade. Most of the districts (Fig. 3A, B) experienced a decreasing trend of GP.  Figure 4 shows that most of the districts of Bangladesh have experienced both positive and negative climatic changes. Out of 64 districts, the average rainfall in only three districts has increased and dropped dramatically in 15 districts ( Figure 4A). The negative temperature change trend in 12 districts has been calculated during the study period while the range was calculated in between -0.146 0 C to -0.0141 0 C/ decade. The temperature change trend in 20 districts was between 0.250-0.381 0 C/decade and between 0.381 to 0.513 0 C/decade in 9 districts ( Figure 4B). The highest temperature increasing rate has been observed throughout the eastern portions of the country.

Urban Lands
Spatiotemporal urban land proportion change analysis shows the rapid urban expansion in all districts from 1990 to 2018. Figure 5A, B shows that the PUL IZ and PUL OZ

GNDVI
Spatiotemporal distribution of mean annual GNDVI including GNDVI IZ and GNDVI OZ for all 64 districts of the country were remarkably different (Fig. 6A, B, C, D The inter-annual variations of GNDVI during 1990-2018 also showed signi cant differences among the districts (Fig. 6A, B, C, D). The mean GNDVI IZ of 9 districts was less than 0.23 in 1990 and increased to 19 in 2018. About 12 districts GNDVI IZ value was greater than 0.4 while one district has a GNDVI IZ value greater than 0.5 in 1990 (Fig. 6A). In 2018, 16 districts' GNDVI IZ value was calculated more than 0.4 (Fig. 6C). The GNDVI IZ value for 20 districts experienced a decreasing trend (Fig. 6A, C) and the GNDVI OZ value for 10 districts experienced decreasing trend (Fig. 6B, D). The average trend rate of GNDVI IZ and GNDVI OZ were 0.021 and 0.026 per decade respectively while the trend rate of GNDVI IZ was greater than 0.05 per decade in the 10 districts and GNDVI OZ greater than 0.05 per decade in the 8 districts in Bangladesh. Districts with increasing GNDVI trends mostly located in the eastern part of the country especially the capital city has experienced the most GNDVI change.

Relationship between Mean Annual GNDVI and Climate
The lower GNDVI IZ was calculated when the GT was more than 31.5 O C and the GP less than 150mm (Fig. 7) and good vegetation density has been observed within the GT range between 30 to 31.5 O C and the GP range between 125 to 300mm.  (Fig. 7B).
The linear regression analysis in Fig. 9 shows that the GNDVI was positively correlated with the climatic factors (GT and GP).  Fig. 9 indicates that the mean annual GNDVI are closely related to the climate variables.

Relationship between GNDVI Variations and Climate Change
The correlation between the GNDVI trend rate with GT, GP, ΔPUL trend rates has been examined through identifying Pearson correlation value and Ordinary least squares regression coe cient value (Fig. 9, 10, 11, and Table 2). The analysis shows that the correlation between the GP and GNDVI was signi cant for most of the districts. Table 2 shows that the correlation between GNDVI trends (Both for IZ and OZ) with GP was negatively correlated so was the proportion of urban land change during 1990-2018. However, Fig. 9 shows that the correlation between GT and ΔPUL was insigni cant for most of the districts of Bangladesh. The Pearson correlation coe cient test results in Table 2 indicate the correlations between the trend rates of GNDVI and GT, GNDVI and ΔPUL, GNDVI and GP are insigni cant. The ΔPUL trends in the districts of Bangladesh located in the western part of the country were found more sensitive GT than the districts located in the eastern part. While the GNDVI trends of the districts located in the Dhaka, Khulna, Sylhet divisions were found less sensitive to climatic factors rather than the Chittagong, Rajshahi, Rangpur divisions. The overall sensitivity analysis of inter-annual variations of GNDVI was found less sensitive to the changes of GP and GT.

Impact of Urbanization on GNDVI Change
The correlation value in Table 2 indicates the positive correlations between GNDVI and ΔPUL indicates that the urban expansion is closely associated with the vegetation cover change in all the districts. The negative trend rate of GNDVI IZ was found for 20 districts and GNDVI OZ for 10 districts, which indicates that the trend of GNDVI rate in the urban areas of Bangladesh is lower than the suburban or rural areas.   (Kafy et al. 2021). In large cities such as Dhaka, Chattogram, Khulna, Comilla, and Rajshahi, urban areas include a large percentage of buildings or high density of built-up areas (Rai et al. 2017) resulting in the lower value of NDVI. According to Mukhopadhyay et al. (2018), the urbanization trend in the suburban areas was moderately high than the urban areas as the cities are expanding. This is con rmed by the GNDVI change in Fig. 6 and the change rate in Fig. 12, where the GNDVI values and GNDVI change trend in OZ were estimated higher than in IZ. Despite this, a positive linear correlation has been found between GNDVI changes and climatic factors (Fig. 8). Therefore, the spatial disparity between the districts with the average GNDVI is related to climate factors, but urban development could play some role at the local level.

Driving Forces of the Temporal Inconsistency of GNDVI
This study found the GNDVI change in all the districts of Bangladesh was less sensitive to change in climatic factors. A moderate number of districts located in the western part of the country (Fig. 9) showed a signi cant correlation between GT and GNDVI while a signi cant positive correlation has been estimated between GP and GNDVI ( Table 2). GP was found more sensitive to GNDVI when monthly total rainfall was more than 220mm i.e., strong sensitivity of vegetation to precipitation in the districts where annual total rainfall ranged from 2200 to 3000mm. The strong sensitivity of vegetation to temperature was found in the districts where the average temperature ranged from 30 to 31 0 C.  (Buyantuyev & Wu, 2009). The discrepancy of the interannual GNDVI variation between IZ and OZ in Fig. 12 ensured the impacts of urban expansion in Bangladesh. The GNDVI change trend in many adjacent districts showed great differences while the climatic factors in the adjacent cities are quite similar. Moreover, the GNDVI trend rate in urban and suburban areas was found different within a district and in suburban areas, urban land is expanding at a higher rate than in urban areas. This difference was considered as a consequence of urban land expansion instead of climate uctuations particularly because, during the study period, Bangladesh underwent rapid urbanization.

Conclusions
Vegetative land cover enriches both urban and rural life with numerous social and environmental services as well as economic services. But historical records show that over the past decades, these vegetative areas have been transformed enormously with the in uence of human activities and climate change. This study investigated the variations of spatiotemporal vegetation cover in Bangladesh and its sensitivity to climatic factor change through assessing the variations of interannual growingseason NDVI, precipitation, and temperature change in all the districts of Bangladesh. The quantitative assessment of the effect of urban expansion on VCC in urban and suburban areas was also done. The study found great differences of mean annual GNDVI, GP, GT, and PUL districts to districts in Bangladesh so as the changes of GP, GT, GNDVI, and PUL. The PUL IZ change rate was found a minimum of 0.53% per decade to a maximum of 21.30% per decade while PUL OZ was a minimum of 0.39% per decade to a maximum of 9.41% per decade indicating that Bangladesh underwent rapid urbanization during 1990-2018. The GNDVI IZ ranged from 0.391 − 0.355 and GNDVI OZ 0.657 − 0.648 per decade. For most of the districts, the suburban area or OZ has experienced more severe vegetative cover loss than urban areas or IZ, the opposite also occurred. The overall study demonstrated that the vegetative cover loss was less sensitive to climatic factors and negatively correlated with the urban expansion in Bangladesh. Thus, this study addressed the long-term impact of human activities on the eco-friendly land Figure 1 Study area map.

Figure 2
Page 13/20 Spatiotemporal distribution of (A) GT and (B) interannual variations of GT during 1990-2018.     Relationship between the trends of GNDVIIZ-OZ and ΔPULIZ-OZ for all the districts of Bangladesh.