Masudur Rahman

The sea level rise (SLR) in the Sundarbans areas is higher than the global-average rate of sea rise, and many studies assume that most of the dry land of the Sundarbans will be inundated by the end of the twenty-first century. This study aims to analyze the amount of dry land that can potentially be inundated by SLR in Sundarbans and the impact under different land cover conditions. Four SLR scenarios, a digital elevation data grid, and net subsidence data are used to map areas that will be potentially inundated by 2100. Results for the low (35 cm), mid (52 cm), high (70 cm), and extreme (147 cm) SLR scenarios indicate that the Sundarbans landmass area will be flooded up to 40 km² (1%), 72 km² (1.8%), 136 km² (3.4%), and 918 km² (23%), respectively, under the current net subsidence rate of −2.4 mm/year by 2100. Except for the extreme scenarios, the low, mid, and high SLR will result in riverbank and beach areas to be covered by water. The potential inundation areas of different vegetation cover classes that already exist today (2020) will be nominal for the low, mid, and high SLR scenarios. We also analyzed the sensitivity of the results through station-based SLR data, which fits with the low (35 cm) SLR scenarios under the −2.4 mm/year subsidence rate. This study concluded that the inundation aspect of SLR will not directly affect the Sundarbans; however, indirectly related threats and anthropogenic disturbances can be major drivers of the Sundarbans’ degradation by the end of the twenty-first century. This work discusses reasonable inundation scenarios integrating SLR and subsidence with a custom land-cover map that includes three forest-density categories. The study’s findings contribute to forest management planning and support the UN goals of the Bangladesh Delta Plan.

Deep Reinforcement Learning (RL) agents often overfit the training environment, leading to poor generalization performance. In this paper, we propose Thinker, a bootstrapping method to remove adversarial effects of confounding features from the observation in an unsupervised way, and thus, it improves RL agents' generalization. Thinker first clusters experience trajectories into several clusters. These trajectories are then bootstrapped by applying a style transfer generator, which translates the trajectories from one cluster's style to another while maintaining the content of the observations. The bootstrapped trajectories are then used for policy learning. Thinker has wide applicability among many RL settings. Experimental results reveal that Thinker leads to better generalization capability in the Procgen benchmark environments compared to base algorithms and several data augmentation techniques.

Changes in land cover are a major driving force behind habitat change, which significantly impacts the distribution of wildlife and ecological systems. However, there is a substantial lack of information on the effects of land cover changes on wildlife habitat and local conservation. Therefore, it is essential to understand how land cover changes may threaten future land cover trends and wildlife habitat loss, especially in protected areas. Landsat satellite imagery uses a geographic information system and remote sensing techniques to determine the spatiotemporal pattern of land cover change and its impact on the human–elephant conflict in the Fashiakhali Wildlife Sanctuary. We found that within the sanctuary (1994–2005), settlements, agricultural land, and bare land increased by 69.8 ha (2.3%), 991.6 ha (32.3%), and 39.5 ha (1.3%), and forest areas and water areas decreased by 1094.1 ha (35.7%) and 6.9 ha (0.2%), respectively. On the other hand (2005–2015), settlements, agricultural land, and water areas increased by 11.7 ha (0.4%), 264.7 ha (8.6%), and 36.2 ha (1.2%), and forest areas and bare land decreased by 308.9 ha (10.1%) and 3.7 ha (0.1%), respectively. Our findings have shown that increased agriculture and settlements have become a severe threat to the ecological sustainability of elephant habitat, resulting in habitat fragmentation and human encroachment of elephant habitats, as well as extreme pressure and competition on resources.

Visceral leishmaniasis is a considerable public health burden on the Indian subcontinent. The disease is highly endemic in the north-central part of Bangladesh, affecting the poorest and most marginalized communities. Despite the fact that visceral leishmaniasis (VL) results in mortality, severe morbidity, and socioeconomic stress in the region, the spatiotemporal dynamics of the disease have largely remained unexplored, especially in Bangladesh. Monthly VL cases between 2010 and 2014, obtained from subdistrict hospitals, were studied in this work. Both global and local spatial autocorrelation techniques were used to identify spatial heterogeneity of the disease. In addition, a spatial scan test was used to identify statistically significant space-time clusters in endemic locations of Bangladesh. Global and local spatial autocorrelation indicated that the distribution of VL was spatially autocorrelated, exhibiting both contiguous and relocation-type of diffusion; however, the former was the main type of VL spread in the study area. The spatial scan test revealed that the disease had ten times higher incidence rate within the clusters than in non-cluster zones. Both tests identified clusters in the same geographic areas, despite the differences in their algorithm and cluster detection approach. The cluster maps, generated in this work, can be used by public health officials to prioritize areas for intervention. Additionally, initiatives to control VL can be handled more efficiently when areas of high risk of the disease are known. Because global environmental change is expected to shift the current distribution of vectors to new locations, the results of this work can help to identify potentially exposed populations so that adaptation strategies can be formulated.

Satellite‐derived lightning data for 17 years (1998–2014) were used to evaluate the relation between environmental factors and lightning activity over the Bangladesh landmass. Time series convective available potential energy (CAPE) data were extracted from ERA‐40 reanalysis data while total and convective rainfalls were obtained from Tropical Rainfall Measuring Mission's monthly products. In addition, the product of CAPE and precipitation was computed and used as an additional variable. Three timescales – monthly, seasonal and annual – were utilized to determine the influence of precipitation and CAPE on lightning activity. The results indicated that CAPE stands out as an important variable at all of these timescales for predicting the occurrence of lightning. The correlation coefficient (r) between CAPE and lightning activity was found to be 0.902 (monthly), 0.703 (pre‐monsoon), 0.550 (monsoon) and 0.702 (annual), respectively. Total rain showed strong positive correlation with lightning on monthly scale (r = 0.734) and in the pre‐monsoon season (r = 0.701). However, such relationship was moderate during monsoon (r = 0.455). In contrast, convective rain showed slightly higher correlation during monsoon (r = 0.587) compared with that of pre‐monsoon season (r = 0.532). Because of strong seasonality in the data, convective rain did not exhibit strong relationship on annual scale (r = 0.227). The product variable (e.g. CAPE × precipitation) showed significant correlation on monthly (r = 0.895) and seasonal scales (r = 0.818 during pre‐monsoon and 0.686 in monsoon) but its influence appears to diminish on a longer timescale (r = 0.375). Spatial maps of correlation coefficient revealed significant positive correlation along relatively drier northern parts of Bangladesh. As lightning‐related fatality is on the rise, this study, first of its kind, is expected to inform public policy and provide information necessary for effective management of this atmospheric phenomenon to save lives and property in Bangladesh.

A database of lightning-related deaths and injuries in Bangladesh was developed from 1990 to mid-2016 from a variety of sources that contains a total of 5468 casualties, composed of 3086 fatalities and 2382 injuries. Spatial, temporal, and demographic aspects of these lightning casualties are evaluated in order to aid relevant entities in effective management of lightning-related meteorological hazards. The annual averages for Bangladesh are 114 fatalities and 89 injuries over the entire period. Weighting by population reveals a fatality rate of 0.92 per million people per year and an injury rate of 0.71. In contrast, the latest 6 years have a fatality rate of 1.6 and injury rate of 1.4. The rural portion of lightning fatalities is 93%. Most fatalities occurred between early morning (0600 LST) and early evening (2000). Through the year, more fatalities occur during the premonsoon season of March through May than during the monsoon season (June–September). The interannual time series of fatalities indicates an increase since the late 2000s, which is a result of greatly improved communications leading to better media reporting of lightning casualties. Bangladesh has also become much more populous in recent years. As a result, the most recent 6 years have 251 fatalities per year, which may be considered as the current estimate. The majority of lightning-related deaths occurred to males. Farming is the major activity at the time of lightning fatalities followed by being inside a dwelling and returning home or walking around homesteads/courtyards.