Assessing the Rapid Urbanization in Tertiary City of Bangladesh by Land Use and Land Cover Change Detection from 2000 to 2024 through NDVI Based Classification and future forecasting for 2032 by Cellular Automata (CA) model in Meherpur District

doi:10.21203/rs.3.rs-5275701/v1

Understanding Land Use and Land Cover (LULC) changes is vital for environmental sustainability, particularly in areas undergoing rapid urban and agricultural transformations. In Meherpur District, Bangladesh, limited research has integrated LULC mapping with predictive models, resulting in a gap in knowledge regarding future land use patterns in this fast-changing region. This study addresses that gap by analyzing LULC changes from 2000 to 2024 using the Normalized Difference Vegetation Index (NDVI), Geographic Information Systems (GIS), and the Cellular Automata (CA) model for predictive analysis. The study reveals significant LULC changes over the 24-year period, including an 18% decrease in vegetation cover, a 6% reduction in agricultural land, and a 4% increase in built-up areas. These trends align with global patterns of urban expansion, often at the expense of agricultural and natural land. Additionally, increases in fallow land (7%) and water bodies (8%) indicate changing land use driven by population growth and infrastructure development. The loss of vegetation, in particular, poses risks to biodiversity, climate regulation, and food security. An innovative aspect of this research is the use of the CA model with the MOLUSCE plugin in QGIS, enabling simulations of future LULC changes up to 2032. This predictive approach offers insights into the impacts of ongoing urbanization, unlike previous studies in Bangladesh that mainly focused on historical LULC changes.

The study suggests several future research directions, including investigating the socio-economic drivers of LULC changes, expanding the geographic scope to neighboring regions, and incorporating advanced remote sensing and machine learning techniques to enhance the accuracy of predictions. In conclusion, this research fills a critical gap in LULC studies in Meherpur by combining historical analysis with predictive modeling, offering valuable insights for policymakers to guide sustainable land use planning amidst ongoing urbanization and agricultural development.

Land use land cover

Supervised classification

Normalized Difference Vegetation Index

Change detection

Geographic Information System

Remote sensing

Meherpur

Land constitutes a crucial element of the global environment and ecosystems, yet its resources face limitations due to burgeoning populations and agricultural demands (Long et al., 2022). Land cover (LC) denotes the physical attributes of the Earth's surface, encompassing both natural entities like water bodies and vegetation as well as anthropogenic features. Conversely, land use (LU) pertains to human activities on the Earth's surface, encompassing activities such as infrastructure development and agricultural practices. In recent times, human actions have exerted significant influence on land use and cover (LULC), profoundly impacting natural resource management. A comprehensive investigation of LULC holds the potential to substantially influence the management of natural resources. The acquisition of precise and adequate information concerning LULC has become imperative in recent years for assessing the social, economic, and environmental implications of these changes and comprehending their ramifications (Roy et al., 2022). Land cover represents the primary attribute of the Earth's surface, delineating its physical state and biotic elements, while land use involves altering land cover to meet human requirements and activities (Prakasam, 2010). Likewise, the process of recognizing alterations in land use and land cover (LULC) over time or between periods is referred to as change detection (Anderson, 1977). Swift transformations in LULC are evident globally, particularly in developing nations, attributable to their substantial dependence on agricultural output and burgeoning populations. Employing the NDVI-Based Classification technique, this research investigates alterations in LULC in Meherpur District, Bangladesh, pivotal for sustainable advancement. The evaluation extends across a substantial timeframe, offering insight into the district's socio-economic and environmental progression. This monitoring is vital for making informed decisions and managing the environment in swiftly changing areas.

A few researchers have examined LULC variations in different parts of the world there are some literature review these given below. A study by Hu et al. (2023) examining southern Punjab, Pakistan, with Landsat and Modis images identified a 14.52% growth in urban areas and a 31.03% reduction in forested regions between 2000 and 2021. Additionally, barren land expanded by 12.87%. These results highlight notable LULC changes over the specified period. Usman et al. (2015) utilized remote sensing data to develop detailed land use maps for the Lower Chenab Canal region in Pakistan, covering the years 2005 to 2012. They assessed accuracy using an error matrix, supplementary data, and previous studies, achieving accuracy rates of 82.83% for winter and 78.21% for summer, highlighting the method's efficacy in detecting land use changes. Ehsan and Kazem (2013) investigated land use and cover changes in Ardakan, Iran, from 1990 to 2006 using NDVI and Landsat ETM + imagery. Their findings revealed that 18.83% of the NDVI values declined by over 10%, while only 1.38% increased, indicating significant changes in vegetation cover over the period.Kindu et al. (2013) assessed land cover changes in Vijayawada, Andhra Pradesh, by analyzing Landsat 8 OLI/TIRS imagery from 2014 and 2020. Through NDVI and PCA, they classified the area into four LULC categories: dense vegetation, vegetation, water bodies, and urban areas. The results showed a 28% reduction in dense vegetation, alongside increases of 10% in vegetation, 23% in urban areas, and 1% in water bodies. The study highlighted the high accuracy of remote sensing and GIS technologies in tracking LULC changes.Landsat TM data was utilized by Yacouba et al. (2010) to monitor changes in Puer-Simao's land cover between 1990 and 1999. The results showed a rise in built-up and agricultural regions and a decrease in forested and barren areas. By combining Landsat TM, NDVI, and DEM data, precise mapping was made possible, even in the face of obstacles like shadow interference. The study's constant land use patterns were found despite the terrain's variation, indicating the efficacy of their approach, which was straightforward and depended on maximum likelihood classification.

The existing body of literature on land use and land cover (LULC) changes has uncovered significant insights, yet several critical gaps remain. For instance, while Hu et al. (2023) documented notable urban expansion and forest loss in southern Punjab, Pakistan, from 2000 to 2021, the study did not explore the socio-economic or environmental drivers behind these changes, nor their impact on biodiversity and ecosystem services. Similarly, Usman et al. (2015) mapped LULC shifts in the Lower Chenab Canal region with high accuracy but only covered a brief seven-year period, lacking long-term analysis or projections, particularly in the context of climate change. Ehsan and Kazem (2013) noted a significant reduction in vegetation in Ardakan, Iran, yet failed to investigate the causes behind this trend. Kindu et al. (2013) highlighted vegetation loss in Vijayawada, India, without addressing mitigation strategies or examining the ecological consequences of urbanization. Yacouba et al. (2010) reported LULC changes in Puer-Simao but overlooked the broader social and environmental impacts. The core issue in these studies is the lack of a comprehensive understanding of the underlying drivers of LULC changes and their broader consequences. Further research is needed to analyze socio-economic, environmental, and climatic influences, assess long-term impacts, and develop predictive models for sustainable land management. Hence, the aims of this research endeavor are to:

(i).To examine Meherpur District's LULC trends between 2000 and 2024.

(ii).To observe changes in vegetation cover over the duration of the study using NDVI. (iii).To determine and evaluate the causes of the observed changes in LULC.

2.1. Study Area

The Meherpur district in Bangladesh's Khulna division (23°44' to 23°59' north latitude and 88°34' to 88°53' east longitude) was the study's primary focus because of its remarkable changes in land use (BBS, 2013). Covering 751.6 square kilometers and encompassing regions like Mujibnagar, Gangni, and Meherpur Sadar.Despite being the smallest district in Bangladesh, Meherpur has experienced noticeable changes in land use and land cover, driven by factors such as population growth and rising temperatures. These significant transformations make it an ideal location for the study.

2.2. Data Collection

Landsat images with a spatial resolution of 30 meters by 30 meters were acquired for the Meherpur district for the years 2000 (Landsat 5 TM), 2008 (Landsat 5 TM), 2016 (Landsat 8 OLI-TIRS), and 2024 (Landsat 9 OLI-TIRS). These images were obtained from the United States Geological Survey (USGS) website https://earthexplorer.usgs. (Accessed on March 27, 2024) for the purpose of assessing land use and land cover changes from January 2000 to March 2024. The selection of Landsat images for February, March, and April prioritized accuracy by minimizing cloud interference. Table 2.1 provides a summary of the Landsat data, outlining their chosen months for their precision and suitability (Hussain et al., 2023).

Table -2.1 properties of satellite image

Access Date	Acquisition	Satellite	Sensor	Band	Path/Row	Resolution (m)	Cloud Cover
28/03/2024	27/02/2000	Landsat 5	TM	(1-7)	138/43, 44	30	0
28/03/2024	21/04/2008	Landsat 5	TM	(1-7)	138/43, 44	30	0
29/03/2024	11/04/2016	Landsat 8	OLI-TIRS	(1-7)	138/44	30	0.19
30/03/2024	08/03/2024	Landsat 9	OLI-TIRS	(1-7)	138/44	30	0.36

3.1 Method of the Study

The study on Meherpur district investigates how perceptions of global warming, land use and land cover (LULC) changes, and climate change have evolved. Initially, data is sourced from the US Geological Survey (USGS) https://earthexplorer.usgs.gov) and is preprocessed using ArcGIS software to prepare for analysis. The Maximum Likelihood Supervised Classification method is employed to categorize the landscape into six distinct LULC classes, which are determined based on Normalized Difference Vegetation Index (NDVI) values. Post-classification, ArcGIS is used to detect changes over time in these LULC classes. To assess the accuracy of the classification results, the Kappa coefficient is calculated, covering the period from 2000 to 2024. The study then utilizes the Cellular Automata (CA) model to forecast future LULC changes. This prediction is carried out using the MOLUSCE plugin in QGIS 2.18.15, an open-source tool. To further analyze changes in NDVI, data is processed and represented using Microsoft Excel and Word. The results are visualized through map creation, which contributes to the overall development of the study's plan and conclusions.

3.2 Process of the work

Landsat images are processed in ArcGIS 10.8 based on the methods of Rahman et al. (2017) and Rani et al. (2018), utilizing image enhancement algorithms to improve quality. Tools such as composite band, copy raster, mosaic to new raster, and clip are employed to standardize images to a 1:25,000 mapping unit, culminating in mosaicking for a comprehensive aerial view.

ArcGIS 10.8 uses a variety of image classification techniques to evaluate changes in LULC, with MLC being the most accurate and commonly used approach (Wu and Shao, 2002).Supervised classification, renowned for its accuracy, was employed to delineate LULC in Meherpur (Saha et al., 2005), utilizing accessible data and local expertise (Mengistu, 2007; Reis, 2008). The research employed Maximum Likelihood Supervised Classification within ArcGIS 10.8 to categorize six types of LULC as outlined by Foody et al. in 1992 and 2002. This approach enhances accuracy by addressing spectral differences. LULC maps for 2000, 2008, 2016, and 2024 were generated in accordance with USGS protocols to ensure uniformity and facilitate comparisons.

This research utilized ArcGIS 10.8 and Google Earth Pro to examine satellite images spanning from 2000 to 2024, detecting changes in LULC. The analysis involved statistical evaluations and accuracy assessments, aided by organized tables for classification, facilitating a comprehensive examination. Diverse techniques before and after classification were employed to ensure a thorough investigation of LULC modifications.

In evaluating satellite image classification accuracy, both overall accuracy and the Kappa coefficient are crucial, as they incorporate user and producer accuracy. To assess classification validity, 30 random ground control points were selected annually, ensuring a minimum distance of 30 meters. These classifications were cross-checked against real land cover data derived from combined baseline datasets. In the year 2000, agricultural land classification achieved a commendable 90% accuracy, with user and producer accuracy reaching 90% and 100%, respectively. The overall accuracy was determined to be 86.67%, accompanied by a Kappa coefficient of 83.52%, based on a sample of 30 observations, which revealed 4 errors.

The Cellular Automata (CA) model was utilized to predict future LULC changes using the MOLUSCE plugin in QGIS 2.18.15, an open-source software. MOLUSCE integrates algorithms like CA and Artificial Neural Networks (ANN) for LULC analysis, modeling, and simulation, with a user-friendly interface for steps such as input, area change analysis, modeling, simulation, and validation. The CA model captures both static and dynamic aspects of LULC changes, enhancing accuracy (Santé et al., 2010). It adopts a bottom-up approach to simulate urban expansion and LULC dynamics (Al Sharif & Pradhan, 2014; Losiri et al., 2016; Yang et al., 2012). The CA model is represented as:

𝑆 (𝑡 + 1) = 𝑓 (𝑆 (𝑡), 𝑁) S (t+1) = f (S (t), N)

Where, 𝑆 (𝑡 + 1) S (t+1) is the system state at time 𝑡 + 1 t+1, influenced by the state probability at time 𝑁 N (Sang et al., 2011). For this study, historical LULC data from 2000 and 2024 were used as dependent variables, with independent variables for input. The ANN predicted LULC transitions with 1000 iterations and a 1-pixel neighborhood. Transition potential images showed land transformation suitability (Balogun & Ishola, 2017). The 2024 LULC map and 2000–2024 transition probabilities were then used to simulate the 2032 LULC map.

(Faisal et al., 2020), NDVI was employed to evaluate vegetation health by analyzing light absorption and reflection. In ArcMap 10.8, preprocessing steps included reprojecting and sub setting to focus on the AOI. NDVI matrices were calculated using Arc Map’s raster calculator, producing temporal maps (for 2000, 2008, 2016, and 2024) to facilitate the analysis of vegetation change over time. NDVI, which is derived from (NIR) and (RED) bands, serves as an indicator of vegetation density, with elevated levels (0.2–0.3) indicating dense forests, intermediate values (0.6–0.8) representing shrub lands and grasslands, and low NDVI (<0.1) indicative of barren terrains (Morshed et al., 2020).

Table 4.1 and Figures 4.1, 4.2, and 4.3 illustrate alterations in LULC between 2000 and 2024 via MLC classification, providing a succinct yet thorough summary of spatial changes occurring at eight-year intervals.

Initially, in 2000, the developed built-up area comprised 11% of the total land area, increasing to approximately 18% by 2008 but returning to 11% by 2016. However, there was a significant recovery, with the developed built-up area representing 15% of the entire region by 2024, reflecting an overall 4% change from 2000 to 2024, as illustrated in Figures 4.2 and 4.3.

Table: 4.1 Meherpur LULC change from 2000 to 2024

LULC	2000		2008		2016		2024		2000 - 2024
LULC	Area (km₂)	Area (%)	Area (km₂)	Area (%)	Area (km₂)	Area (%)	Area (km₂)	Area (%)	Change %
Built up Area	79.09	11%	132.11	18%	77.71	11%	110.92	15%	4
Vegetation	411.70	56%	266.19	36%	105.93	14%	278.12	38%	-18
Agricultural land	208.19	28%	129.73	18%	203.95	28%	160.25	22%	-6
Fallow land	11.89	2%	53.99	7%	304.58	41%	68.59	9%	7
Water bodies	16.25	2%	103.08	14%	16.46	2%	73.82	10%	8
Roads	10.68	1%	52.71	7%	29.20	4%	46.12	6%	5

In 2000, vegetation covered a substantial 56% of the area. However, it experienced a significant decline, dropping by around 36% by 2008 and further diminishing to only 14% by 2016. Nevertheless, there was a notable resurgence, with vegetation expanding to encompass 38% of the total area by 2024, as depicted in Figure 4.2. Despite this recovery, there was still an overall reduction of 18% from 2000 to 2024, as highlighted in Figure 4.3.

In 2000, agricultural land accounted for 28% of the total area. There was an 18% decrease by 2008, but by 2016, it had recovered to reach 28% again. However, by 2024 (as depicted in Figure 4.2 for 2000, 2008, 2016, and 2024), agricultural land had notably diminished to 22% of the entire area, marking a 6% decline overall from 2000 to 2024, as referenced in Figure 4.3.

In 2000, fallow land comprised merely 2% of the area. By 2008, there was a noticeable 7% increase. Surprisingly, by 2016, it had surged dramatically to 41%. However, by 2024 (as depicted in Figure 4.2 for 2000, 2008, 2016, and 2024), it had significantly decreased, accounting for only 9% of the entire area, although still reflecting a 7% increase from 2000 to 2024, as referenced in Figure 4.3.

In 2000, water bodies accounted for 2% of the area. By 2008, they had increased by 14%, only to revert back to 2% by 2016. Subsequently, there was significant growth, with water bodies comprising 10% of the total area by 2024 (referenced in Figure 4.2 for 2000, 2008, 2016, and 2024), indicating an 8% overall increase from 2000 to 2024, as highlighted in Figure 4.3.

In 2000, roads occupied 1% of the total area. By 2008, there was a noticeable increase to 7%. However, by 2016, they had decreased to 4% of the total area. Yet, by 2024, there was significant expansion, covering 6% of the entire area (referenced in Figure 4.2 for 2000, 2008, 2016, and 2024), indicating a 5% overall increase from 2000 to 2024, highlighted in Figure 4.3.

Table: 4.2 LULC area change from 2000-2024km for Meherpur districts.

LULC	2000-2008	2008-2016	2016-2024	2000-2024
BUILT UP AREA	53.02	-54.4	33.21	31.83
VEGETATION	-145.51	-160.26	172.19	-133.58
AGRICULTURAL LAND	-78.46	74.22	-43.7	-47.94
FALLOW LAND	42.1	250.59	-235.99	56.7
WATER BODIES	86.83	-86.62	57.36	57.57
ROADS	42.03	-23.51	16.92	35.44

Assessing the accuracy of remote sensing data plays a vital role in its processing and analysis, as it determines how reliable the data is for users (Fung and Drew, 2022). From 2000 to 2024, user, producer, overall accuracies, and coefficient K were computed for each classification category (a = built-up area; b = vegetation; c = agricultural land; d = fallow land; e = water bodies; and f = roads) to examine the quality of satellite images used in the research. The calculation of accuracy assessment indices for the entire study period (2000–2024) is detailed in Table 4.3.

Table: 4.3 Results of accuracy assessment indices for the current study

LULC	Users Accuracy %						Producer Accuracy%						OA	(K)
LULC	a	b	c	d	e	f	a	b	c	d	e	f	%	%
2000	100	100	85.7	75	71.43	33.33	100	71.42	66.67	100	83.33	100	80%	75.31
2008	66.67	100	100	75	100	33.33	100	66.67	70	100	100	100	83.33	78.60
2016	100	75	100	33.33	100	80	100	100	66.67	50	100	100	86.67	83.52
2024	83.33	66.67	75	80	100	80	100	50	60	80	100	100	83.33	79.84

Table 4.4: Class statistics

Class	2000	2024	Change sq. Km	2000%	2024%	Change%
Built up area	0.02	0.03	0.01	11.09	15.03	3.94
Vegetation	0.11	0.08	-0.04	54.58	37.04	-17.55
Agricultural land	0.06	0.04	-0.01	28.80	21.64	-7.16
Fallow land	0	0.02	0.02	1.65	9.38	7.73
Water bodies	0	0.02	0.02	2.24	10.28	8.04
Roads	0	0.01	0.01	1.64	6.63	4.99

The Land Use and Land Cover (LULC) changes in Meherpur District from 2000 to 2024, simulated using the Cellular Automata (CA) model, demonstrate significant transformations across various land-use categories. The transition matrix in Table 4.5 highlights the probabilities of land transitioning among six distinct classes: built-up areas, vegetation, agricultural land, fallow land, water bodies, and roads.

Table 4.5:Transition Matrix

	Built up area	Vegetation	Agricultural land	Fallow land	Water bodies	Roads
Built up area	0.13	0.36	0.24	0.09	0.12	0.06
Vegetation	0.15	0.37	0.21	0.10	0.10	0.07
Agricultural land	0.16	0.37	0.21	0.09	0.10	0.07
Fallow land	0.12	0.38	0.25	0.07	0.11	0.07
Water bodies	0.16	0.36	0.20	0.09	0.12	0.06
Roads	0.16	0.35	0.22	0.10	0.10	0.06

During the period from 2000 to 2024, built-up areas expanded by 3.94%, growing from 11.09% in 2000 to 15.03% in 2024, which indicates ongoing urbanization in the district. Conversely, vegetation experienced a significant decline of 17.55%, shrinking from 54.58% to 37.04%. This reduction likely points to deforestation or the conversion of vegetation to other land-use types, such as urban or agricultural areas. Agricultural land also saw a decrease of 7.16%, which can be attributed to shifts in land-use patterns, potentially due to urban expansion or changes in agricultural practices. Fallow land and water bodies, which were less prominent in 2000, increased notably by 7.73% and 8.04%, respectively, by 2024. These changes may be driven by either natural factors or human activities, such as seasonal land abandonment or the development of water bodies for irrigation purposes. Additionally, the road network grew modestly by 4.99%, in alignment with infrastructural development that accompanies urban growth.

Table: 4.6 Artificial Neutral Network (Multi-layer Perception)

Artificial Neutral Network ( Multi-layer Perception)
Neighborhood	1 px
Learning rate	0.100
Maximum iterations	1000
Hidden layers	10
Momentum	0.050
Overall Accuracy	-0.01136
Min Validation Overall Error	0.03015
Current Validation Kappa	0.76875

Looking ahead, the LULC map projected for 2032 using the CA simulation model suggests that urban expansion will continue, encroaching further on agricultural and vegetative land. Built-up areas are expected to keep growing, spurred by ongoing urbanization and infrastructure development. Both vegetation and agricultural land are forecasted to continue declining due to urban pressure and shifts in land management practices. The forecast also indicates modest increases in fallow land, water bodies, and roads, reflecting continued infrastructure development and environmental management efforts in the region.

Figure 4.6: Neural Network learning curve

Regarding model accuracy, the Artificial Neural Network (ANN) used alongside the CA simulation achieved a validation Kappa of 0.76875, suggesting good predictive reliability. However, the overall accuracy of the model was relatively low, at -0.01136, which could be due to over fitting or limitations in data quality. Nevertheless, the ANN's ability to minimize validation errors to 0.03015 demonstrates that the model effectively matches observed LULC patterns and transition probabilities. In conclusion, the results reveal notable LULC changes in Meherpur District from 2000 to 2024, primarily driven by urbanization and the conversion of vegetative and agricultural land. The projections for 2032 suggest these trends will persist, presenting challenges for sustainable land management as urban areas continue to grow. Despite some accuracy limitations, the CA model combined with ANN provides valuable insights into future LULC dynamics.

Table 4.7 Calculated NDVI for LULC in Meherpur from 2000-2024

Years	High value	Low value	Mean value	Std. dev.
2000	0.67	-0.55	0.03	0.12
2008	0.33	-0.34	0.07	0.11
2016	0.45	-0.03	0.19	0.08
2024	0.55	-0.08	0.30	0.09

The NDVI changes over the years 2000, 2008, 2016, and 2024 are illustrated in Figure4.9, respectively. In 2000, the NDVI values varied between -0.55 and +0.67 (Figure 4.9), with a mean of 0.03 and a standard deviation of 0.12. By 2008, there was a decrease in the range of NDVI values, ranging from -34 to +0.33 (Figure 4.9), with a mean of 0.07 and a decreased standard deviation of 0.11. Moving to 2016, the range shifted from -0.03 to +0.45 (Figure 4.9), with a mean of 0.19, indicating an increase, and a reduced standard deviation of 0.08. In 2024, there was a further increase in both the lowest and highest values, which ranged from -0.08 to +0.55 (Figure 4.9). The mean value for 2024 was 0.30, representing a continued increase from 2016, and the standard deviation also increased to 0.09 compared to the previous timeframe. These NDVI observations across different years are summarized in Table 4.7.

The NDVI is a robust indicator of vegetation health and productivity, with higher values indicative of lush areas like forests (Ahmed, 2012). Its ratio-based approach mitigates noise from various sources, allowing for reliable assessment of vegetation changes (Hussain et al., 2019). However, declining water availability has led to reduced vegetated areas, impacting ecology and biodiversity. Addressing these challenges is crucial for enhancing agricultural productivity and livelihoods (Jensen and Cowen, 1999).

The study of land use and land cover (LULC) changes in Meherpur District, Bangladesh, from 2000 to 2024 provides insights into the district’s evolving landscape. Using Landsat satellite imagery, the Normalized Difference Vegetation Index (NDVI), and Geographic Information Systems (GIS), the research highlights key land cover changes over a 24-year span. The findings are compared with similar studies from other regions to offer a broader understanding of global trends. Additionally, the study explores the implications of these changes, their limitations, and the use of the Cellular Automata (CA) model to predict future land use patterns.The analysis reveals significant trends in Meherpur’s land use. Over the two decades, vegetation cover decreased by 18%, built-up areas increased by 4%, agricultural land decreased by 6%, and fallow land increased by 7%. These patterns mirror those found in other regions. For example, Hu et al. (2023) studied Southern Punjab, Pakistan, from 2000 to 2021, noting a 14.52% rise in urban areas and a 31.03% decline in forest cover. Both Meherpur and Southern Punjab experienced urban growth and vegetation loss, although the forest decline was more pronounced in Southern Punjab due to regional variations.In Pakistan’s Lower Chenab Canal region, Usman et al. (2015) studied LULC changes from 2005 to 2012, finding similar trends due to agricultural intensification and infrastructure growth. Ehsan and Kazem (2013) reported an 18.83% reduction in vegetation in Ardakan, Iran, from 1990 to 2006, nearly identical to Meherpur’s decline. However, Ardakan did not see an increase in fallow land, reflecting different land management practices. In Vijayawada, India, Kindu et al. (2013) observed a more significant decrease in vegetation and a larger rise in urban areas, highlighting regional differences in urbanization rates.

The implications of these LULC changes are significant for both the environment and local livelihoods. The 18% reduction in vegetation cover raises concerns about biodiversity loss, climate regulation, and ecosystem services like carbon sequestration. The depletion of vegetation could exacerbate global warming by reducing carbon sinks and increasing emissions. Additionally, it may lead to soil erosion and habitat destruction, with long-term environmental consequences. The 6% decline in agricultural land raises concerns about food security in a district dependent on agriculture. The 4% rise in built-up areas and infrastructure development encroach upon agricultural land, which could reduce future productivity. The 7% increase in fallow land and the 8% rise in water bodies suggest a shift in agricultural practices, possibly due to economic factors. While the expansion of water bodies could improve irrigation, it may also point to a shift towards water-intensive farming, which could worsen water scarcity.Urban expansion presents both advantages and challenges. On one hand, it can drive economic growth and infrastructure improvements, but on the other, it leads to the loss of natural habitats and increased environmental pressures. As urban areas grow, local resources such as water, energy, and air quality are under greater strain, which can result in increased pollution and environmental degradation. While the study provides important insights, it also has limitations. One limitation is the reliance on satellite imagery for LULC classification. Techniques like Maximum Likelihood Classification (MLC) and NDVI are generally reliable but may be affected by cloud cover and atmospheric disturbances. Although efforts were made to minimize cloud interference, some inaccuracies may still have impacted the results.

Another limitation is the study’s focus on Meherpur District alone, limiting the ability to generalize the findings to other areas of Bangladesh or beyond. Moreover, the study only covers a 24-year period, which may not capture longer-term changes such as shifts in government policies or the broader impacts of climate change.The study also lacks an in-depth investigation of the socio-economic factors driving LULC changes. While urbanization and population growth are likely contributors, examining other factors like migration, economic policies, or infrastructure development could provide more insight into the underlying causes of these transformations. Understanding these factors is crucial for developing effective land management strategies. The study employs the Cellular Automata (CA) model, which simulates future LULC changes based on the interactions between neighboring land parcels. This model is effective for forecasting urban growth and land use patterns. In this study, the CA model was implemented using the MOLUSCE plugin in QGIS, with LULC data from 2000 and 2024 as inputs. The model predicted that urban expansion will continue, with more encroachment on agricultural and vegetative land by 2032. However, the accuracy of the CA model depends on the quality of the input data and the assumptions made during modeling. The Kappa coefficient for the Artificial Neural Network (ANN) combined with the CA model was 0.76875, indicating reasonable predictive reliability, though the overall accuracy was low (-0.01136), suggesting potential issues with data quality or over fitting. To enhance the accuracy of future CA models, researchers should include more detailed socio-economic data, government policies, and land use regulations. Additionally, advanced remote sensing technologies and machine learning algorithms could improve LULC classification accuracy. Expanding the geographic scope to include surrounding districts could provide a more comprehensive understanding of land use trends across the region. Future research should also examine the impacts of LULC changes on ecosystem services like water regulation, carbon sequestration, and biodiversity to better assess the broader environmental consequences.In conclusion, the study of LULC changes in Meherpur District between 2000 and 2024 highlights significant transformations, including a reduction in vegetation and agricultural land and growth in urban areas. These changes present challenges for environmental sustainability, food security, and community well-being. Despite some limitations, the CA model provides useful predictions for future LULC trends. To better manage land use in the region, future research should focus on improving data accuracy and understanding the socio-economic drivers behind these changes.

The study on Land Use and Land Cover (LULC) changes in Meherpur District, Bangladesh, from 2000 to 2024 reveals significant shifts in land cover, including a decrease in vegetation and agricultural land, and an increase in urban areas. These changes are critical to understanding the districts environmental and socio-economic transformations, highlighting the urgency for sustainable land management as urbanization and infrastructure development continue to influence the region. A notable aspect of this research is its use of advanced remote sensing techniques, particularly the combination of the Normalized Difference Vegetation Index (NDVI) and Landsat satellite imagery. This approach, alongside Geographic Information Systems (GIS), provides a robust and accurate assessment of LULC changes in Meherpur. Additionally, the application of the Cellular Automata (CA) model with the MOLUSCE plugin in QGIS enables more precise predictions of future LULC trends. This innovative integration of tools offers a comprehensive view of both past and future changes, which remains relatively underutilized in similar studies in Bangladesh. The study's findings raise important concerns. The 18% reduction in vegetation is especially troubling, as it impacts local biodiversity and diminishes the region’s ability to sequester carbon, thereby contributing to climate change. The 6% decline in agricultural land poses a threat to food security, given the district’s dependence on agriculture. The 4% increase in built-up areas and urban sprawl not only disrupt local ecosystems but also place increasing pressure on essential resources like water, energy, and air quality.

However, the study is not without its limitations. The accuracy of the LULC classification depends heavily on the quality of satellite imagery, and although efforts were made to mitigate cloud interference, some atmospheric disturbances may still have impacted the data. Furthermore, the focus on Meherpur limits the broader applicability of the findings to other regions in Bangladesh. The absence of a detailed analysis of socio-economic drivers behind the LULC changes, such as migration trends, government policies, and economic pressures, is another shortcoming. Future research should address these gaps by expanding the geographic scope to neighboring regions, providing a more comprehensive understanding of LULC trends across Bangladesh. Additionally, more attention should be given to socio-economic factors that contribute to LULC changes, such as shifts in policy and climate change. Advanced machine learning techniques could be integrated to enhance the accuracy of both LULC classification and predictive modeling. Regular monitoring using high-resolution satellite data and incorporating detailed socio-economic and environmental data into the CA model would improve the accuracy of future predictions, aiding policymakers in developing effective, sustainable land management strategies tailored to regions like Meherpur. In conclusion, this study offers a thorough examination of LULC changes in Meherpur, providing valuable insights into the district’s evolving environmental and socio-economic landscape. Despite its limitations, the innovative use of remote sensing and predictive modeling presents a significant step forward in understanding both historical and future LULC dynamics. Future research should focus on expanding the geographic and socio-economic scope, as well as improving predictive accuracy to better inform land use management decisions.

Author Contribution

Author contribution:Md. Ibrahim Hossain (M.I.H): Writing – original draft, Investigation, Data curation, Conceptualization, Methodology, Resources. Md. Mostafizur Rahman(M.M.R): Writing – review & editing, Validation, Formal analysis, Supervision.

Acknowledgement

AcknowledgmentAuthors thankfully acknowledge that these images were obtained from the United States Geological Survey (USGS) website https://earthexplorer.usgs. (Accessed on March 27, 2024) to assess land use and land cover changes from January 2000 to March 2024.

Al-Saady, Y., Merkel, B., Al-Tawash, B., & Al-Suhail, Q. (2015). Land use and land cover (LULC) mapping and change detection in the Little Zab River Basin (LZRB), Kurdistan Region, NE Iraq and NW Iran. FOG - Freiberg Online Geoscience, 43, 1–32.
Anderson, J M, Hardy, E E, Roach, J T and Witmert, R E. 1976. A Land Use Classification System for Use with Remote Sensing Data. U.S. Geological Survey Professional Paper, No. 964, Washington D. C.: Government Printing Office.
Ahmad,F. Areview of remote sensing data change detection algorithms: Comparison of Faisalabad and Multan Districts, Punjab Province, Pakistan. J. Geogr. Reg. Plan. 2012, 5, 236–251.
Al sharif, A.A., Pradhan, B., 2014. Monitoring and predicting land use change in Tripoli metropolitan city using an integrated markov chain and cellular automata models in GIS. Arabian J. Geosci. 7, 4291–4301.
Athick, A.M.A.; Shankar, K.; Naqvi, H.R. Data on time series analysis of land surface temperature variation in response to vegetation indices in twelve Wereda of Ethiopia using mono window, split window algorithm and spectral radiance model. Data Brief 2019, 27, 104773.
Balogun, I., Ishola, K., 2017. Projection of future changes in landuse/landcover using cel- lular automata/markov model over Akure city, Nigeria. J. Remote Sens. Technol. 5, 22–31.
Congalton, R.G. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sens. Environ. 1991, 37, 35–46.
Ehsan, S., & Kazem, D. (2013). Analysis of land use-land covers changes using normalized difference vegetation index (NDVI) differencing and classification methods. African Journal of Agricultural Research, 8(37), 4614–4622. https://doi.org/10.5897/ajar11.1825
Foody, G M, Campbell, N A, Trodd, N M and Wood, T F. 1992.Derivation and applications of probabilistic measures of class membership from maximum likelihood classification, Photogrammetric engineering and remote sensing 58: 1335-1343.
Foody, G M. 2002. Status of Land Cover Classification Accuracy Assessment, Remote Sensing of Environment, 80:185-201.
Fung,T.; LeDrew, E. For Change Detection using Various Accuracy. Photogramm. Eng.RemoteSens.1988,54,14491454.Availableonline:http://www.asprs.org/wpcontent/uploads/pers/1988journal/oct/1988_oct_1449-1454.pdf (accessed on 20 December 2022).
Faisal, B.M.R.; Rahman, H.; Sharifee, N.H.; Sultana, N.; Islam, M.I.; Habib, S.M.A.; Ahammad, T. Integrated Application of Remote Sensing and GIS in Crop Information System—A Case Study on Aman Rice Production Forecasting Using MODIS-NDVI in Bangladesh. Agriengineering 2020, 2, 264–279.
Hussain, S.; Mubeen, M.; Ahmad, A.; Majeed, H.; Qaisrani, S.A.; Hammad, H.M.; Amjad, M.; Ahmad, I.; Fahad, S.; Ahmad, N.; et al. Assessment of land use/land cover changes and its effect on land surface temperature using remote sensing techniques in Southern Punjab, Pakistan. Environ. Sci. Pollut. Res. 2022, 1–17.
Hu, Y., Raza, A., Syed, N. R., Acharki, S., Ray, R. L., Hussain, S., Dehghanisanij, H., Zubair, M., & Elbeltagi, A. (2023). Land Use/Land Cover Change Detection and NDVI Estimation in Pakistan’s Southern Punjab Province. Sustainability (Switzerland), 15(4). https://doi.org/10.3390/su15043572
Hussain, S.; Mubeen, M.; Ahmad, A.; Akram, W.; Hammad, H.M.; Ali, M.; Masood, N.; Amin, A.; Farid, H.U.; Sultana, S.R.; et al. Using GIS tools to detect the land use/land cover changes during forty years in Lodhran District of Pakistan. Environ. Sci. Pollut. Res. 2019, 27, 39676–39692.
Hussain, S.; Mubeen, M.; Karuppannan, S. Land use and land cover (LULC) change analysis using TM, ETM+ and OLI Landsat images in district of Okara, Punjab, Pakistan. Phys. Chem. Earth Parts A/B/C 2022, 126, 103117.
Hadeel, A.; Jabbar, M.; Chen, X. Application of remote sensing and GIS to the study of land use/cover change and urbanization expansion in Basrah province, southern Iraq. Geo-Spat. Inf. Sci. 2009, 12, 135–141.
Holtz, T.S.U. Introductory Digital Image Processing: A Remote Sensing Perspective, Third Edition. Environ. Eng. Geosci. 2007, 13, 89–90.
Jensen, J.R.; Cowen, D.C. Remote Sensing of Urban/Suburban Infrastructure and Socio-Economic Attributes. Photogramm. Eng. Remote Sens. 1999, 65, 611–622.
Mengistu D., A., and Salami A.T., 2007, Application of remote sensing and GIS inland of Environmental Science and Technology 1(5): 099-109.
Morshed, S.R.M.R.; Fattah, A.; Rimi, A.A.; Haque, N. Surface temperature dynamics in response to land cover transformation. J. Civ. Eng. Sci. Technol. 2020, 11, 94–110.
Masum, K.M.; Alam, M.S.; Al Mamun, M.M.A. Ecological and economical significance of homestead forest to the household of the offshore island in Bangladesh. J. For. Res. 2008, 19, 307–310.
Naeem,M.; Farid, H.U.; Madni, M.A.; Ahsen, R.; Khan, Z.M.; Dilshad, A.; Shahzad, H. Remotely sensed image interpretation for assessment of land use land cover changes and settlement impact on allocated irrigation water in Multan, Pakistan. Environ. Monit. Assess. 2022, 194, 1–18.
Rahman, M.T.U.; Tabassum, F.; Rasheduzzaman, M.; Saba, H.; Sarkar, L.; Ferdous, J.; Uddin, S.Z.; Islam, A.Z.M.Z. Temporal dynamics of land use/land cover change and its prediction using CA-ANN model for southwestern coastal Bangladesh. Environ. Monit. Assess. 2017, 189, 565.
Rwanga,S.S.; Ndambuki, J.M. Accuracy Assessment of Land Use/Land Cover Classification Using Remote Sensing and GIS. Int. J. Geosci. 2017, 8, 611–622.
Rani, M.; Kumar, P.; Pandey, P.C.; Srivastava, P.K.; Chaudhary, B.; Tomar, V.; Mandal, V.P. Multi-temporal NDVI and surface temperature analysis for Urban Heat Island inbuilt surrounding of sub-humid region: A case study of two geographical regions. Remote Sens. Appl. Soc. Environ. 2018, 10, 163–172.
Reis S. 2008, Analyzing land use/land cover changes using remote sensing and GIS in Rize, North-East Turkey, Sensors, 8: 6188 6202
Santé, I., García, A.M., Miranda, D., Crecente, R., 2010. Cellular automata models for the simulation of real-world urban processes: a review and analysis. Landsc. Urb. Plan. 96, 108–122.
Sang, L., Zhang, C., Yang, J., Zhu, D., Yun, W., 2011. Simulation of land use spatial pattern of towns and villages based on CA–Markov model. Math. Comput. Model. Dyn. Syst. 54, 938–943.
Tian, Y.; Tsendbazar, N.-E.; van Leeuwen, E.; Fensholt, R.; Herold, M. A global analysis of multifaceted urbanization patterns using Earth Observation data from 1975 to 2015. Landsc. Urban Plan. 2021, 219, 104316.
Usman, M., Liedl, R., Shahid, M. A., & Abbas, A. (2015). Land use/land cover classification and its change detection using multi-temporal MODIS NDVI data. Journal of Geographical Sciences, 25(12), 1479–1506.
Wu, W. and Shao, G. 2002; “Optimal Combinations of Data, Classifiers, Remote Sensing, 28(4): 601-609.
Yacouba, D., Guangdao, H., & Xingping, W. (2010). Assessment Of Land Use Cover Changes Using Ndvi And Dem In Puer And Simao Counties ,Yunnan Province , China. Report and Opinion, 2(9), 7–16.

No competing interests reported.

Assessing the Rapid Urbanization in Tertiary City of Bangladesh by Land Use and Land Cover Change Detection from 2000 to 2024 through NDVI Based Classification and future forecasting for 2032 by Cellular Automata (CA) model in Meherpur District

Status:

Version 1

Abstract

Figures

1. Introduction

2. Study Area and Dataset

3. Methodology

4. Study Result

5. Discussion

6. Conclusion

Declarations

Author Contribution

Acknowledgement

References

Additional Declarations

Status:

Version 1