Mapping and analyzing temporal variability of spectral indices in the lowland region of Far Western Nepal

Geospatial methods, along with Landsat imagery, have recently garnered unprecedented interest due to their vital role in providing information about the scale and configuration of changes in vegetation, urban areas, and bodies of water. This knowledge is of utmost significance for purposes such as urban and landscape development planning, management of waterbodies, risk assessment, disaster management, and the sustainable management of environmental resources (Uddin et al. 2010; Palacios-Orueta et al. 2006; Gu et al. 2007; Gautam et al. 2015; Karanam & BabuNeela 2018; Roy & Bari 2022). Satellite imagery provides a diverse array of opportunities for swiftly monitoring the environment, particularly in regions where conducting on-site surveys is hindered by factors such as challenging topography, dense vegetation, or other local constraints (Van Dessel et al. 2008; C. Huang et al. 2021; Espinoza-Molina et al. 2022). Remote sensing (RS) satellite imagery stands as a powerful tool with the capacity to benefit local communities and decision-makers by offering timely, cost-effective, and accurate information regarding vegetation, built-up areas, and water bodies (Valor & Caselles 1996; Jat et al. 2008; Singh et al. 2022). The classification of these features not only supports localized planning and resource management but also addresses broader issues related to environmental sustainability, disaster resilience, and informed governance. As technology continues to advance, the importance of satellite imagery in addressing these multifaceted challenges is only expected to grow (Mirchooli et al. 2022; Njumbe et al. 2023).

RS data play a vital role in the examination of environmental processes at both local and global levels. They serve as fundamental resources for identifying changes that have occurred in recent decades. Datasets such as Landsat, Sentinel, and Spot images are immensely valuable for tasks like visualizing, categorizing, and analyzing geographical areas. These datasets can be categorized based on several factors, including resolution, electromagnetic spectrum, energy source, imaging medium, and the number of spectral bands they offer. It's worth noting that higher-resolution satellite data, encompassing spatial, spectral, radiometric, and temporal aspects, result in more accurate classification outcomes (He et al. 2010; Kshetri 2022). The utilization of Geographic Information Systems (GIS) and RS methods has provided us with valuable knowledge about the attributes of alterations in land features (Pattanayak & Diwakar 2018). This includes understanding the scale and spatial patterns of diverse changes in the spectral indices over time.

The acquisition of data through RS demonstrated to be exceptionally efficient when it comes to the mapping and identification of shifts in land use, encompassing areas like agricultural land, urban developments, water bodies, and unproductive terrain (Sun et al. 2012; Saikia & Thapa 2022). Numerous observation satellites of Earth, each offering various levels of spatial and spectral resolution, are available for utilization in these analyses (Van Dessel et al. 2008). Spectral indices, particularly the water index, built-up index, and vegetation index have been employed to identify urban expansion, and vegetation dynamics and visualize water presence areas (Wicaksono & Wicaksono 2019; Ali Shah et al. 2022).

Vegetation indices (VIs) obtained from satellite data find widespread application in ecological studies, the assessment of ecosystem benefits, and the monitoring of land surfaces (Nath 2014; Pattanayak & Diwakar 2018; C. Huang et al. 2021). The NDBI is defined as a composite of the near-infrared (NIR) band and either the middle infrared (MIR) or shortwave infrared (SWIR) band (Kshetri 2022). The primary purpose of the NDBI is to streamline the procedure of delineating urbanized regions. Urbanized areas are accurately delineated by performing mathematical operations on reprocessed images of the NDVI and NDBI, which are derived from Landsat Thematic Mapper (TM) imagery (Zha et al. 2003). The NDBI is computed by taking the ratio of the SWIR to the NIR bands, resulting in index values that fall within the range of −1 to 1 (Estoque & Murayama 2015).

The NDVI, perhaps the most commonly employed vegetation index, applies a spectral analysis technique in RS to monitor the Earth's surface (Shi et al. 2023). The NDVI data derived from Landsat imagery is widely regarded as one of the most valuable datasets for conducting long-term trend analyses of the NDVI (C. Huang et al. 2021). This dataset boasts an extended temporal span of over 30 years and a high spatial resolution of 30 meters. Consequently, it serves as a robust tool for comprehending the historical trends in vegetation growth, monitoring present conditions, and preparing for future environmental changes (C. Huang et al. 2021). The NDVI can be computed by combining the Red and NIR bands of a sensor system (Espinoza-Molina et al. 2022; Kshetri 2022). NDWI is utilized for the examination of aquatic features, with this index making use of data from the Green and NIR bands in RS imagery (Singh et al. 2022).

Examining NDWI, NDVI, and NDBI together at the district level is vital because it gives a complete picture of an area's water bodies, vegetation dynamics and urban growth. At regional and national levels, satellite imagery insights are crucial for formulating policies and strategies concerning urban development, environmental protection, disaster management, and more. Nevertheless, despite the ongoing advancements in spectral indices for swift and accurate classification of vegetation, urban areas, and water bodies in satellite imagery, there remains a notable absence of a comprehensive comparison across these indices when applied to different satellite datasets.

The study area spans rural, semi-urban, and urban settings, and this particular research, which has not been undertaken previously, holds significant value. This study represents one of the first pioneering efforts and aims to demonstrate the potential of combining these spectral indices. Also, it aims to evaluate the changing trends in Spectral Indices over time, with the potential to address critical concerns like urban sprawl, agriculture practices, water resource management, and environmental preservation in the lowland region of Far Western Nepal. This study seeks to fill these existing research gaps and serve as a foundational reference for local policymakers, guiding decisions related to land use, natural resource management, and environmental sustainability. Furthermore, it can set the precedent for future studies with similar objectives in this region. The main objectives of this study encompass two key aspects: (1) connecting extensive RS data to improve our comprehension of the reciprocal effects between urban growth and shifts in waterbodies and vegetation, and (2) investigating temporal dynamics and the interconnections among these indices. To attain these objectives, the research was carried out to analyze the temporal trends of spectral indices and their intricate relationships within the lowland area of Far Western Nepal.

The present study was conducted in the Kailali district (Figure 1), which is situated within the geographical coordinates of 80°15″–81°15″ E longitude and 28° 22″–29° 00″ N latitude. Kailali District lies in the lowlands (terai) region of the Far Western part of Nepal and, is positioned approximately 800 km west of Nepal's capital city, Kathmandu. Encompassing an area of 3,235 km², the district comprises about 40% of the lower Terai terrain, characterized by a relatively flat landscape, while the remaining portion constitutes the Chure hill range. The majority of the district falls within the Tarai belt, spanning an altitudinal range of 109–1,950 m above mean sea level (msl). The district shares its borders with Bardiya and Surkhet to the east, Kanchanpur to the west, Doti and Dadeldhura to the north, and India to the south (Figure 1).

The Terra Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (GDEM V003) were retrieved from (https://search.earthdata.nasa.gov/search) earth data. Landsat-derived images with a spatial resolution of 30 m × 30 m, covering the Kailali district, were procured for the 1991 Landsat 5 Thematic Mapper (TM), 2001 Landsat 5 (TM), 2013 Landsat 8 Operational Land Imager-Thermal Infrared Sensor (OLI–TIRS), and 2022 of Landsat 8 (OLI–TIRS). These images were obtained from (https://earthexplorer.usgs.gov/), the United States Geological Survey (USGS) website and were employed to analyze variation between spectral indices i.e. NDVI, NDBI and NDWI. Specifically, the images selected corresponded to the months of March and April, as these periods exhibited optimal quality and precision in Landsat imagery in the case of Nepal. To ensure the reliability of the data, very less cloud cover Landsat images were chosen for data collection. This approach aimed to maintain a high level of accuracy in the analysis.

The retrieved imageries were projected to the Universal Transverse Mercator (UTM) projection system (UTM Zone 44N). Subsequently, the obtained satellite images underwent processing within the geospatial software environment of ArcGIS 10.8. Following this, a shapefile representing the Kailali District was extracted from the complete scene using the Extract by Mask tool. Tables 1 and 2 summarize an overview of the acquired Landsat data and spectral band information.

This research is conducted in three phases: (1) acquiring RS data, (2) preprocessing the RS data, and (3) analyzing the temporal changes in spectral data and their interrelationships spanning from 1991 to 2022.

This article focuses on three major land cover categories: vegetation, water bodies, and built-up areas, employing three distinct indices – NDVI, NDBI and NDWI, as outlined in Table 3. The NDVI is a fundamental metric for assessing the greenness of the Earth's surface, achieved through a linear combination of the NIR and Red bands (Chen et al. 2013). For NDVI computation, the wavelengths of the NIR and Red bands were utilized (Table 3). NDVI values range from 0 to 1, with higher values indicating denser vegetation cover. The NDBI, on the other hand, automatically delineates urban built-up features (Chen et al. 2013), computed based on the SWIR and NIR wavelengths (as detailed in Table 3). The NDBI values span a scale from −1 to +1, where negative values signify the presence of water bodies and vegetation, positive values indicate the presence of constructed or built-up areas, and slightly positive values correspond to barren soil types (Freidoony et al. 2015).

Table 3

Indication of spectral indices, adopted bands, and respective formulae

Indices Formula .	Reference .
	C. Huang et al. (2021)
	He et al. (2010)
	Espinoza-Molina et al. (2022)

Indices Formula .	Reference .
	C. Huang et al. (2021)
	He et al. (2010)
	Espinoza-Molina et al. (2022)

View Large

The NDWI was employed to determine moisture content within vegetation canopies (Ashwini & Sil 2022), using Landsat 8 imagery Bands 5 and 6 (as specified in Table 3). The NDWI scale extends from −1 (negative) to +1 (positive). Water bodies are depicted by values falling within the range of 0 to +1 (positive), and values closer to 1 signify a high concentration or density of water bodies (Freidoony et al. 2015). Table 3 provides a comprehensive overview of the indices, their corresponding bands, and the formulas used to calculate these indices. The study utilized Landsat TM satellite imagery from March 16, 1991, and April 28, 2001, as well as Landsat 8 OLI/TIRS satellite imagery from April 13, 2013, and March 05, 2022. ArcGIS 10.8 was employed for the processing of all satellite images, and map generation.

Correlation analysis involves examining two or more variables to assess the extent of their relationship, as quantified by the correlation coefficient. The correlation coefficient, denoted as ‘r,’ spans from −1 to +1. Positive values of ‘r’ denote a positive correlation, while negative values indicate a negative correlation (Yang et al. 2023). A larger absolute value of ‘r’ signifies a stronger degree of correlation. In this study, Spearman's correlation was used to understand the relationship between NDVI, NDBI, and NDWI. OriginPro 2023b was utilized to prepare a correlation plot. Trend graphs have been visualized as scatter plots, utilizing x and y coordinates to plot data points. This graphical representation may reveal patterns and relationships among the plotted values, helping us distinguish data trends.

To create a correlation graph between selected spectral indices in ArcMap 10.8, the following steps were performed: First, import the image containing pre-calculated indices values into ArcMap. Next, use the ‘Create Fishnet’ tool through the Data Management tool to generate a grid pattern. After the ‘Create Fishnet’ process is completed, employ the ‘Extract Multi Values to Point’ tool within the Spatial Analysis tool on both the selected images. Subsequently, apply the ‘Clip’ tool to the output produced by the ‘Extract Multi Values to Point’ operation. Within the attribute table of the clipped image, need to notice the addition of two new fields of selected images. Proceed to the ‘Create Graph’ option, select the scatter plot, and designate one image for the x-axis and another for the y-axis for all the indices, respectively. Finally, click ‘Next’ to generate the correlation graph, illustrating the relationship between selected indices. This process resulted in trend graphs for all the indices, facilitating the identification of correlations among the variables. ArcGIS 10.8 was employed for the grid-based analysis, processing of all satellite images, and map generation.

The RS datasets are used extensively for evaluating the NDVI, NDBI, and NDWI variability from 1991 to 2022. The health of vegetation in a particular area can be depicted using the NDVI, which serves as an indicator of the vegetation's condition (S. Huang et al. 2021). In this study, the maximum value of the NDVI from 1991 to 2022 is in increased order, except in the year 2001. Similarly, minimum value of the NDVI from 1991 to 2022 is in increased order, except in the year 2022. These variations in the NDVI values are mainly due to the intense pressure of the human population or as a result of anthropogenic activities (Nath 2014). NDVI values of 1991 imagery indicate the highest and lowest vegetation 0.42 to −0.09. The maximum values indicate the vegetation canopy over an area is highest, and the minimum values are displayed on the water body, barren land, and sandy areas (Ali Shah et al. 2022).

While the NDVI value of the 2001 imaginary shows a maximum value of 0.37 and a minimum value of −0.06 over the study area. The imagery 2013 NDVI represents the highest value on vegetation, agriculture crops and pasture at 0.45, on the other hand, the minimum value showed the NDVI −0.07 over the sandy, barren and built-up areas.

Comparing NDBI of 1991–2022, the built-up is highest in 2001. While the lowest values were computed in 2022 imagery. Hence, NDBI was found to be a good index for distinguishing the built-up areas from OLI data, which is similar to the study carried out by (Ali Shah et al. 2022) in Khangarh Taluka, Ghotki. Figures 5 and 7 demonstrate the range of minimum to maximum NDBI values during 1991–2001–2013 and 2022, respectively.

Between 1991 and 2001, there was a decline in the distribution of higher NDVI values, while the lowest NDVI values increased. From 2001 to 2013, we observed a reverse trend, with the distribution of higher NDVI values increasing and the lowest NDVI values decreasing. From 2013 to 2022, both the highest and lowest NDVI values exhibited an increase in trends. During the 1991–2001 period, both the maximum and minimum NDBI values experienced an increase trends (Figure 7). However, from 2001 to 2022, there was a continuous decrease in the maximum NDBI value, while the minimum NDBI value showed a continuous increase trend. In terms of NDWI, between 1991 and 2001, the distribution of maximum NDWI values increased, while the lowest NDWI values decreased. From 2001 to 2022, both the highest and lowest NDWI values experienced an increase in trends. Notably, the minimum NDWI value decreased continuously from 2001 to 2022 (Figure 7). However, higher NDWI values decreased from 2001 to 2013 and then increased from 2013 to 2022 (Figure 7).

Regarding the temporal variations in spectral indices, it was observed that regions with both high and low values of NDVI, NDBI, and NDWI are distributed throughout the area under study. High NDVI values were predominantly noted in the northern and Middle Eastern sections of the study area, consistent with earlier research (Ali Shah et al. 2022; Espinoza-Molina et al. 2022). These areas are characterized by open impermeable surfaces. As for NDWI, the regions along the river exhibit elevated NDWI values, while the urban areas with exposed impermeable surfaces tend to have the lowest NDWI values, aligning with findings from previous studies (Gautam et al. 2015; Guha et al. 2018; Ali Shah et al. 2022). Nonetheless, notably elevated NDBI values were predominantly detected in the southern, eastern, and Middle Eastern sectors of the district. Conversely, the northern district exhibited the lowest NDBI values, indicating a lesser degree of urbanization, in accordance with prior research findings reported by (Guha et al. 2020) and (Mishra & Garg 2023). Therefore, based on the findings, it is evident that urbanized regions exhibit notably higher NDBI compared to areas covered by vegetation and bodies of water. A similar outcome was also observed by (Ghosh et al. 2019) and (Chen et al. 2022). The spatial variability in thermal patterns within urban areas is influenced by a multitude of factors, encompassing environmental, social, and urban design elements (Ghosh et al. 2019; Chen et al. 2022; Espinoza-Molina et al. 2022). Over the period from 1991 to 2022, there has been a notable upward trend in NDVI across substantial portions of the Kailali district. The result obtained in the present study is consistent with the previous study (Baniya et al. 2018), carried out in Nepal. Typically, elevated NDVI values, whether measured through integrated NDVI or the annual maximum NDVI, signify favorable plant productivity and biomass conditions (Baniya et al. 2018). A positive trend in NDVI implies an improvement in ecological and ecosystem conditions. This positive shift in NDVI is attributed to ecological restoration efforts and recent environmental changes, which have significantly contributed to the expansion of green vegetation in the Kailali district. Similarly, In Kailali district, there has been a noticeable upward trajectory in NDWI from 2001 to 2022, particularly in Southern, central and eastern parts. Typically, elevated NDWI values, whether assessed through integrated NDWI or the annual maximum NDWI, signify satisfactory conditions for waterbodies. It's worth noting that NDWI levels across all land covers tend to be higher in January and April when the soil is moist (De Alwis et al. 2007). Since the data extracted for this study pertains to the months of March and April, it explains the higher values observed during the 2001–2022 period.

The increase and decrease trends in NDVI, NDBI, and NDWI values from 1991 to 2022 may be attributed to a combination of factors, including, alterations in land usage, fluctuations in climatic conditions, enhancements in agricultural methodologies, the regeneration of natural vegetation and advancements in satellite technology (Gautam et al. 2015; Chapungu & Nhamo 2016; C. Huang et al. 2021; Ali Shah et al. 2022; Sarkar & Patra 2022; Singh et al. 2022; Frimpong et al. 2023).

Land use modifications, such as urbanization or alterations in vegetation cover, might have led to reduced water absorption and increased surface water, resulting in higher NDWI values (Ali Shah et al. 2022). Changes in precipitation and hydrological conditions over this time span could have influenced water availability, potentially leading to elevated NDWI values. The decrease in NDBI values from 1991 to 2022 may be attributed to several factors related to urbanization and land use. One possible explanation is that urban development and built-up areas expanded significantly between 1991 and 2022, resulting in a higher proportion of impervious surfaces like buildings and roads. As a result, the NDBI values decreased because the index measures the presence and extent of built-up or urban areas in comparison to natural or non-built-up areas. To ascertain the precise reasons for the decrease, a comprehensive analysis considering urban growth patterns, land use changes, and policy developments in the specific area of interest would be necessary. Furthermore, advancements in RS technology and data processing methods may have contributed to more accurate and higher-resolution NDVI, NDWI and NDBI measurements in 2022 compared to 1991.

Similarly, in the year 2001, we observed analogous correlation trends. Notably, there was a substantial positive correlation (r = 0.64, p < 0.05) between NDBI and NDWI, indicating a connection between urban development and water. Conversely, significant negative correlations (r = −0.96 and r = −0.76, p < 0.05) were found between NDVI and both NDWI and NDBI, respectively, during the same year, implying an inverse link between vegetation and both water and built-up areas. A similar pattern persists in 2013, with a significant positive correlation (r = 0.82, p < 0.05) between NDBI and NDWI and significant negative correlations (r = −0.96 and r = −0.090, p < 0.05) between NDVI and NDWI and between NDVI and NDBI during the same year. The year 2022 also exhibits a substantial positive correlation (r = 0.82, p < 0.05) between NDBI and NDWI, compared with significant negative correlations (r = −0.98 and r = −0.88, p < 0.05) (Figure 8) between NDVI and NDWI and between NDVI and NDBI.

These findings underscore a clear trend across the years, revealing a positive association between NDWI and NDBI, signifying a parallel increase in urban development and water bodies. Conversely, the NDVI consistently displayed a negative correlation with both NDWI and NDBI over the study period, indicating a divergent trend in vegetation health. These results were similar to previous studies conducted in Kathmandu Valley, Nepal, metro cities of India, Noida City, India, Hyderabad City (Telangana), India, Pabna municipality in Bangladesh, Khangarh taluka, district Ghotki, Pakistan, Densely Populated Cities of South Asia, Greater Arba Minch Area, Rift Valley, Ethiopia, Al-Hashimiya district, Iraq, and Florence and Naples city, Italy (Chen et al. 2006; Kikon et al. 2016; Guha et al. 2018; Pattanayak & Diwakar 2018; Sarif et al. 2020; Shahfahad et al. 2020; Abir & Saha 2021; Jothimani et al. 2021; Maharjan et al. 2021; Access, n.d.; Ali Shah et al. 2022).

Globally, the expansion of urban areas is driven by human activities like changing land use, urbanization, population growth, and industrialization. Land use changes have varied effects on vegetation, water bodies and urban areas. When analyzing NDVI, NDWI, and NDBI from 1991 to 2022, a negative relationship between vegetation with urbanization and waterbody was observed due to terrain size and vegetation cover. Therefore, this study discusses and compares the results of both indices in four years (1991, 2001, 2013, and 2022). These results highlight the inverse relationship between vegetation, water bodies, and urban areas; as one increases, the others decrease, consistent with prior studies (Palacios-Orueta et al. 2006; Gu et al. 2007; Sun et al. 2012; Pattanayak & Diwakar 2018; Sarif et al. 2020; Shahfahad et al. 2020; Jothimani et al. 2021; Ali Shah et al. 2022; Singh et al. 2022; Frimpong et al. 2023; Gupta et al. 2023). Researchers are now focusing on sustainable management because unchecked urban growth could potentially wipe out vegetation and vital resources. This poses a significant threat to biodiversity and the existing ecosystem. Figure 9 illustrates the scatter plot depicting the relationships between spectral indices spanning from 1991 to 2022.

Although this study presents a captivating exploration of the interplay among spectral indices within the largest lowland district in Far Western Nepal, it is essential to acknowledge its inherent limitations. One noteworthy limitation is the absence of an investigation into the spatial variations and dynamics of the relationships among these spectral indices. To address this limitation and gain a more comprehensive understanding, future research should incorporate a spatial analysis. This would involve examining how the associations between spectral indices change across different geographical areas within the district. Such an approach could reveal valuable insights into the spatial distribution of these relationships and help identify any localized patterns or anomalies. Additionally, the study's reliance on a limited set of RS indices for exploring these connections is another aspect that warrants attention. To enhance the interpretation of the relationship between spectral indices, future research endeavors should consider utilizing a broader array of RS indices. This broader dataset can provide a more nuanced view of the complex interactions between spectral indices and their implications for the district's characteristics. This study focused solely on one city, and future research should extend its analysis to multiple urban areas.

The study relied on the Landsat satellite sensor, but the efficacy of the research could be further assessed by considering alternative satellite sensors with varying resolutions, such as MODIS, ASTER, Sentinel, and others. Furthermore, it is advisable to expand the scope of future research by delving into various factors that can influence these relationships. Factors such as population growth, socio-economic parameters, and natural driving forces play a pivotal role. By considering these factors, researchers can better grasp the multifaceted nature of the relationships between spectral indices and the evolving landscape of the district in Far Western Nepal.

The study employed Landsat 5 TM and Landsat 8 OLI/TIRS data to compute NDVI, NDBI, and NDWI trends spanning the years 1991–2022. This study explores the intricate relationships among these indices within this timeframe. The study revealed that temporal variations in spectral indices demonstrated diverse patterns across the study area, with high NDVI values in the northern and Middle Eastern sections, elevated NDWI values along the river, and notably higher NDBI values in the southern, eastern, and Middle Eastern sectors, illustrating the influence of urbanization and various factors on these indices. The results unveiled that regions characterized by human settlement, barren terrain, open impermeable surfaces, and sandy expanses exhibited significantly elevated NDBI values in 1991, 2001, and 2013 compared to 2022, while NDWI and NDVI values were consistently lower in 1991, 2001, and 2013 compared to 2022.

Furthermore, an intriguing finding was the robust negative correlation observed between the NDVI and both the NDBI and NDWI from 1991 to 2022. Conversely, a robust positive correlation was consistently identified between the built-up index and the water index during this same time frame. These outcomes underscore the utility of NDVI, NDBI, and NDWI as invaluable parameters for diverse applications, such as monitoring vegetation health, estimating crop acreages, assessing urban development, and delineating aquatic features. In summary, this research emphasizes the potential of utilizing a combination of vegetation, built-up, and water indices derived from Landsat TM and OLI/TIRS bands to facilitate the mapping of vegetated zones, urban expanses, and water bodies. The findings provide valuable insights into the temporal dynamics of these spectral indices, emphasizing their crucial role in effective land management, fostering sustainable urban development, facilitating environmental monitoring, and managing water resources sustainably. In light of these findings, it is recommended that governmental bodies and policymakers allocate support for further research endeavors focusing on these indices in conjunction with environmental factors. This approach would enable a deeper understanding of evolving land use patterns, particularly in urban and agricultural domains. Additionally, it is crucial to consider the influence of factors such as rapid population growth, landscape alterations, and socio-economic parameters in future research endeavors. These considerations will undoubtedly contribute to informed decision-making and sustainable development planning.

The author expresses deepest gratitude to the USGS for the cost-free satellite images of the study region.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.