ML‐Based Hybrid SAR and Optical Image LULC Mapping and Change Analysis With Variations in the Air Quality of the Imphal Valley, North‐East India

Imphal Valley, situated in Manipur, India, stands as an intermontane valley of great ecological significance, notably hosting Loktak Lake. This research delves into the Land Use Land Cover Change (LULCC) within the Imphal Valley from 2016 to 2021 and assesses their impact on air quality across distinct land cover types. We used Sentinel‐1 & 2 data, ALOS PALSAR Digital Elevation Model, and applied Random Forest (RF), a machine learning algorithm for effective land use and land cover (LULC) mapping. Additionally, Sentinel‐5P data was utilized to monitor air quality parameters (CO, HCHO, NO2, SO2, Aerosol Index) spanning 2019 to 2021. The overall accuracies for the LULC maps employing a k (k = 3) fold approach for accuracy assessment varied between 88% and 92%, with corresponding Kappa coefficient ranging from 0.85 to 0.90. Noteworthy trends emerged from our analysis, revealing an increase in settlements and horticulture farms and a decline in forested areas and phumdis (floating biomass). Our findings highlight a mean concentration of CO ranging between 0.045 mol/m2 and 0.055 mol/m2 in different land cover types during February and March (2019–2021). Furthermore, we observed maximum mean HCHO, NO2, SO2, and aerosol index concentrations in March. Pollution levels surged during forest fires and shifting agriculture seasons, while aerosol levels declined during the lockdown period. This integrated approach emphasized on the comprehensive analysis of the dynamic interplay between LULCC and air quality in the Imphal Valley. This intricate relationship between LULC changes and air quality dynamics in the Imphal Valley, contribute to our understanding of the environmental dynamics in this ecologically vital region.


Introduction
Land use and land cover changes (LULCC) have widespread impacts on ecosystems, air quality and climate patterns in the long run, making it a global concern.Urbanization, deforestation, agricultural expansion, forest fires, and shifting cultivation practices driven by human activities have accelerated LULCC rates, posing environmental challenges (Borgohain et al., 2023;Gupta et al., 2023;Puri et al., 2011;Roy et al., 2022;Yilmaz et al., 2023).LULCC significantly affects air quality by influencing the dispersion and concentration of air pollutants in the atmosphere (Yilmaz et al., 2023;Yin et al., 2019).Understanding the relationship between LULCC and air quality is crucial for sustainable development.Many previous studies have primarily relied on optical remote sensing data for LULC mapping (Gupta & Shukla, 2016, 2020;Mohajane et al., 2018;Niraj et al., 2020), which can be limited by cloud cover and atmospheric conditions.In this regard, the utilization of Sentinel 1 Synthetic Aperture Radar (SAR) data offers a unique advantage, as it can overcome these limitations and provide accurate information on land cover changes, even in cloudy or rainy conditions (Tavares et al., 2019).
Numerous studies have explored the relationship between LULCC and air pollutant concentrations, providing insights into this relationship's complex dynamics and implications.Global and regional studies have highlighted the significant contribution of LULCC to the release of air pollutants (Yin et al., 2019).For example, Raihan (2023) emphasized the substantial impact of deforestation on carbon dioxide (CO 2 ) emissions, stressing the need for effective land management strategies.Similarly, research by Huang et al. (2023) demonstrated the role of LULCC, particularly the conversion of forests to agricultural land, in increasing GHG emissions.It is crucial to study LULCC and their impact on air quality.Air pollutants, such as Carbon Monoxide (CO), Formaldehyde (HCHO), Nitrogen dioxide (NO 2 ), and Sulfur dioxide (SO 2 ), significantly affect Earth's energy balance and climate patterns (Borgohain et al., 2023).CO originates from the combustion of fossil fuels, biomass burning, and the atmospheric oxidation of methane and other hydrocarbons (Copernicus Sentinel data, 2023;Landgraf et al., 2022).NO 2 and NO-together usually referred to as nitrogen oxides (NO x = NO + NO 2 )-are important trace gases that enter the atmosphere due to anthropogenic activities (notably fossil fuel combustion and biomass burning) and natural processes (such as microbiological processes in soils, wildfires and lightning) (Yang et al., 2021).The primary HCHO source in the remote atmosphere (Earth's atmosphere that are distant from direct human and local pollution source) is CH 4 oxidation.While sulfur dioxide (SO 2 ) emission influences the climate by contributing to radiative forcing by forming sulfate aerosols (Sentinel 5P ATBD-SO 2 User Documents, 2022).Activities like fossil fuel combustion, forest fire and changing land use practices have led to a substantial increase in emissions of these pollutants.
Forest fires release large amounts of smoke, particulate matter, and gases, while shifting cultivation involves burning vegetation, contributing to air pollution (Prasad et al., 2000).In 2022, the world witnessed devastating heatwaves (Ma et al., 2022), which have regional and global implications, impacting air quality in neighboring areas.Sustainable land management practices, including forest fire prevention and control, are crucial for mitigating these impacts and reducing air pollutant emissions (Smith et al., 2016).The Paris Agreement, which aims to reduce and mitigate greenhouse gas emissions, recognizes the implications of LULCC on air quality (Enoh et al., 2023).
In the northeastern region of India, forest fires and shifting cultivation practices have been identified as significant contributors to the rise in air pollutants (Gupta & Shukla, 2022).Due to this practice of shifting cultivation, the air in this region is deteriorating, and it is anticipated that the Imphal Valley, a bowl-shaped valley surrounded by mountains all around, will be most affected.Loktak Lake, one of the largest freshwater lake in South Asia is located in this valley.It is a Ramsar site renowned for its floating islands, called Phumdis, composed of a diverse mixture of soil, vegetation, and organic matter.Loktak Lake is listed in the Montreux Record as it is among the wetland sites of international importance where human interference or pollution have caused changes in its ecological character.The lake is home to the beautiful Keibul Lamjao National Park (KLNP), the world's only floating park (Gupta & Shukla, 2023).
Moreover, the densely populated valley surrounding Loktak Lake provides shelter for over two-thirds of Manipur's population.Localized studies in specific regions, like the Imphal Valley, Manipur, are necessary to capture the intricate dynamics of LULCC and its influence on concentration of air pollutants.Thus, by studying the impact of forest fires (Borgohain et al., 2023), shifting cultivation (Puri et al., 2011), and LULCC on air pollutants, this research is in line with the goals of the Paris Agreement.
Therefore, we aim to contribute to the existing body of knowledge by investigating the relationship between LULCC and air pollutants in Imphal Valley, Manipur, using Sentinel 1 (SAR), Sentinel 2, and ALOS PALSARderived topographic data.This paper aims to quantify LULCC between 2016 and 2021 using Machine Learning based model integrating optical and SAR data sets, to investigate the relationship between land use changes and concentration of air pollutants (CO 2 , NO 2 , SO 2 , HCHO, aerosol).Further this research assesses the contribution of different land use categories toward emission of air pollutants and provides recommendations for sustainable land management practices.By examining the interactions between LULCC and air pollutant emissions in the study area, including forests, agriculture, urban areas, and wetlands, this research seeks to enhance understanding of regional air pollution dynamics and support environmental management strategies and policy recommendations for sustainable land use and air pollution control.

Study Area
Imphal Valley, situated in the lesser Himalayan region of North East India (Figure 1), is an intermontane valley located within the Manipur state.Spanning an area of approximately 1864.44 sq.km., this valley is characterized by its geographical coordinates between 24°-25°N latitudes and 93.50°-94.25°Elongitudes (Gupta & Shukla, 2022).The valley is traversed by numerous small rivers, such as the Imphal River and its tributaries Iril, Thoubal, Khuga, and Sekmai, originating from the surrounding hills and flowing in a north-south direction (Kumari et al., 2018).Imphal Valley is famous for Loktak Lake and KLNP (the only floating park in the world).The lake is renowned for its unique floating islands, locally known as phumdis.Phumdis are heterogeneous masses consisting of vegetation, soil, and organic matter in different stages of decomposition that have consolidated into a solid form.Loktak Lake is among the 48 wetland sites worldwide listed in the Montreux Record, which documents Ramsar sites experiencing or at risk of ecological character changes (Das Kangabam et al., 2017).With a subtropical monsoon-type (Köppen: Cwa) climate (Sahu et al., 2020), the Imphal Valley experiences maximum summer temperatures ranging from 32 to 34°C, while winter temperatures can drop to a minimum of 1-2°C.According to the 2011 census (Home|Government of India, 2024), Imphal Valley accounts for approximately 60% of the population of Manipur, that is, 1,713,476 persons (Das Kangabam et al., 2017;Gupta & Shukla, 2022).

Satellite Used
The Sentinel 1 (COPERNICUS/S1_GRD) data set contains satellite images from the Copernicus Open Access Hub's Sentinel-1 mission's Ground Range Detected (GRD) products (Copernicus Sentinel data, 2023).Sentinel-1A and 1B are positioned 180°apart in the same orbit, providing global coverage with a revisit time of 6-12 days.The C-band SAR data offers a spatial resolution ranging from 5 to 20 m.The Level-1C processing of Sentinel-1 GRD products includes noise removal, signal calibration, compensation for instrument-specific characteristics, data focusing, transformation into ground range coordinates, radiometric calibration, and terrain correction.This results in georeferenced ortho-images in the Universal Transverse Mercator projection with the WGS84 datum (Copernicus Sentinel data, 2023;Steinhausen et al., 2018).Details of the data used are given in Table 1.The Sentinel-2 Multi-Spectral Instrument (MSI) data set comprises satellite images from the Sentinel-2 mission's Level-1C products, available through the Copernicus Open Access Hub.Sentinel-2A and Sentinel-2B, positioned 180°apart in the same orbit, enable a global revisit time of up to 5 days.The Sentinel-2 MSI instrument captures images in 13 spectral bands, ranging from visible to shortwave infrared, with spatial resolutions between 10 and 60 m (Sentinel-2 User Handbook, 2023; Steinhausen et al., 2018).Details of the data used are given in Table 1.We obtained the SO 2 , NO, CO, HCHO, CH 4 and aerosol data from the Sentinel 5P satellite (ESA, 2018), launched in October 2017 and orbiting at approximately 824 km, using the Tropospheric Monitoring Instrument (TRO-POMI) instrument.Details of the data used are given in Table 1.

Data Used
The data utilized in this study comprises Level-1 GRD products generated from Focused SAR data.This data was processed through detection, multi-looking, and projection to ground range using an Earth ellipsoid model (Sentinel-1 SAR technical guide, 2023).Additionally, Sentinel-1 (S1) ortho-corrected SAR data (Sentinel-1 User Handbook, 2023) was employed, specifically utilizing the S1 GRD scenes processed with the Sentinel-1 Toolbox.These data were combined with Sentinel-2 (S2) (Sentinel-2 User Handbook, 2023) surface reflectance tier 1 data set from 2016, up to 2021.We also used ALOS PALSAR data (https://search.earthdata.nasa.gov/)EARTH-DATA, 2015 for LULC mapping, utilizing bands B1-B7 and the Normalized Difference Vegetation Index (NDVI) (Whyte et al., 2018), features from S2, as well as VV and VH from S1, and Digital Elevation model (DEM) (Chen et al., 2017), and slope from ALOS PALSAR.
The Random Forest (RF) classifier was employed using all bands (Bl-B7) from S2, NDVI, DEM, slope, VV, and VH from S1 for the LULC mapping process.The analysis was conducted on the Google Earth Engine (GEE) platform and further analyzed using ArcMap 10.5.Later, we utilized Sentinel-5P near-real-time (NRTI) (January 2019-December 2021) to monitor air quality, employing near real-time data for NO 2 , SO 2 , CO, HCHO, and UV aerosol index.It allowed us to interpret the changes in trace gas concentrations across the Imphal Valley.The NRTI product is available within 3 hr after acquisition, and off-line (OFFL) is available within a few days, so we preferred to use NRTI products.

Methodology
We applied a quantitative research approach to examine LULCC between 2016 and 2021.This involves comparing LULC classes in two time periods, 2016 and 2021, and calculating the spatial extent of changes in each class.Specifically, we determined the difference in land area (sq.km.) for each category, within Imphal Valley.Then, we tried to relate it to variations in the air quality parameters (such as CO, HCHO, NO 2 , and SO 2 concentrations) within Imphal Valley, Manipur, from 2019 to 2021.The methodology (Figure 2) integrates remote sensing analysis, geospatial techniques, and measurements of trace gas concentrations.

Data Pre-Processing
First, we have imported Sentinel 1 SAR, Sentinel 2, and ALOS PALSAR data sets into GEE.Then, all the data sets were preprocessed.SAR data is preprocessed to enhance image quality and reduce noise and speckle.We have applied a morphological mean filter with a 50 m smoothing radius to remove speckle noise on Sentinel 1 data (Gupta & Shukla, 2023), while cloud masking was applied to Sentinel 2 (as mean of original cloud cover percentage in Sentinel 2 was found as 26.341) images to remove cloud cover.Then the data was filtered from February to April for each year and clipped according to geometry.This data generated in pre-processing stage was used for classifying the satellite image to prepare the LULC map.

LULC Mapping
A number of approaches have been put forth and used for LULC mapping using remote sensing.These approaches include pixel-based image classification methods (Aguirre-Gutiérrez et al., 2012;Belgiu & Csillik, 2018;Hussain et al., 2013); object-based image analysis (Blaschke, 2010;Ez-zahouani et al., 2023), and its integration with fuzzy soft classification Fuzzy-OBIA (Feizizadeh et al., 2013;Feizizadeh, Garajeh, et al., 2021), and machine learning techniques (Liao et al., 2019;Shetty, 2019;Zhao et al., 2020).RF algorithm, which is a recommended ML approach used for LULC classification due to its resilience, ability to handle high-dimensional data, and flexibility to multi-class situations (Shetty, 2019).When it comes to LULC, RF does exceptionally well at producing clear and understandable results, especially when dealing with noisy or ambiguous data.RF benefits when it integrates Sentinel data and SAR.Sentinel's spectrum information is complemented with SAR data, which captures vegetation structure and surface roughness, improving classification accuracy (Niculescu et al., 2018;Rajah et al., 2019).The algorithm's temporal data handling capabilities complement Sentinel's timeseries data, guaranteeing classification consistency between time points.Additionally, the advantage of RF's resilience to cloud cover is that it may be combined with SAR, which is less impacted by clouds (Zhao Earth and Space Science  (Gupta & Shukla, 2023;Santos et al., 2022).Hence, we have chosen the RF algorithm to prepare the LULC map.Random training samples for various land cover types: waterbody, phumdis, settlements, horticulture farms, paddy fields, forests/vegetation, and bareland were taken as 300, 160, 260, 400, 180, 285, and 30 respectively.These samples were used to train the RF algorithm with these parameter settings: number of trees: 300, variable per split (Null), minimum leaf population: 1, bag fraction: 0.5, maximum nodes: null, seed: 0. An accuracy assessment was conducted to evaluate the accuracy and precision of the land use maps.

Accuracy Assessment
The accuracy evaluation utilized a subset of testing data, constituting 25% of the total data points.
The assessment employed traditional methods based on the kappa coefficient.Overall accuracy (Equation 1) was determined by quantifying the number of pixels correctly classified in the image under examination.
The formula (Equation 2) used to calculate the kappa coefficient was as follows (Petropoulos et al., 2015;Verma et al., 2020).

Overall Accuracy
Here, n ii is the number of correctly classified pixels and N is the total number of pixels.
Here, r represents the number of rows in the matrix.X denotes the number of observations in row i and column i (the diagonal elements).The terms x i+ and x ii ˙refer to the marginal totals of row r and column i, respectively.N represents the total number of observations.

Air Quality
Trace gas concentrations, specifically CO, HCHO, NO 2 , SO 2 and aerosol index, were obtained on the classified LULC feature collection.We initially upscaled the LULC maps to a resolution of 1113.2 m, equivalent to the resolution of air quality parameters.Subsequently, we extracted the feature collection for each class.The air quality parameters were then obtained for each class feature collection and plotted.These measurements were obtained at regular intervals from 2019 to 2021 to capture monthly temporal variations of these contaminants in the Imphal Valley.

LULCC Analysis
The overall LULCC analysis revealed significant transformations in the study area.Paddy fields experienced a slight decrease of 3.04% in area, suggesting possible shifts in agricultural practices or land use patterns.Horticulture farms exhibited a 1.85% increase in area, indicating potential expansion of horticultural activities.Settlements demonstrated a notable 3.49% increase in area, highlighting urbanization and population growth.Phumdis experienced a 1.43% decline in area, potentially due to environmental factors or human activities.
Forest/Vegetation was reduced by around 2%, implying potential land degradation or conversion impacts.At the same time, the water body appears to be at the same level (Table 2).
A series of thematic maps were generated to depict the spatial distribution of land cover classes to visually highlight the extent and changes in land cover types, such as paddy fields, horticulture farms, settlements,

Variations in Air Quality Parameters
The overall assessment of air quality parameters across different land cover types revealed intriguing patterns and potential associations with LULCC.CO concentration demonstrated an upward trend during shifting cultivation or forest fires season, with elevated levels observed from January to April in 2019, 2020, and 2021.This increase in CO concentration can be attributed to the burning of vegetation during shifting cultivation and forest fires (Borgohain et al., 2023).Furthermore, the HCHO, aerosol index, and NO 2 concentrations analysis demonstrated similar trends across the different land cover types and seasons with slight variations in SO 2 .
Line graphs were utilized to depict the temporal trends of air quality parameters to the shifting cultivation or forest fires season and the lockdown period.These line graphs show the mean concentration of NO 2 , SO 2 , CO, HCHO and aerosol index, over time, specifically from January to April 2019, 2020, and 2021, and during the September 2020 to May 2021 lockdown period.The line graphs facilitate the visual identification of any consistent patterns or significant fluctuations in air quality parameters during these periods (Figures 4a-4e).It highlights the intricate relationship between land cover types, seasonal factors, and human-induced events, providing information about into the fluctuations in concentration of these air quality parameters., the maximum in this seasonal trend of 3 years.Almost the same CO concentration is seen for all the classes.In contrast, the dip in the line graphs reaches 0.029 mol/m 2 , consistent among all classes in the study area (Figure 4c).Mean HCHO concentration peaks in March-April; maximum values are seen for phumdis.The mean HCHO concentration lies between 0.00005 mol/m 2 and 0.00035 mol/m 2 .
In March 2020, it was found to be 0.00024 mol/m 2 , and 0.00023 mol/m 2 in March 2021.While for forested land mean HCHO concentration reached the maximum in March with a 0.0002 mol/m 2 value.And lowest values were seen in August for all land types in 2020 (Figure 4d).Aerosol index found its maximum at 0.191193 in December 2021 after the rise on March 21, attaining a value of almost 1.091193 in all classes.Aerosols were found more in forested land, horticulture and paddy fields than in phumdis (Figure 4e).

Discussion
The study demonstrated that the RF ML algorithm is suitable for complex and heterogeneous landscapes, as it effectively captures the spatial patterns and spectral information in the data (Fu et al., 2017).The study by Schulz et al. (2021) also utilized RF to classify different LULC classes using Sentinel-2 satellite imagery, achieving a high classification accuracy.Combining separate data sets enables the use of bands from various electromagnetic spectrum regions (Nuthammachot & Stratoulias, 2019).Thus, combining the optical and SAR data helped us to avail the benefits of both data types (D'Addabbo et al., 2016;Nuthammachot & Stratoulias, 2019;Tavares et al., 2019).Each sensor can better classify certain land cover types over others.SAR data alone is more efficient in delineating built-up areas, whereas the optical data is inept in classifying the natural vegetation (Jiang et al., 2020).Hence the combination of both types of data may reduce the problem of misclassification.The feature importance analysis revealed that VV, VH, NDVI, and DEM were the most influential features in the classification process.VV and VH bands from Sentinel 1 captured the radar backscatter signals and delineated the information about the surface characteristics more efficiently.NDVI, derived from Sentinel 2 bands, helped assess vegetation density and health; additionally, including DEM and slope as topographic features contributed to the classification accuracy by considering the terrain characteristics (Chen et al., 2017).The significance of these selected features highlights their role in effectively differentiating LULC classes.Thus, the inclusion of VV, VH, NDVI (Whyte et al., 2018), and DEM-derived products (e.g., slope, aspect, etc.) (Chen et al., 2017) proved successful in accurately classifying LULCC in the study area.The robustness and reliability of the classification results can be attributed to the careful selection and integration of these influential features.
The analysis of LULCC in the study area revealed significant transformations in various land cover types.The observed changes indicate a transformation in settlement areas, horticulture farms, phumdis, vegetation, and paddy fields.There has been a noteworthy increase in settlement areas.This can primarily be attributed to factors such as population surge, urbanization, and tourism growth.Imphal Valley is characterized as an intermontane valley and the only region in Manipur that caters to most of the state's agricultural, economic, and other administrative needs.Between 2001 and 2021, the population of Manipur surged by more than 33% (Home| Government of India, 2024), along with the rapid urbanization of more than 35% (De & Devi, 2023).Such a significant increase in the human population within the valley exerts immense pressure on the natural resources, such as forests, lakes, etc.Moreover, the development of small-scale industries and associated facilities like schools and hospitals also contributed to the growth of settlement areas.Such developmental activities played a significant role in supporting the livelihoods of local communities, along with horticulture and tourism industry serving as crucial economic drivers.With the increasing number of tourists visiting popular places like Loktak Lake and KLNP, there is a growing demand for accommodations such as hotels and cottages.Thus, the expanding built-up areas and residential zones in the Imphal Valley are driven by the growing demand for socio-economic development and tourism.As a result, there has been a transformation of land cover types and an increase in human settlements in the region.
In 2023, Gupta and Shukla's study showed grave concern regarding the most pronounced changes in phumdis, due to the increase in settlement on them.Specifically, the developmental plans focused on various sectors within the basin have strained the ecosystem services and biodiversity of the Loktak Lake (Manipur Loktak Lake (Protection) Act, 2006;Meitei & Singh, 2018).In the study, we observed decline in phumdis, flora, and paddy fields.Phumdis, which contain plant, soil, and organic material, have decreased significantly.Anthropogenic activities, such as encroachment, habitat degradation due to eutrophication, causing disturbances to the natural ecology, may be responsible for this decline (Kangabam et al., 2019).Thus, activities such as illegal settlements and pollution are disrupting the natural ecology of phumdis, making them more susceptible to degradation and natural decline (Devi et al., 2021).Increased use of fertilizers and its runoff into the Loktak Lake would be causing harm to phumdis and other aquatic life.Changes in water levels, flow rates and sediment deposition can disrupt the growth and sustainability of phumdis leading to their further decline.Climate change related factors such as altered precipitation (Roy et al., 2022), increased temperature and extreme weather events can affect phumdis ecosystems.Overexploitation and unsustainable practices such as overfishing, overharvesting vegetation, and unsustainable resource extraction from phumdis can deplete them and disturb the delicate ecological balance (Anand & Oinam, 2020;Kumari et al., 2018).Horticultural farms have experienced a slight increase.This can be attributed to the growing interest in horticultural practices like growing fruits, vegetables, flowers, herbs, and other plants.
Increased economic benefits, market demand, quick returns, short growing seasons, and favorable climatic conditions in the region may drive the expansion of horticulture farms.Also, Loktak Lake has high monsoonal fluctuations (Tripathi et al., 2018), which gives the farmers the benefit of cultivating the temporarily reclaimed land.Furthermore, due to illegal encroachment on the periphery and in the Loktak lake, the people have deliberately reclaimed the land from the lake and are using it for agricultural purposes (Laishram & Dey, 2013).Deforestation has directly and indirectly contributed the most to the rapid and large increase in settlement areas.Thus, land clearing and the transfer of wooded regions to other land uses may all contribute to the decline in vegetation coverage (Table 2).Decrease in paddy fields may be attributed to less monetary benefits and long growing seasons relative to vegetables and herbs (Gupta & Shukla, 2023).The intricate relationship between land use and air quality dynamics is well studied all over the world (Heald & Spracklen, 2015;Sun et al., 2016).Major contributors to declining air quality are the anthropogenic sources and increasing population.The population of the Imphal valley has increased significantly in the last two decades (K.J. Singh, 2018).Hence, the expanding urban footprint witnessed a marked decline in forest cover and the unique phumdis of Loktak Lake.The expansion of settlement areas fostered the increased usage of vehicles of all kinds.Thus, the changes in demography and urbanization cumulatively altered the LULC of Imphal Valley which eventually escalated the emission of many harmful gases.
Various air quality parameters such as NO 2 , SO 2 , CO, HCHO are studied.We analyzed NO 2 , SO 2 , as these are good indicators of the larger group of pollutants called NO x and SO x oxides.Seasonal variations in the concentrations of all five contaminants can be seen in graph 4. A more pronounced dip can be observed during July and August due to the impact of monsoons, and this is further diluted by the clean air (Gaur et al., 2014).Due to urban expansion, the agricultural land on the periphery of the towns was converted into residential areas and roads, which led to increased vehicular traffic, resulting in elevated emissions of NO 2 , SO 2 , CO, and aerosols (A.Kumar & Pandey, 2013).More than three-fourths of the population of Manipur relies on agriculture as one of the main sources of livelihood (Reimeingam, 2017).Additionally, there are no industries in the Imphal Valley that would significantly contribute to environmental pollution (S.I.  , 2015).Paddy is grown in 82% of the total cultivated area of Manipur (Thangjam & Jha, 2020).After harvesting paddy in the months of October and November, the farmers start burning paddy stubble in November and early December (Sahu et al., 2021), which mainly increases the concentration of aerosols like HCHO, CO, and SO x in the atmosphere (P.Kumar & Joshi, 2013), which is captured in graph 4. Since November, the weather in the valley, becomes drier and windy.Thus, farmers start preparing for Jhum cultivation during mid-January to February, by clearing a patch of forest (Thong et al., 2020), burning the twigs, branches and waste from the cleared area, releasing substantial amounts of CO, NO x , SO x , and aerosols in the surroundings (Bhagawati et al., 2015).This augments the spike in the concentration during the fire season of February and March, which is also reflected from Figure 4.
The surge in population and increased tourism at Loktak Lake has led to illegal encroachments upon the lake (Paonam & Chatterjee, 2022).This is slowly destroying the ecosystem of phumdis.Phumdis decompose faster due to eutrophication induced by the amplified use of pesticides and fertilizers in surrounding agricultural and horticultural farms (Kangabam et al., 2019).Thus, the heightened algal bloom in response to these chemicals contributes to formaldehyde (HCHO) emission upon decomposition.As discussed earlier, many phumdis are being utilized for residential purposes, and the wetland is being reclaimed for both paddy cultivation and horticultural activities.Since wetlands serve as natural sinks for pollutants, their deterioration could potentially release stored carbon, thereby contributing to elevated levels of CO, HCHO, and other substances.Also, the imprudent waste management practices, like the discharge of solid and liquid waste in Loktak and the piling up of solid waste and small landfills in the valley (N. S. Kumar, 2017), add another layer, augmenting the emission of NO 2 , SO 2 , CO and other pollutants.According to ISFR 2017, 77.69% of the total geographical area of Manipur was under forest cover, which decreased to 74.34% in 2021 by ISFR 2021.We too have seen on the LULC map that there is a slight loss in agricultural land, along with a reduction in the forest cover.The loss of forest cover disturbs the delicate equilibrium between emission and absorption of aerosols which might trigger an imbalance, further deteriorating the air quality.Covid 19, undoubtedly a nightmare for the whole world, also has at least one merit to see.Due to the lockdown, most of the daily commuting was halted, which resulted in less vehicular use and hence less vehicular emissions in the valley (Behera et al., 2022;Ghahremanloo et al., 2021;Ghasempour et al., 2021).The same is depicted by the graph.One can clearly see the sharp trough or the dip in the emission of all the contaminants during the March-April 2020 that is, the Covid I season.
Imphal Valley is affected by natural wildfires (Mamgain et al., 2023;Sanjenbam, 2023) which is usually initiated by anthropogenic activities.Wildfire is a direct cause and an indirect consequence of LULCC (Viedma et al., 2006), yet the changes in LULC impact the wildfire risk (Vilar et al., 2016).Hence, not only it is indirectly related to the emission of gases like NO 2 , SO 2 , CO, HCHO and aerosols but it is also an important contributor to the elevated levels of all the five pollutants.In February and March farmers start preparing the fields.We can notice in Figure 4, that all the five pollutants show a spike during mid-February and March, which is the fire season in the valley.In essence, the Imphal Valley stands at a crossroads where human induced LULC changes intricately interplay with the emission dynamics of critical air pollutants.As population increases and urbanization burgeons, the valley grapples with visible transformations and subtle, insidious alterations in air quality.Addressing these issues mandates a nuanced approach, where mitigating emissions is not detached from sustainable land-use policies and conscientious waste management but is viewed as integral components of a unified environmental strategy.The story woven by this data illuminates a path forward, urging strategic interventions to ensure a harmonious co-existence between human activities and the delicate ecosystem of the Imphal Valley.The findings highlight the need for sustainable land management practices, conservation efforts, and effective urban planning strategies to balance development with environmental preservation.

Conclusion
This study provides valuable insights into the intricate relationship between LULCC and their impact on air quality in the ecologically significant Imphal Valley.This study used SAR and Sentinel 2 data to effectively map LULC (2016)(2017)(2018)(2019)(2020)(2021).Band combinations B1-B7 (S2), NDVI(S2), VV and VH(S1), DEM(ALOS) and Slope (ALOS) were utilized to classify the region accurately.The feature importance analysis revealed that VV, VH, NDVI, and DEM were the most influential features in the classification process among all other elements.The Earth and Space Science 10.1029/2023EA003176 GUPTA ET AL.
accuracy assessment of the land cover maps demonstrates the robustness of our methodology, with overall accuracies ranging for the 2020 LULC Map found to be 92% and Kappa coefficient values of 0.86.Further Sentinel 5P data, namely NO 2 , SO 2 , CO, HCHO, and UV aerosol index, were used to monitor air quality parameters from 2019 to 2021.Urbanization, forest fires and shifting cultivation activities contributed to the deterioration of air quality, releasing pollutants and negatively affecting the surrounding environment.These findings underscore the need for effective measures to mitigate air pollution, aligning with the United Nations Framework Convention on Climate Change (UNFCCC) goals and the international effort to combat climate change.To advance our understanding of the complex interplay between LULCC and air quality in the Imphal Valley, future research should delve into various avenues.First and foremost, an exploration of long-term trends in LULCC and air quality is essential, offering a more comprehensive perspective over an extended period.Investigating the factors influencing the observed shifts, such as the increase in settlements and horticulture farms and the decline in forested areas and phumdis, will provide valuable insights into the underlying dynamics.Advanced data integration is imperative for enhanced precision, involving the incorporation of additional satellite data sources or advanced sensors, as well as exploring the potential of emerging technologies like high-resolution satellite imagery and advanced ML algorithms.A more in-depth temporal analysis, capturing seasonal variations and considering the influence of climatic events, will contribute to a nuanced understanding.Validation through ground truth or field studies is crucial to bolster the reliability of LULC maps and air quality assessments.Additionally, exploring policy implications for regional development and urban planning, while acknowledging the constraints of remote sensing data and the impact of training points on LULC quality, will guide future research endeavors.

Figure 1 .
Figure 1.Map of study area Imphal Valley showcasing the land cover types visible on Google Earth within this region used to prepare the land use and land cover.(a) Map of India, (b) Map of Manipur, (c) Map of Imphal Valley, (d) Waterbody, (e) Phumdis, (f) Settlement, (g) Horticulture farm, (h) Paddy Field, (i) Forest/Vegetation, (j) Bare land (figures d-g is taken at same 100 m scale).

Figure 2 .
Figure 2. The methodology used in the study comprises processing in mainly two software, Google Earth Engine and ArcMap.

Figure 3 .
Figure 3. (a) Land use and land cover Maps derived from Random Forest (RF) ML model with seven distinct classes waterbody (rivers and wetland), phumdis (mass of vegetation, soil and organic matter), settlement (built-up areas, industries, factories, roads), horticulture farms (fruits, vegetables, flowers, herbs and other plants), paddy (rice crop), vegetation (trees and forested area), bare land (barren land).(b) The figure represents the vicinity of Loktak Lake, zooming into the map to demonstrate the successful classification performance of the RF algorithm in delineating key features such as Imphal River, phumdis, circular rings within Loktak Lake, floating Keibul Lamjao National Park situated on phumdis, and the village of Thanga located within the lake.The figure also includes corresponding crowd-sourced images sourced from various online media platforms.Source (a): Source: https://www.taleof2backpackers.com/loktak-lake-manipur/ news story 24 February 2023, Source (c and d): https://www.jpl.nasa.gov/images/pia22369-loktak-lake-india,news story 7 May 2018.

Figure 4 .
Figure 4.The visual analysis shows variation of Mean NO 2 , Mean SO 2 , Mean CO, Mean HCHO, and Mean AI in different land use and land cover types (Water, Forest, Paddy, Horticulture farms, Settlement, Phumdis).The line graph displays temporal trends of the mean concentration of these parameters.

Table 1
The Table Below Provides Comprehensive Details on the Data Utilized in the Study, Encompassing Information About Sentinel Satellite Images Including Their Spatial Resolution, Bands, Image Collection and Data Source 2 https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S5P_NRTI_L3_SO2GUPTA ET AL.
al., 2020).It is better than other classifiers regarding classification output accuracy and robustness to noise in comparison to other classifiers as it achieves better classification results GUPTA ET AL. et Notably, NO 2 concentration consistently increases during shifting cultivation or forest fires from January to April 2019, 2020, and 2021.Phumdis have the highest NO 2 concentration during these events.The purple line indicates phumdis showed a higher concentration of SO 2 than other classes from January to April 2020.Forest showed a higher concentration of SO 2 during fire season from February 21 to April 21 than other classes.Mean CO concentration rises during shifting cultivation or forest fires season (January-April 2019-2021) due to burning vegetation and forest fires, emitting substantial CO.During January-April 2019-2021, HCHO levels increased due to shifting cultivation practices and forest fires, releasing HCHO into the atmosphere.Forest fires also increased HCHO levels through organic matter and vegetation combustion.Aerosol Index increases during shifting cultivation/forest fires (January-April 2019-2021), indicating an environmental impact.Shifting cultivation (Controlled burning) releases aerosols.Forest

Table 2 Table Presents an
Overview of the Changes in Land Use and Land Cover Area Extent (sq.km and %) for Various Classes From 2016 to 2021 2 .Phumdis experienced its peak in March every year in 2019 it reached 0.40 mol/m 2 , in 2020 it reached 0.45 mol/m 2 and in 2021 it reached 0.55 mol/ m 2 .In 2021, it reached 0.055 mol/m 2 Singh et al., 2019).Hence, most pollution due to 2 , SO 2 , and CO can be attributed to vehicular emissions.The only thermal power plant in Leimakhong (Central Electricity Authority, 2021) in the northwestern part of the valley might contribute minutely to the NO 2 levels.Out of the 22,327 sq.km geographical area of Manipur, 12.98% (2,895 sq.km.) is used for cultivation, and 52% of it is confined in the valley (Official Website of Manipur NO