Measuring the control of landscape modifications on surface temperature in India

Abstract This study highlights the land use/land cover change, and its implications on the long-term changes in various hydrometeorological parameters in India during 2001–2018. The study exhibited a decrease in grassland (15.1%), barren land (10.2%), and shrubland (9.5%) in contrast to an increase in deciduous broadleaf forest cover (27.4%) and croplands (25.2%) in India. The study exhibited 31.7% of the geographical area observed a low warming (<0.8 °C), in contrast to 52.1% of the area observed a low cooling (> −0.8 °C) in India over the observation period. The transformation of (a) barren land to shrubland and cropland (Western India) and (b) cropland, and grassland to forest cover (Central India) led to the cooling effects. While Indo-Gangetic Plain and North-Eastern parts observed warming effects due to greenness and wetness change. The findings highlight the relevance of bio-geophysical and land-climate feedback, which may help in developing integrated and effective climate mitigation and adaptation strategies.


Introduction
The variation in the atmosphere's lower boundary is mainly associated with land use/land cover change (LULCC) (Zipper et al. 2019), which affects climate at local to global scales (Findell et al. 2007). Though, great attention has been given to significant environmental changes concerned with the atmospheric compositions in recent times in contrast to anthropogenic influences on regional climate. The irreversible alterations in land use/land cover are driven by anthropogenic activities including industrialization, rapid population growth, urbanization, and deforestation (Eigenbrod et al. 2011;Gollin et al. 2016;Piao et al. 2020). The accelerated population growth (361-1221 million during 1951-2011) in India over several decades led to the most important perceptible changes in LULC (Roy and Roy 2010). Post-Independence, the green revolution (the mid-1960s) (Chakravarti 1973;John and Babu 2021) to meet the high demands of food grain production eventually led to land degradation, soil nutrient depletion, and deforestation. These bring about the altercation of warmness, moistness, carbon, and energy between the earth's surface to the atmosphere (Bonan 1997;Bonan 2008). Also, a reduction in global forest cover by 0.90% (31.62% to $30.72%) during 1990-2016 considerably disturbed ecology and wildlife habitat. Nevertheless, greening has increased over the decades in India (Lal et al. 2019), which is primarily due to a rise in agricultural activities (Chen et al. 2019).
The increase in carbon storage, contributing to combating climate change, is directly linked with an increase in vegetation (Bastin et al. 2019;, which increases the precipitation in the tropics and sub-tropics and intensifies the hydrological cycle (Pitman et al. 2011). While anomalous growth of the urban population in the past few decades induced high consumption of wood (76%) for industrial, residential, and commercial purposes (Brown 2001). The transformation of vegetated cover to urban surfaces alters the hydrological cycle ) and biogeochemical cycle, which directly impacts global climate (Pataki et al. 2006). In contrast, wastes generated from cities lead to air and water pollution , altering biogeochemical cycles at local, regional, and global scales (Luck et al. 2001). It is expected that the newly expanded built-up surfaces will invade predominantly agricultural land (for urban land of $ 50%-63%) followed by forest area (for urban land of $26%-30%) ) thus regarded as a primary concern for environmental change. This will induce a direct impact on food production under various socioeconomic pathways ) and also affect global mean surface temperatures in response to the alterations incurred in bio-geophysical parameters (Prevedello et al. 2019).
Previous studies reported LULCC driven triggering of a series of changes in hydrometeorological observations as, an increase in Leaf Area Index and conversion of arid to non-arid land, which led to an increase in precipitation between 2002 and 2015 . The change in forest cover has a considerable impact on moisture supply of the summer monsoon as observed through simulations of climate models (Paul et al. 2016). Also, the increase temperature in India has been attributed to the large-scale transformation of forest to cropland, while a decrease in temperature referred to the conversion of grassland to cropland . The change in albedo and evapotranspiration governs the change in near-surface temperature, as dense forest has a lot more evapotranspiration compared to the cropland, while grassland has less water content than cropland. Although water content in cropland is a short-term phenomenon but creates a long-term change in the climate. The LULCC impact on climate is quite uncertain as LULCC in the form of deforestation led to a cooling effect in mid-and high latitudes, whereas in tropical and subtropical regions it contributes to warming during winter and spring (Pitman et al. 2009;Lawrence and Chase 2010). Apart from the forest, cropland, and grassland, urban regions also play a dominant role in altering the microclimate conditions ultimately influencing the regional and global climate system. The micro-climate, particularly in an urban environment, largely goes un-observed and un-attended today due to the difficulty to observe spatial patterns of temperature in the urban and peri-urban regions. The rapid urbanization exacerbates irreversible alterations of natural surfaces (vegetation and grassland) (Diksha and Kumar 2017; Tripathy and Kumar 2019; Kumari et al. 2022) and intensifies the urban heat island (UHI) effect (Kafy et al. 2021;Das et al. 2021;Xie et al. 2022). UHIs change the urban thermal environment and lead to a series of problems by affecting the regional climate, urban hydrology, air quality, and altering the human thermal comfort (Qiao et al. 2019;Ren et al. 2022). The combined changes in LULC create a global concern as land surface temperatures have been increasing globally due to climate change and attributed to anthropogenic influences on a great extent  researchers have attempted to study land surface temperature at the country level, which is either limited to the city-scale (Kayet et al. 2016;Sahana et al. 2019) or focused on some specific homogenous regions. We hypothesize that the vegetation vigour decreases due to higher anthropogenic influence that leads to an increase in land surface temperature. This modifies the various hydrometeorological properties and has potentially negative effects on land surfaces and climate. Therefore, in the present study, (a) the changes in land use/land cover have been investigated at a pan-India scale, and (b) the intrinsic relationship between LULC change and LST and various other hydro-meteorological parameters were evaluated.

Data used
Various sources of datasets viz., observation, reanalysis, and satellite products were used to analyse the changes in LULC and land surface temperature in India. The International Geosphere-Biosphere Programme (IGBP) classification of Moderate Resolution Imaging Spectroradiometer (MODIS) MCD12Q1 LULC (https://cutt.ly/xjXlqBr) was considered for LULC change analysis between 2001-2010-2018. MODIS aqua (MOD11C3) and terra (MYD11C3) day and night datasets were used to analyse LST change during 2001-2018. ERA 5 land reanalysis data (latent heat flux, sensible heat flux) from European Centre for Medium-Range Weather Forecasts (ECMWF) was used to analyse the change in flux due to modification of agriculture and vegetation changes during 2001-18 and used to deduce its relationship with LULCC. Tropical Rainfall Measurement Mission (TRMM) 3B42v7) precipitation data (https://giovanni.gsfc.nasa.gov/giovanni/), MODIS Aqua and Terra evapotranspiration data, and Global Land Data Assimilation System (GLDAS) root zone soil moisture data have been used for analysing the trend with significance and to establish a relationship with a different variable. Detailed information on different datasets is provided in Table S1.

Methodology
The data for LULC change (as a footprint of anthropogenic influence) and LST were acquired from MODIS at a pan-India scale for the post-2000 periods and analysed on a yearly basis. Initially, MODIS MCD12Q1 LULC yearly product was used to analyse the LULC change and LULC transformation over the past 18 years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018). The selection of MODIS LULC datasets was made considering its consistent sensors over time as well as the availability of a variety of products (viz., reflectance data, vegetation indices, biophysical products) and thematic products, which are used to describe different LULC classes with a quality assessment per pixel (Garc ıa-Mora et al. 2012). Due to considerable underestimation of urban and built-up classes for the entire India in the MODIS-based LULC, urban change mapping is not recommended using MODIS-based LULC datasets. Since the focus of the present study is mainly focused on the impacts of agriculture and vegetation cover change, the LULC product of MODIS was used.
LST was retrieved from MODIS and its monthly product was converted into a yearly mean to analyse the change in surface temperature in the last 19 years using the Theil-Sen slope and significance test using Mann-Kendall (M-K). Before performing trend analysis of data, pixel quality checks were performed, and it was observed that $5% of pixels were bad, and 2% were missing. Each month's missing pixel was filled using a Harmonic Analysis of Time Series (HANTS) algorithm. HANTS algorithm was initially applied to NDVI by Roerink et al. (2000) to extract the phenology by generating the cloud-free data and it worked on fast Fourier transformation (FFT) (Xu and Shen 2013). In this study, the HANTS algorithm was applied using MATLAB 2017a (The Mathworks Inc., Natick, MA, USA), and the code used for the above analysis was provided by Abouali (2020). It applies a least-square curve fitting procedure based on harmonic components (sines and cosines), and amplitude-phase defined by an iterative process. The hi/lo was set at low since the cloud influence will lower the LST. The range of valid values was set from À70 C to 70 C, and the fit error tolerance was set at 3 C. Seasonal fluctuations must have followed in the time series, so the Degree of Dependence was set at 105 to achieve the best fitting results.
The change in vegetation vigour (greenness) was estimated using the Normalized Difference Vegetation Index (NDVI) and analysed by calculating the percent change between the years 2001-2003 and 2016-2018. The periodic estimation (average of three years' NDVI) was considered to reduce the yearly biases. For various meteorological parameters, trend analysis and significance were estimated using the Mann-Kendall (M-K) and Theil-Sen slope estimator in the National Center for Atmospheric Research (NCAR) Command Language (NCL). To infer the relation between various bio-geophysical parameters, correlation analysis is performed in Google Earth Engine (GEE), and code can be accessed through GEE developers (https://bit.ly/3oDKUzx).
The LULC change including change in vegetation vigour and its implications on LST was evaluated in a spatial environment and analysed to deduce the possible relationship between the variables at the country scale over the years.
India experienced a major land transformation from forest cover to cropland (27,510 km 2 ), followed by shrubland to cropland (22,272 km 2 ), grassland to cropland (83,919 km 2 ), and barren land to cropland (6,288 km 2 ) during 2001-18 (Tables 1 and 2; Figure S1). The large-scale land transformation was evident in the barren land, which was primarily transformed into shrubland (18,858 km 2 ), grassland (6,288 km 2 ), and snow cover (5,444 km 2 ) ( Table 1 and S2). Also, the increase in the urban area was observed primarily at the cost of cropland (1,099 km 2 ), and forest cover (305 km 2 ). Forest cover (combined with evergreen needleleaf forest (ENF), evergreen broadleaf forest (EBF), deciduous needleleaf forest (DNF), and mixed forest (MF)) was observed increasing and primarily transformed from grassland (27,764 km 2 ), cropland (35,253 km 2 ) and barren land (434 km 2 ) during the period (Figure 1).  Indian region is characterised by large scale spatio-temporal variability in meteorological parameters and has a large scale interannual and inter-decadal variability in temperature (Krishnan et al. 2020). Thus, the trends of LST over the pan Indian region were retrieved using MODIS satellite data for the period 2000-2018. LST is the skin temperature of the earth's surface directly modulated by the change in LULC. LULCC affects boundary layer dynamics and modifies surface and near-surface meteorological variables, leading in microclimatic variations and changes in mesoscale weather patterns (Paul et al. 2018). LULCC altered the landscape temperature patterns owing to multiple anthropogenic influences (Cui et al. 2017) and evidently increased LST two-fold as much as global temperatures since pre-industrialisation (1850-1900) (Jia et al. 2018). In India, a moderate to high decrease ($À0.4 C to À1.2) in land surface temperature was observed in major parts (central, southern, and extreme northern parts) during 2001-2018. In contrast to the Indo-Gangetic Plains (IGP), western and extreme southern parts observed a moderate rise ($0.4 C to 0.6 C) in LST, with statistically high significance (Figure 2a). The seasonal change pattern exhibits the overall comparable pattern of changes as observed at the annual scale barring very few exceptions. Although few exceptions exist in the IGP and western India (WI) during different seasons. In IGP, minor variation in terms of a slight decrease in LST ($0.2 C) was observed during the winter season (DJF), in contrast to a higher increase of LST ($1.5 C-1.6 C) during the pre-monsoon season (MAM) and monsoon season (JJA) along with a decrease in NDVI (by 5% to 10%) and increase in the evapotranspiration (by 2-3 mm/year) during 2001-2018. In the post-monsoon season, Western India experienced a decrease in LST (by $0.4 C-0.6 C) with an increase in greenness (NDVI by $10-20%), evapotranspiration (2-4 mm/year), LHF (by > 0.5 w/m 2 ) in contrast to decrease in SHF (by 0.5 w/m 2 ) (Figures 3 and 4). The relationship between various biophysical and bio-geophysical parameters was analysed that exhibits a negative correlation between NDVI -LST (À0.8 to À0.9), and evapotranspiration -LST (À0.6 to À0.8) over the observation years ( Figure 5 and Table 3).

Discussion
Major changes in forest cover were observed in the north-eastern states, particularly in the Himalayan regions, where natural forests were rapidly transformed into temporary agricultural lands (Jhum farming), leading in a drastic change in regional phytomass (Choudhury et al. 2013). The expansion of agricultural lands in response to the rising demand improved accessibility, and technological advancement resulted in irreversible loss of natural vegetation, affecting agrobiodiversity and sustainability (Lambin 1997). In contrast, better conservation measures, afforestation activities, agroforestry as observed primarily in eastern (Chhattisgarh, Jharkhand, Bihar, Odisha), central (Madhya Pradesh), western (Goa, Gujarat), north-western (Uttarakhand), and upper southern parts (Karnataka, Andhra Pradesh) led to the improvement (gain) in forest cover (canopy density) ( Figure 6) (Hansen et al. 2020).
A slight decrease in LST in IGP during the winter season (DJF) was attributed to the rising concentration of aerosol optical depth (Mandariya et al. 2020;Kutty et al. 2020). While a higher increase in LST during pre-monsoon season (MAM) and monsoon season (JJA) was attributed to increasing in sensible heat flux (SHF) (0.5-0.6 w/m 2 ) (Figure 4c) together with a decrease in precipitation (À1 to À5 mm/year) ( Figure 1c) and NDVI (5-10%) in the region during 2001-2018 (Figure 3). Although the decrease in NDVI during the monsoon season is linked to increased incidences of agricultural drought conditions and delays in the crop sowing in major parts of IGP . The higher increase of SHF in the monsoon season is mainly observed due to a decrease in NDVI (Bounoua et al. 2000), and precipitation leading to the warming of land surfaces (Myhre et al. 2018). This has elevated the evapotranspiration in the region (Byrne and O'Gorman 2015;Milly and Dunne 2016), resulting in the rise of stress in root zone soil moisture (Sepaskhah and Ilampour 1995;Anderson and Kustas 2008). The changes in LST during the post-monsoon season (SON) have conformity to the annual change in the major parts of the Indian region barring a minor exception in western India, where a decrease in LST is associated with the increased precipitation (by $2 mm/year but not significant). Similar results have also been reported during the monsoon and post-monsoon seasons in the western arid zone of India (Gupta and Jain 2020;Maharana et al., 2019;Mishra et al., 2020;Saini et al., 2020), with an increase in evapotranspiration ($1-1.5 mm/year) and root zone soil moisture ($3 mm/year (significant)). Following all the hydro-meteorological parameters, an increase in vegetation cover was observed in western India with an increase in greenness (NDVI by $10-20%) leading to a decrease in SHF (by 0.5 w/m 2 ) and an increase of LHF (by > 0.5 w/m 2 ) (Figure 4).
The significant parts of upper south India (SI) and central India (CI) observed a cooling in temperature primarily due to the significant land conversion from grasslands to croplands. Parts of western India (WI) were also converted from shrubland to croplands, and recorded an increase in precipitation in the region, leading to a significant cooling of land surface temperature (Figure 2c). In contrast, a warming trend was observed in the extreme western part due to extensive coverage of the desert landscape (Thar desert),   causing enhanced local warming in the region. It is to note that the cropland acts cooler in contrast to the leaves of grassland and shrubland, due to the higher availability of leaf water content and less surface roughness in croplands. In India, demand and supply for irrigation and intensive agricultural activities have tremendously increased due to the advancement of machinery and tools in recent decades with the increased support of local governments (Roy et al. 2007;Rodell et al. 2009). Similar results of cooling trends are also reported due to the accelerated irrigation activities in India during the post-green revolution (Roy et al. 2007). At the same time, Central India (Madhya Pradesh) observed a significant transformation of cropland to forest cover and grassland leading to the warming trends (Table S3).
Since Central India has the highest forest cover in the country, the region observed an increase in forest cover due to effective measures of the government policies to protect forests. Although the increase in forest cover leads to increased surface roughness and therefore led to warming impacts, various densely forested regions observed cooling due to increased evapotranspiration. The western part of Central India and eastern part of Western India observed an increase in precipitation with no significance though a high evapotranspiration rate with high significance (surface roughness gets suppressed by increasing evapotranspiration) ( Figure 2) and increased NDVI (5-30%) (Figure 3), led to the cooling effects. A similar observation was recorded in sub-tropics and tropic regions of the globe (Green et al. 2017;Devaraju et al. 2018).
The snow cover (2411.80 km 2 ) in high-altitude regions of northern India led to a decrease in land surface temperatures during 2001-2018 (Table S3; Ladakh) due to the high albedo ( Figure 4) as also reported by Kumar et al. (2019) based on modelling. The land surface temperatures significantly increased in IGP during 2001-2018 due to the fact that there have been no significant land conversions in the region and linked with the land use function (agriculture, agro-industries, high transportation flow, and high population density). The IGP is predominantly occupied under agricultural practices over the decades and is under the impact of global warming. The warming impact was prominent in the IGP, NE, and EI regions, where there was a decreasing or no change in NDVI (Figure 2a). Furthermore, the rise in sensible heat flux and evapotranspiration together with a decline in latent heat flux and precipitation (highly significant) corroborates warming impacts ( Figure  4a). Latent heat flux increases and sensible heat flux decreases (Figure 4) with increasing vegetation (Figure 3) leading to cooling impacts (Zong-Ci et al. 2013;Shastri et al. 2015). A high rate of deforestation in the NE regions (Table S3) coupled with the decline in precipitation causes albedo warming over evapotranspiration cooling (Bonan 2008). The eastern coast (EC) and eastern CI observed cooling effects (highly significant LST) due to (Table  S3: Orissa) increased evapotranspiration and forest cover. Furthermore, the increased frequency of tropical cyclones and cyclonic precipitation during pre-monsoon and post-monsoon in the EC (Gogoi et al. 2019) induced a low negative trend in precipitation (not significant). The decrease in soil moisture (Figure 2d) over India during 2001-2018 was attributed to vegetation greening and linked with high evapotranspiration (Jasechko et al. 2013). The increased greening due to forest cover and croplands leads to comparatively higher loss of water through the transpiration process from the extended leaf area (Bernacchi and VanLoocke 2015) and thereby reduce the soil evaporation due to the shadow of larger foliage/leaf area index (Zhang et al. 2016). Thus, the extension of such conditions led to the probability of intensified hydrological drought in the future. The increased NDVI signifies healthy vegetation or agriculture crops that may have high water content. This high negative correlation between NDVI and LST is attributed to high leaf   water content that produces a cooling effect. The high to a moderate positive correlation between (a) evapotranspirationprecipitation and (b) NDVI and precipitation signifies accelerated evapotranspiration that induces the probability of convective precipitation. While the major parts of India experienced an opposite trend with rising evapotranspiration and declining precipitation primarily due to the greening impact (Figure 2b-c).
The proper and sustainable land use transformation is one of the major challenges and therefore included in the Sustainable Development Goal (SDG, Agenda 21). Agenda 21 implies land resource sustainability through coordinating the interrelationship between the six factors (environment, society, economy, population, culture, and politics). Land alteration needs to be controlled and managed, primarily the further conversion of forest cover to mitigate climate change impacts. Effective land management including preservation of the quantity and quality of forest resources, maintaining biodiversity, afforestation, urban green space, avenue plantation, agroforestry may effectively be implemented to meet climate goals considering the geographical perspective and local scenarios under the Paris Agreement and the Aichi Biodiversity Targets under the UN Convention on Biological Diversity (CBD), with multiple targets, explicitly referring to sustainable agriculture and forestry (OECD, 2020). Also, regular assessment of land use/land cover change needs to be assessed and considered for evaluation and improvisation of land-use policies.

Caveat
The caveat of the study is the use of reanalysis data to analyse a few biophysical variables, while a certain regional climate model and radiative forcing may provide magnified interrelationships between the climate extremes and LULCC.

Conclusion
The study exhibited cooling impacts in the major parts of India, primarily manifested by land use/land cover changes, moisture content, irrigation practices, soil type, latitudinal and altitudinal gradients. The study reported a decrease in grassland (15.1%), barren land (10.2%), and shrubland (9.5%) in contrast to an increase in forest cover in India during 2001-18. A cooling impact was observed in Western India, Central India, and the upper part of Southern India during 2001-2018, while warming in the Indo-Gangetic Plain and Northeast India. The analogous patterns of LST were evident in major parts of India barring Indo-Gangetic Plain, where warming effects were recorded during the pre-monsoon and monsoon seasons. The changes in LULC influence the thermodynamic pattern of the environment, and in response, alter the natural cooling and warming trends of land surfaces. The transformation of barren land into shrubland and later to cropland in Western India, and the transformation of cropland and grassland into forest cover in Central India led to cooling effects. In contrast, Indo-Gangetic Plains and Northeast India have observed warming effects due to a decline in greenness and wetness. The study necessitates the inclusion of bio-geophysical variables in land-climate feedback to develop integrated and effective climate mitigation and adaptation strategies.