Flip flop of Day-night and Summer-Winter Surface Urban Heat Island Intensity in India

The difference in land surface temperature (LST) between an urban region and its nearby non–urban region, known as surface urban heat island intensity (SUHII), is usually positive as reported in earlier studies. India has experienced unprecedented urbanization over recent decades with an urban population of 380 million. Here, we present the first study of the diurnal and seasonal characteristics of SUHII in India. We found negative SUHII over a majority of urban areas during daytime in pre-monsoon summer (MAM), contrary to the expected impacts of urbanization. This unexpected pattern is associated with low vegetation in non-urban regions during dry pre-monsoon summers, leading to reduced evapotranspiration (ET). During pre-monsoon summer nights, a positive SUHII occurs when urban impacts are prominent. Winter daytime SUHII becomes positive in Indo-Gangetic plain. We attribute such diurnal and seasonal behaviour of SUHII to the same of the differences in ET between urban and non-urban regions. Higher LST in non-urban regions during pre-monsoon summer days results in intensified heatwaves compared to heatwaves in cities, in contrast to presumptions made in the literature. These observations highlight the need for re-evaluation of SUHII in India for climate adaptation, heat stress mitigation, and analysis of urban micro-climates.

for the estimation of the UHI characteristics. The vegetation indices are obtained from  Aqua product (MYD13A, 1000 m) at temporal resolution of 16 days. The QC information 49 available with the data product is utilised to filter out good quality data to be utilised for the 50 analysis. 51 We further obtain the evapotranspiration (ET) data from MODIS-Aqua product (MYD13A,52 1000 m) at a temporal resolution of 16 days. The available information about the quality 53 control is used to filter out good quality data to be utilised for this analysis. 54 All three MODIS data products for the present study are obtained from climate data archive  61 A cell resizing operation S12 popularly known as resampling is carried out over the acquired 62 LC data. The resampling methodology operates a pixel scale calculation to change spatial 63 resolution of the dataset without losing useful information available with the data. The 500 m 64 spatial resolution LC images are interpolated to 1000 m resolution dataset to match with the 65 available LST data. There are many resampling methods available, through a variety of 66 platforms; here we apply the widely used Nearest Neighbor (NN) S13 resampling technique. 67 The NN algorithm maintains the original brightness values of MODIS pixels as well as has a 68 lower processing time of spatial resampling S14 . 69 The city clustering algorithm 35 is used to determine the extent of urban area for all the 84 70 cities. The following steps are followed to estimate UHI intensities over the selected cities: 71 1. An urban map is prepared based on the geographical location of centre of each big city.

72
The urban map is used to identify first urban node (root node here after) from the resampled 73 MODIS LC data. The root nodes are queued and processed one by one to identify their 74 respective urban and non-urban region. A root node is selected from the queue. 75 2. We examine eight neighbouring nodes around the root node. If the land cover type of the 76 neighbouring nodes has urban land cover, the neighbouring nodes are added into a queue, 77 assigning attribute of the neighbouring nodes as an urban node, otherwise the attribute of the 78 neighbouring node is assigned as a non-urban node. 79 3.
Step 2 is repeated for each node in the queue till the queue is empty. When the queue is 80 empty; we quit the search and return the urban map to grow the cluster of the successive root 81 node. 82 4. After the urban map is returned, suburban area is defined as the buffer zone, which is a 83 ring around urban area that consists of the nonurban nodes excluding water. 5. The urban centres with the count of buffer zone grids covering the 50-150% of urban 85 nodes are selected for the SUHII analysis.

86
Supplementary Figure S1 shows an example of identified urban cluster by CCA and selection 87 of corresponding non-urban region along the urban boundary. The figure S1 presents four including emissions from industry, transport, residential, agriculture and informal industry 133 sector including fuel consumption, process and fugitive emissions and solvent use S15-S16 .
These five sectors were disaggregated further into 13 source categories and ~75 technologies/activities for estimating 2010 emissions. Black carbon emissions arise largely 136 from traditional biomass technologies, characterized by inefficient combustion and 137 significant emissions, are widely used in residential cooking and "informal industries" 138 including brick production, food and agricultural product processing operations like drying 139 and cooking operations related to sugarcane juice, milk, food-grain, jute, silk, tea and coffee.

140
In addition, seasonal agricultural residue burning in field is a discontinuous source of crop .production, and climatological MODIS active fires S16 . From the dataset, emission 154 densities of BC in northern India ranging 60-120 Tongrid-1mon-1, in winter months, are 155 larger than those of 20-100 Tongrid-1mon-1, in pre-monsoon summer months. 157 We found that the opposite seasonal patterns of SUHII between pre-monsoon summer and 158 winter exist in north and central India (Figure 1) with positive winter daytime SUHII, which is in contrast to negative pre-monsoon summer daytime SUHII. Daytime land-surface temperature could also be influenced by the atmospheric abundance of radiation absorbing 161 constituents, such as black carbon (BC) aerosols, which are pollution particles emitted from 162 incomplete combustion of fuels that strongly absorb radiation over the entire solar 163 spectrum 30 . Radiation absorption from BC can lead to heating of the atmospheric layers in 164 places where these particles are abundant at instantaneous rates of up to several degrees 165 Kelvin per day S17 . However, it should be noted that BC results into increased air temperature, 166 but decreased LST and hence the direct impacts of BC do not explain the winter day-time 167 SUHII behavior. On the other hand, the lowered LST possibly results into decreased ET with 168 a modification in the latent heat flux that results into a feedback to LST. This is a complex 169 process and needs model driven studies to understand the same.

170
There is a strong haze from aerosol pollution over north and central India S18 , particularly in 171 winter months. Such aerosol, which includes a significant fraction of BC, can lead to surface 172 radiation balance changes that can affect land surface temperature and consequently, SUHII.

173
To examine the possible role of BC aerosols, the spatial plot of SUHII was overlaid on BC 174 emission fluxes, spatially distributed on a 25 km grid that was calculated in a recent 175 emissions inventory for India S15-S16 [Details in Supplementary    The UHI intensity is quantified with mobile surveys carried out from January 2011 to March 2013. The pre-dawn UHI is observed to be more intense than early night UHI, also intensity in winter is stronger than in summer. The study area is classified into different local climate zones (LCZ), thermal gradient and cooling rates are observed within the zones and validated with the LCZ classification. The maximum UHI intensity is seen in the central part of the city. 13 Nagpur S31 Traverse surveys were carried during the summer and winter seasons, for the years 2012-2014, to measure night time mean canopy UHI intensity. Canopy UHI effects were found to be most prevailing in high building and population density areas. The negative impact of vegetation and positive effect of population density is revealed. 14 Pune S32 Dry and wet bulb temperature data obtained by a mobile survey conducted in April 1997. The results indicate that at night, the core of the city appears as both heat and moisture islands whereas at the time of sunrise as heat and dry islands. Situated in a basin-like topography, the city experiences stronger influence of winds rather than the thermal circulation systems arising from spatial in homogeneity in thermal and moisture patterns. 15 Thiruvananthapuram S33 Air temperature variations across the urban centre were recorded by mobile traverse method on June 29, 2010. Cooling and warming rates in the urban centre and suburban area were derived from stationary air temperature recorders. The study observes significant difference in the urban and rural cooling rates with the UHI intensity reported as 2.4 o C. 16 Bopal S34 The study estimates UHI with the help of Landsat TM data of the year 2006. The study reports a prominent UHI effect over the city. The UHI intensity is observed to be higher over the roads and industrial zones and lower over the vegetated areas as well as the residential regions with lighter roofing. 17 Ahmedabad S35 In UHI effect is studied using Landsat ETM satellite data along with field measurements using Infra Red Gun in various zones of the city. The surface temperature near industrial areas and dense urban areas is reported to be higher as compared to other suburban areas in the city. 18 Mumbai S36 The UHI is estimated for the time period 1976-2007 of based on the meteorologically observed surface air temperature over the two urban stations of the urban region and two peripheral non urban stations. The study reveals prominent UHI during winter season for both day and night time than the summer season.

Noida S37
The UHI is assessed with the help of field data, meteorological observations and Landsat thermal dataset for year 2000 and 2013. Estimated UHI showed a negative correlation between NDVI, Emissivity and temperature whereas, NDBI, Albedo and temperature showed a positive correlation. The change in temperature is reported mainly due to increase in impervious areas.