Spatial and temporal variation of daytime and nighttime MODIS land surface temperature across Nepal

LUINTEL Nirajan, MA Weiqiang,c, MA Yaoming,c, WANG Binbin and SUBBA Sunil

aInstitute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China; bCollege of Earth and Planetary Science, University of Chinese Academy of Sciences,Beijing,China; cKey Laboratory of Tibetan Environmental Changes and Land Surface Processes,Beijing,China

ABSTRACT Land surface temperature (LST) is an important variable for assessing climate change and related environmental impacts observed in recent decades.Regular monitoring of LST using satellite sensors such as MODIS has the advantage of global coverage, including topographically complex regions such as Nepal.In order to assess the climatic and environmental changes,daytime and nighttime LST trend analysis from 2000 to 2017 using Terra-MODIS monthly daytime and nighttime LST datasets at seasonal and annual scales over the territory of Nepal was performed.The magnitude of the trend was quantified using ordinary linear regression, while the statistical significance of the trend was identified by the Modified Mann-Kendall test.Our findings suggest that the nighttime LST in Nepal increased more prominently compared to the daytime LST,with more pronounced warming in the pre-monsoon and monsoon seasons.The annual nighttime LST increased at a rate of 0.05 K yr-1(p<0.01), while the daytime LST change was statistically insignificant.Spatial heterogeneity of the LST and LST change was observed both during the day and the night.The daytime LST remained fairly unchanged in large parts of Nepal,while a nighttime LST rise was dominant all across Nepal in the pre-monsoon and monsoon seasons. Our results on LST trends and their spatial distribution can facilitate a better understanding of regional climate changes.

KEYWORDS Daytime; nighttime; land surface temperature; MODIS;Nepal; climate change

1. Introduction

Land surface temperature(LST)is an indicative factor for climatic and environmental changes,and has been extensively used in the assessment of environmental features such as urban heat islands(Peng et al.2018),vegetation conditions (Li et al. 2016), land-use and land-cover changes (Muro et al. 2018), drought severities (Karnieli et al.2010),and climate changes(Eleftheriou et al.2018;Khorchani et al.2018),among others.The recent advancements in remote sensing have made the acquisition of LST data at high spatial and temporal resolutions feasible.Satellite-borne thermal sensors can efficiently record LST signals and have shown promising prospects for global monitoring by leveraging the acquisition of temperature data even from topographically complex regions where maintenance of observation stations is challenging.Among them, the Moderate Resolution Imaging Spectroradiometer(MODIS)is the most widely used,due to its high spatial and temporal resolution (1 km and 4 times a day),global coverage,and long-term dataset.

MODIS LST data have been extensively used for spatial and temporal analysis globally. MODIS LSTbased global temperature trend analysis shows that the temperature change varies from region to region,as cooling has been observed in the central and eastern regions of the Pacific Ocean, northern regions of the Atlantic Ocean, northern regions of China, Mongolia,southern regions of Russia, western regions of Canada and America, eastern and northern regions of Australia,and southern tip of Africa, while warming has been observed in other regions, mostly in the Southern Hemisphere (Mao et al. 2017). Furthermore, differing patterns of daytime and nighttime LST changes have been observed in the Terra-MODIS dataset in Greece(Eleftheriou et al. 2018).

Nepal is a mountainous country situated on the southern flank of the central Himalaya and has been identified as one of the main hot spots with respect to global warming and climate change. Previous studies have recognized the signals of climate change in Nepal mainly using air temperature and precipitation data(Baidya, Shrestha, and Sheikh 2008; DHM 2017).Increased temperatures (1971-2014) at the rate of 0.056°C yr-1(DHM 2017) and the proliferation of irregular precipitation (Talchabhadel et al. 2018) are the major climate change signals observed in Nepal.Nevertheless,the majority of climate studies have relied on observation station data that are not uniformly distributed; that is, no stations above 4100 m above sea level (http://dhm.gov.np/meteorological-station/).Hence, to better understand the temperature trends in Nepal, data obtained from high mountain regions are required. The remote sensing of temperature presents a unique avenue for continuously inspecting the thermal environment across the high mountain region of Nepal. Thus, the evaluation of spatial and temporal variation of LST across Nepal not only supplements climate studies in Nepal but also adds value to regional and global climate studies related to the surface thermal environment and its variation.

The main objective of the current research was to explore the annual and seasonal spatiotemporal trends of the clear-sky daytime and nighttime LST across Nepal using 18 years of the MODIS monthly LST product.

2. Materials and methods

2.1. Study area

Nepal is located towards the southern slope of the central Himalaya between China and India, extending from 26.35°N to 30.45°N and 80.06°E to 88.20°E,with an area of 147 181 km2(Figure 1). The topography is dominated by high-elevation-rugged terrain in the north with a narrow band of low-elevation and flat land in the south,with the elevation ranging from 60 to 8848 m above sea level within the north-south span of approximately 200 km.Such diversity generates varied climatic zones stretching from a subtropical climate in the southern plains to a tundra climate in the high mountains in the north, thereby enabling the occurrence of a diverse eco-environment in a relatively small area.The temperature decreases from the south to the north, as the temperature is strongly influenced by the elevation.The seasons are divided into four based on the monsoon circulation pattern: pre-monsoon(March, April, May); monsoon (June, July, August,September); post-monsoon (October, November); and winter(December,January,February).

Figure 1. Elevation map of Nepal.

2.2. Data and preprocessing

2.2.1. MODIS

MODIS is a multispectral sensor on board sunsynchronous near-polar orbital satellites - namely,Terra and Aqua - that belong to NASA's Earth Observing System. MODIS images have been used to develop a number of data products for environmental monitoring, including LST products. Terra products have been available since March 2000 and Aqua products since July 2002. They have a revisit time of one to two days depending on the location and record two observations per day,which results in up to four observations per day between them.In the current study,a Terra product was used, due to the availability of a longer time series. The Terra satellite passes over Nepal at around 10:45 and 21:45 every day.

MODIS LST data are produced using the generalized split-window algorithm and the day/night algorithm,which have been continuously optimized for quality improvement (Wan 2014; Wan and Dozier 1996).Although a daily 1 km spatial resolution dataset is available, monthly products on a 0.05° geographical grid (MOD11C3) have been used in this study because the availability and reliability of LST data increases with spatial aggregation (Bosilovich 2006) and temporal aggregation (Li et al. 2018), resulting in fewer gaps.

We downloaded MOD11C3 (version 6) LST data from March 2000 to February 2018 from the USGS LPDAAC ftp server (https://lpdaac.usgs.gov/data_access/data_pool).In this study, the time period of the study is referred to as 2000-17 for simplicity. The quality flag in MOD11C3 was used to filter out the poor quality observations. An LST with an LST error greater than 3 K was nulled and excluded from further evaluation. Such missing data accounted for only 0.62% and 0.85% of daytime and nighttime pixels inside the study area,respectively.

2.3. Trend analysis

Trend analysis was performed with annual and seasonal means of daytime and nighttime LST.First,the temporal evolution of mean LST across the spatial domain of Nepal was computed using ordinary linear regression (OLR)between the year (the independent variable) and LST(the dependent variable), and the significance of the observed trend was evaluated using the non-parametric Modified Mann-Kendall test (Hamed and Ramachandra Rao 1998). Second, the spatial distribution of the LST trend was analyzed using OLR, and the Modified Mann-Kendall test at each pixel in a similar manner as above.In the present work,all the trend values that were significant at the 0.05 level were defined as a significant trend.

3. Results

3.1. Temporal evolution of LST

Figure 2 shows the temporal evolution of the annual and seasonal daytime and nighttime LSTs across Nepal from 2000 to 2017.In general,LST was found to be increasing throughout the year,but the trends were non-significant during the daytime;the significant trends were observed in the nighttime LST, particularly in the warmer seasons and at the annual scale. The monsoon nighttime LST increased at the fastest rate, with a trend of 0.08 K yr-1(p < 0.05), while the annual nighttime LST increase was the most significant (0.05 K yr-1, p < 0.01).

Tables 1 and 2 summarize the percentage of pixels showing increasing or decreasing daytime and nighttime LST trends,respectively,grouped according to the statistical significance of the trend.A positive trend was found in more pixels compared to a negative trend across the study area,irrespective of the statistical significance of the trend. However, at the 95% confidence level (p < 0.05),pixels with non-significant(p>0.05)trends outnumbered pixels with significant trends, especially in the daytime.Nevertheless, the area with a significant LST rise was dominant over the area with a significant LST decrease,except in the case of winter daytime LST. The nighttime LST rise was more consistent compared to the daytime LST rise in all seasons.Warming,in both the daytime and nighttime, was subdued in the colder seasons (postmonsoon and winter) in comparison to warmer seasons(pre-monsoon and monsoon). Interestingly, a significant negative trend exceeded a significant positive trend by a small margin in the winter daytime LST. A significant increase in the nighttime LST(p<0.05)was observed in a higher percentage of pixels in the warmer seasons(43.02%in the pre-monsoon and 31.15%in the monsoon season) and also at an annual scale. Therefore, the temporal evolution of LST was closely related to the percentage of the study area showing a similar trend.

3.2. Spatial pattern of LST and LST trends

The spatial distribution of LST was largely controlled by the elevation, with higher LST recorded in the lowelevation southern plains and lower LST recorded in the high mountains in the north, in the daytime as well as in the nighttime in each season (Figure S1).The southern belt features subtropical climate with hot summers and mild winters, while the northern belt features tundra climate with sub-freezing temperatures throughout the year. This gives rise to a large variation in LST across the spatial domain of Nepal,with daytime LST exceeding 305 K in the southern plains during the pre-monsoon and monsoon seasons,while nighttime LST dropping below 255 K in the winter season along the Central Himalayan range in the north.

Figure 2.Temporal evolution of LST in Nepal:(a)pre-monsoon season;(b)monsoon season;(c)post-monsoon season;(d)winter;(e)annual.

Table 1.Percentage of pixels over Nepal showing trends in daytime LST.p<0.05 denotes trends that are significant and p > 0.05 denotes those that are non-significant, at the 95% confidence level.

Table 2.Percentage of pixels over Nepal showing trends in nighttime LST. p<0.05 denotes trends that are significant and p > 0.05 denotes those that are non-significant, at the 95% confidence level.

Figure 3. Spatial distribution of (a-e) daytime and (f-j) nighttime LST trends in Nepal: (a, f) pre-monsoon season; (b, g) monsoon season; (c, h) post-monsoon season; (d, i) winter; (e, j) annual.

The spatial distribution of the daytime and nighttime LST trends at the seasonal and annual scales at the 95%confidence level are shown in Figure 3. The figure illustrates that significant warming was more evident during the nighttime as compared to daytime, and in the premonsoon and monsoon seasons in comparison to the post-monsoon and winter seasons. Significant trends were not observed in most parts of Nepal during the daytime, apart from mild heating in the low-elevation southern belt in the pre-monsoon and monsoon seasons.Meanwhile,a significant positive LST trend was the dominant pattern during the nighttime,especially in the warm pre-monsoon and monsoon seasons when the monsoon LST rise was the most intense. In the post-monsoon season, slight warming was observed in the eastern region both in the daytime as well as at nighttime.Notably, in the winter season, unlike other seasons, the central region appeared to be cooling slightly, particularly in the daytime.At an annual scale,the southern lowelevation belt showed a slight increase in daytime LST,while most regions-particularly the northern regionshowed no change. A rise in annual nighttime LST was observed in large part of Nepal. Thus, the LST trend showed interseasonal variability, with different patterns and intensities in the daytime and nighttime.

4. Discussion

4.1. Comparison of observed trends

The nighttime LST trends were more significant (lower p-value) and had larger magnitude as compared to daytime LST trends. The difference could be due to the difference in the degree of influence of climatic and environmental factors. The daytime LST was more variable in the temporal domain (Figure 2), while the nighttime LST was more variable in the spatial domain(Figure S1)(quantitative data are presented in Table S1).This indicates that the daytime LST change was influenced by the factors that are less variable in space and more variable in time, such as solar radiation,while the nighttime LST change was controlled by the factors that are more variable in space and less variable in time,such as longwave radiation exchange, land cover change, soil moisture change, etc.

A comparison of the results in the current study to those from other studies on LST trends in recent decades reveals that our findings are, in general, in agreement with theirs. The increasing trend of the nighttime LST in the present study is concurrent with the findings in the neighboring area of the Tibetan Plateau(Ouyang et al. 2018; Qin et al. 2009). Similarly, the result presented herein is consistent with LST warming observed in the South Asia region (Mao et al. 2017).

The LST rate of increase in Nepal, especially the nighttime LST rise (0.05 K yr-1, p < 0.05), was found to be much faster compared to the rate of regional and global warming.The rate of increase in global temperature from 2001 to 2017, calculated using the surface temperature monthly data from Climate Research Unit,was 0.027°C yr-1(p < 0.05). Similarly, the global mean LST trend, as well as that of the LST across the global land surface and the LST across Asia from 2001 to 2012,calculated using the results reported in Mao et al.(2017), were 0.004 K yr-1(p > 0.05), 0.009 K yr-1(p >0.05), and 0.003 K yr-1(p > 0.05), respectively. These trend values are smaller than that in Nepal, which further demonstrates that Nepal lies in one of the hot spots for global warming.

4.2. Possible driving factors of LST change

LST is the product of the interaction of climatic and environmental components with the land surface,apparently influenced by atmospheric and land processes. Changes in LST are attributable to one of these factors or their combined effects. Generally, the LST trend across Nepal is in agreement with the global warming phenomenon. Global warming is indisputably attributable to an enrichment of greenhouse gases in the atmosphere (IPCC 2013), which has increased quickly over the last two decades (https://www.esrl.noaa.gov/gmd/ccgg/trends/gl_full.html). The enhanced greenhouse effect promotes positive radiative forcing,which subsequently increases the radiative temperature of the earth surface, i.e.,the LST.This effect of radiative forcing is elevated during the high insolation period of the summer (Khorchani et al. 2018).

Apart from this global-scale driver,regional and local factors also influence LST changes. The southern lowelevation belt,where the major land cover type is cropland, showed consistent warming in the warmer seasons,especially at nighttime,which could be attributed to land cover changes in the region. The conversion of forests to agriculture and settlement areas, as well as the conversion of croplands to settlement areas,reduces the evaporative cooling, resulting in accelerated warming. Such land cover change has occurred in the low-elevation southern belt and Churia hills of Nepal in recent decades (Li, Deng, and Zhao 2017;Rimal et al. 2018) (Figure S2).

Furthermore, in a unique manner, the winter LST is decreasing in the central Nepal low-elevation zone. This anomalous pattern can be partly explained by the foginduced effect. The fog cuts offthe radiation intake and reduces surface heating during the day(Sathiyamoorthy,Arya,and Kishtawal 2016).Evidence shows that the number of foggy winter days induced by increased aerosol emissions has been increasing in the Indo-Gangetic plain(IGP)(Sathiyamoorthy,Arya,and Kishtawal 2016).The fog in the IGP extends into the southern part of Nepal but does not affect the northern regions because aerosols are not easily transported to higher elevation regions during stable atmospheric conditions,which prevail in winter,as such aerosol-induced fog in the winter prevails below the 2000 m elevation mark(Yan et al.2016).The effect of fog is much more pronounced in the southernmost lowelevation plain,where a maximum aerosol concentration is found in the winter season (Alvarado et al.2018).This results in subdued warming or even cooling, especially during the daytime, in the low-elevation zone, while warming in the high-elevation zone (above the 2000 m elevation level) remains unabated. This inconsistent pattern observed in low-elevation plains in Nepal is in close relation to the proliferation of the number of cold days around the IGP in recent decades(Sharma 2017).

The decline in precipitation observed in recent decades in the study area could possibly influence LST changes,especially in the monsoon season. The increase in net radiation intake under a drier climate intensifies the warming effect through reduced cloudiness and a decrease in the moisture available for evaporative cooling, which gives rise to higher surface temperatures. The precipitation across Nepal has been observed to decrease,prominently in the monsoon season (DHM 2017; Karki et al. 2017). Although the long-term decreasing trends are insignificant, the recent trends (after the 1990s) are more significant(Shrestha,Yao,and Adhikari 2019).

5. Conclusions

In the present study, the temporal and spatial variation in the daytime and nighttime LSTs across Nepal were analyzed using a MODIS LST dataset from 2000 to 2017.The results indicate that the fingerprints of global warming and climate change are observed in Nepal,especially through the nighttime LST.Moreover,the nonuniformity in the spatial distribution shows that the LST is highly variable in space and time. Our work on the LST trend across Nepal using satellite data supplements a better comprehension of climate and climate change status in Nepal in relation to environmental impacts.Furthermore,we conclude that Nepal is warming faster in comparison to global warming rates. Hence, our findings contribute to climate and climate change studies at the regional and global scales. Nonetheless, the results presented here are to be considered with caution, due to a few limitations; for example, lack of precision, since an LST with an error less than 3 K is considered valid, and the use of coarse resolution(0.05°)data,which might not be able to capture small-scale changes in the mountainous area. Finally, we recommend a comprehensive study for future endeavors,using high spatial resolution long-term datasets of LST, and land cover, albedo, and aerosol datasets for better insight into the impact of different environmental and climate factors on LST modification.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This study was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences [grant numbers XDA2006010103 and XDA19070301]and the National Natural Science Foundation of China [grant numbers 41830650,91737205, 91637313, and 41661144043]. The authors are grateful to NASA for the production of open access data.