G ES LAND SURfACE TEMPERATURE DYNAMICS IN DRY SEASON 2015-2016 ACCORDING TO LANDSAT 8 DATA IN THE SOUTH-EAST REGION Of VIETNAM

Located in Southeast Asia, Vietnam is one of the most severely affected countries by climate change and faces to series of challenges related to climate change, in which droughts are one of the most serious natural disasters. Land surface temperature (LST) is important factor in evaluating soil moisture and drought phenomenon. Remote sensing technique with many advantages, compared with traditional methods, can be used effectively for retrieving LST. This article presents study on the application of LANDSAT 8 multi – temporal data for monitoring LST changes in dry season 2015 – 2016 in Loc Ninh district, Binh Phuoc province in Southeast region of Vietnam. LST was derived using SplitWindow (SW) algorithm. The results showed that the LST at the end of 2015 – 2016 dry seasons (in February and March) is much higher than at the early of dry season. The area with LST higher than 309 K increases very fast in dry season 2015 – 2016, from less than 1% of the total study area in November and December to 19.59% in February and 30.74% in March. The results obtained in this study can be used to create the LST distribution map and to monitor drought phenomenon. kEY wORDS: remote sensing, LST, drought, thermal infrared, LANDSAT, Vietnam CITATION: Quang Khanh Nguyen, Le Hung Trinh, Khanh Hoai Dao, Nhu Duan Dang (2018) Land Surface Temperature Dynamics In Dry Season 2015-2016 According To Landsat 8 Data In The South-East Region Of Vietnam. Geography, Environment, Sustainability, Vol.12, No 1, p. 75-87 DOI-10.24057/2071-9388-2018-06


INTRODUCTION
Vietnam is likely to be one of the several countries most adversely affected by climate change.During the last 50 years, Vietnam's annual average surface temperature has increased by approximately 0.5 -0.7 0C (Vietnam assessment report on climate change).LST is important fac-tor in global change studies, in estimating radiation budgets in heat balance studies and as a control for climate models.LST can provide important information about the surface physical properties and climate which plays a role in many environmental processes (Mallick et al. 2008;Mira 2007).The estimation of LST pays important role in numerical modeling especially in phys-ical based hydrological models where water balance/budgeting of the catchment is an important component (Thakur and Gosavi 2018).
Many researchers have used remote sensing data to determine and monitor LST distribution.Retrieval of LST using thermal infrared bands of satellite images is the most effective way to derive energy balance and evapotranspiration (ET) on regional basis (Pariada et al. 2008).Since the last of 20 th century, satellite-derived surface temperature data have been utilized for regional climate analyses on different scale (Carnahan and Larson 1990).Beginning with Landsat 4, Landsat satellite series provides the data for retrieval of LST for longer period of time.Landsat 5 TM and Landsat 7 ETM+ data were used to estimate LST in urban area (Alipour et al. 2007;Mallick et al. 2008;Kurma et al. 2012; Balling and Brazel 1988;Grishchenko 2012;Marchokov and Trinh 2013;Tran et al. 2009;Trinh 2014).The results of these studies have demonstrated that in the big cities, urban heat island effect is becoming a problem due to increasing coverage of land with asphalt pavements.
The relationship between LST and vegetated areas has been documented in the many studies.Cueto et al. (2007) found correlation between surface temperature in Mexicali (Mexico) and land use by using remote sensing data.Hyung Moo Kim et al. (2005) proposed algorithm to estimate the statistical correlation between LST and vegetation index.A study by Cai et al. (2017) analyzed the relationship between LST and land cover changes in zhengzhou city (Huabei Plain) using multi-temporal satellite data.They examined the usefulness of Landsat 5 TM imagery for classifying land cover/land use and using thermal infrared band (band 6) to produce a thermal map of zhengzhou city.
LST is also one of the most important factors in studying drought phenomenon, as well as input parameters for climate models (Alshaikh 2015).Many studies have proven that a combination of surface temperature and normalized difference vegetation in-dex (NDVI) can provide information about surface soil moisture.A study by Lambin and Ehrlich (1996) reviewed extensively the drivers between normalized difference vegetation index (NDVI) and brightness temperature (BT), and described a general spatial pattern of relationships between NDVI and BT, related to land cover.They concluded that BT/NDVI slope could be used to classify land cover, and monitor land cover changes over time when associated to seasonality information, retrieved from NDVI annual variations alone (Julien and Sobrino 2009).Sandholt et al. (2002) proposed a drought index called Temperature Vegetation Dryness Index (TVDI), which is calculated from satellite derived vegetation index (NDVI) and surface temperature.This drought index is also used in many other studies to assess soil moisture and drought status (zverev and Trinh 2015;Chen et al. 2011;Shang et al. 2017).
A number of algorithms have been used to estimate the LST using remote sensing thermal infrared (TIR) data as it is capable to decipher the thermal characteristic of the land surface.These algorithms are namely mono-window (MW), split-window (SW), dual-angle (DA), single-channel (SC)… (Galve et al. 2008;Rongali et al. 2018).The studies carried out in different areas, such as the northern Negev Desert, Israel (Du et al. 2014;Rozenstein et al. 2014) and the Beas River basin, India (Rongali et al., 2018) show that the split-window algorithm can be adjusted for estimating LST from Landsat 8 data to get better accuracy.The objective of our paper is to evaluate the dynamics of LST in Loc Ninh district, Binh Phuoc province of Southeast region of Vietnam during 2015 -2016 dry season using Landsat 8 multi-temporal data.SW algorithm was used to calculate LST from Landsat 8 data in this case study.

STUDY AREA AND MATERIALS
Loc Ninh is a mountainous district in the northwestern border of Binh Phuoc province, with a borderline of over 100 kilometers.It is bordered by Kratie and Cong Pong Cham provinces (Cambodia).The area is bounded between latitude 11 0 39'6.09''N to 12 0 05'50.7''Nand longitude 106 0 24'39.5''E to 106 0 59'19.3''E(Fig. 1) (locninh.binhphuoc.gov.vn, 2018).The district covers an area of 853.95 km 2 and had a population of 115268 people (locninh.binhphuoc.gov.vn, 2018).Loc Ninh district has a high terrain in the north and low terrain in the south.It located in tropical monsoon region; the climate is divided into two seasons: the rainy season from May until October and the dry season from November to April, while March and April are the warmest and driest months.This is the largest pepper growing area of Binh Phuoc province and also is one of the most regions severely affected by drought in Southeast region of Vietnam.In this study, five multispectral cloud -free LANDSAT 8 OLI -TIRS images (path 124, row 52) with a spatial resolution of 30x30 meters were acquired from November 24, 2015, December 26, 2015, January 11, 2016, February 28, 2016and March 31, 2016 (Fig. 2).The LANDSAT 8 data was the standard terrain correction products (L1T), downloaded from United States Geological Survey (USGS -http://glovis.usgs.gov)website.The data used in this study was grouped into two categories (Table 1): the thermal infrared data (band 10) was used to calculate temperature, the red (band 4) and near infrared band (band 5) to calculate surface emissivity based on normalized difference vegetation index (NDVI).
The SW algorithm is based on the different atmospheric absorption behavior of two ra-01|2019

GES
(1) (3) determined by using method based on NDVI image, which proposed by Valor and Caselles (1996)  For calculating NDVI index, the digital number of red and near infrared band was converted to surface reflectance value.In this study, one very advance atmospheric approach (FLAASH) has been applied on the Landsat 8 multispectral image, and then, the NDVI is calculated according to Eq. 7.
The land surface emissivity images of bands 10 and 11 are used to calculate mean and difference emissivity: In last step, LST can be calculated by following equation ( 1).

RESULTS AND DISCUSSION
The reflectance values for red and near infrared channels of LANDSAT 8 data was used to calculate normalized difference vegetation index (NDVI).For determining surface emissivity by this methodology, values of soil and vegetation emissivity are needed.The emissivities of pure soil and pure vegetation cover were calculated from the MODIS UCSB emissivity library using method proposed by Tang (Tang et al. 2011).Soil and vegetation emissivities for Landsat 8 TIRS bands are listed in Table 5 (Yu et al. 2014).
Basing on the emissivities values of soil and vegetation, emissivity image was prepared using method of Valor and Caselles by using formula (5).
From brightness temperature and land surface emissivity images, the LST image was obtained by using Spatial Modeler of ERDAS Imagine 2014 program.Fig. 4 shows the spatial distribution of LST in Loc Ninh district, Binh Phuoc province (Southeast region of Vietnam) in dry season 2015 -2016.
The LST ranged from 296.85 to 316.04 K in November 24, 2015;292.52 to 311.17 K in December 26, 2015;298.85 to 313.92 K in January 11, 2016;298.42 to 321.27 K in February 28, 2016 and 300.20 4).These areas have sparse vegetation cover and uncultivable land.High LST are also recorded in agricultural land use and residential land in the center and southern part of the study area.Meanwhile the area with full vegetation coverage in center of study area has much lower LST.
LST data from 10 measurements points at February 28, 2016 (Fig. 5d) were used in the comparison with temperature calculated from Landsat 8 image, which acquired from same day.This in situ data were observed in the framework of the ministry-level project (Ministry of Natural Resources and Environment (Vietnam), No. 2015.08.10) and were collected from 10:00 am to 11:00 am local time in the day selected.Comparison between LST at the measurement points and the results calculated from the Landsat 8 satellite image (February 28, 2016) based on SW and MW algorithms is presented in Table 6.It can be seen that LST at the measurements points is lower than the temperature calculated from the Landsat 8 image.The biggest difference between in situ data and LST calculated from the Landsat 8 image is 1,75 (K) degree.In addition, it can be seen that the LST value determined by SW and single-channel (SC) algorithms are highly similar.However, overall the LST value determined by used SW algorithm tends to be smaller than using the SC algorithm (Table 6).Thus, the difference between the LST values determined by used SW algorithm and the in situ data is lower than using the SC algorithm.
The LST distribution map of the study area displays the different zone of temperatures.The density sliced image shows seven temperature zones that represents greater than 310, 309 -310, 307 -309, 305 -307, 303 -305, 301 -303

Table 1 . Overview of the satellite scenes applied in this study
6 -SW coefficients values.The values of SW coefficients are given in Table 2 (Sobrino et al. 2006; Skokovic et al. 2014).The flowchart of SW algorithm utilized in the present study for the estimation of LST is shown in Fig. 3.

Table 2 . SW coefficient values for TIRS band of Landsat 8 imagery
In first step, OLI and TIRS band data must be converted to TOA spectral radiance using the radiance rescaling factors provided in the metadata file (Landsat.usgs.gov,2018):where: L λ -TOA spectral radiance (Watts/( m 2 * srad * μm)), M L -Band-specific multiplicative rescaling factor from the metadata (RADIANCE_ MULT_BAND_x, where x is the band number) (Landsat.usgs.gov,2018),A L -Band-specific additive rescaling factor from the metadata (RADIANCE_ADD_ BAND_x, where x is the band number) (Landsat.usgs.gov,2018),Q cal -Quantized and calibrated standard product pixel values (DN).Fig.3.

Table 6 . Compare the LST at the measurement points and the results calculated from the Landsat 8 satellite image increase
, corresponding to 20,46% and 31,20% of the total study area on February 28 and March 31, 2016.Thus, it can be noticed, LST of Loc Ninh district tend to increase significantly in dry season 2015 -2016, in that March is considered to be the hottest month.This is also consistent with monitoring data at meteorological stations and in situ data of ministry-level project (Ministry of Natural Resources and Environment (Vietnam), No. 2015.08.10).