An Assessment of the Accuracy of MODIS Land Surface Temperature Over Egypt Using Ground-Based Measurements

remote sensing

Article An Assessment of the Accuracy of MODIS Land Surface Temperature over Egypt Using Ground-Based Measurements

Ahmed M. El Kenawy 1,2,* , Mohamed E. Hereher 1,3 and Sayed M. Robaa 4 1 Department of Geography, Sultan Qaboos University, Al Khoud, Muscat 123, Oman 2 Department of Geography, Mansoura University, Mansoura 35516, Egypt 3 Department of Environmental Sciences, Damietta University, New Damietta 34511, Egypt; [email protected] or [email protected] 4 Department of Astronomy, Space Science and Meteorology, Faculty of Science, Cairo University, Cairo 12613, Egypt; [email protected] * Correspondence: [email protected]

Received: 30 August 2019; Accepted: 8 October 2019; Published: 12 October 2019

Abstract: Space-based data have provided important advances in understanding climate systems and processes in arid and semi-arid regions, which are hot-spot regions in terms of climate change and variability. This study assessed the performance of land surface temperatures (LSTs), retrieved from the Moderate-Resolution Imaging Spectroradiometer (MODIS) Aqua platform, over Egypt. Eight-day composites of daytime and nighttime LST data were aggregated and validated against near-surface seasonal and annual observational maximum and minimum air temperatures using data from 34 meteorological stations spanning the period from July 2002 to June 2015. A variety of accuracy metrics were employed to evaluate the performance of LST, including the bias, normalized root-mean-square error (nRMSE), Yule–Kendall (YK) skewness measure, and Spearman’s rho coefficient. The ability of LST to reproduce the seasonal cycle, anomalies, temporal variability, and the distribution of warm and cold tails of observational temperatures was also evaluated. Overall, the results indicate better performance of the nighttime LSTs compared to the daytime LSTs. Specifically, while nighttime LST tended to underestimate the minimum air temperature during winter, spring, and autumn on the order of 1.3, 1.2, and 1.4 C, respectively, daytime LST markedly overestimated the maximum − − − ◦ air temperature in all seasons, with values mostly above 5 ◦C. Importantly, the results indicate that the performance of LST over Egypt varies considerably as a function of season, lithology, and land use. LST performs better during transitional seasons (i.e., spring and autumn) compared to solstices (i.e., winter and summer). The varying interactions and feedbacks between the land surface and the atmosphere, especially the differences between sensible and latent heat fluxes, contribute largely to these seasonal variations. Spatially, LST performs better in areas with sandstone formations and quaternary sediments and, conversely, shows lower accuracy in regions with limestone, igneous, and metamorphic rocks. This behavior can be expected in hybrid arid and semi-arid regions like Egypt, where bare rocks contribute to the majority of the Egyptian territory, with a lack of vegetation cover. The low surface albedo of igneous and limestone rocks may explain the remarkable overestimation of daytime temperature in these regions, compared to the bright formations of higher surface albedo (i.e., sandy deserts and quaternary rocks). Overall, recalling the limited coverage of meteorological stations in Egypt, this study demonstrates that LST obtained from the MODIS product can be trustworthily employed as a surrogate for or a supplementary source to near-surface measurements, particularly for minimum air temperature. On the other hand, some bias correction techniques should be applied to daytime LSTs. In general, the fine space-based climatic information provided by MODIS LST can be used for a detailed spatial assessment of climate variability in Egypt, with important applications in several disciplines such as water resource management, hydrological modeling, agricultural management and planning, urban climate, biodiversity, and energy consumption, amongst others.

Remote Sens. 2019, 11, 2369; doi:10.3390/rs11202369 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, 2369 2 of 29

Also, this study can contribute to a better understanding of the applications of remote sensing technology in assessing climatic feedbacks and interactions in arid and semi-arid regions, opening new avenues for developing innovative algorithms and applications speciﬁcally addressing issues related to these regions.

Keywords: air temperature; MODIS; LST; validation; lithology; Egypt

1. Introduction In the era of climate change, an accurate characterization of climate variability and its impacts on natural and human environments requires climatic data at high resolution over space and time [1,2]. Amongst all climatic records, air temperature is an important input variable for both water and energy cycles, being a key indicator of the land surface–atmosphere interactions and feedbacks [3,4]. As such, it has a great importance from the view of various disciplines, including hydrology, agriculture, ecology, ecosystem, health, and energy, among others. Unfortunately, arid and semi-arid environments suffer from a lack of adequate meteorological networks that can properly reflect the main climatological conditions, particularly at the regional and local scales. Specifically, climatic data are often unevenly distributed over space, with temporal discontinuities and inhomogeneities. Furthermore, due to changes in the location of observatories, observers, observation practices, or instruments, climatic records often have a relatively short duration, with frequent gaps, highlighting the limitations of data temporal sampling and spatial coverage. Moreover, in areas of complex topography or high spatial variability of climate, meteorological observatories may fail to capture the high variability of the climate. In developing countries, stationary observation data could have a degree of uncertainty, due to human interference, as most of the meteorological stations are placed in airports under urbanization effects, which could induce records significantly biased from real ambient temperatures [5]. Overall, in data-sparse regions, the quality of climate records is mostly impacted by ageing infrastructure and the inherent costs of manipulating and maintaining observation networks, combined sometimes with a history of unrest, ethnic conflicts, and political and social instability (e.g., Syria, Iraq, Libya, Yemen, and Sudan) [6–8]. In Egypt, the available meteorological records are generally sparse over space and time. Specifically, the majority of the meteorological stations are situated in urbanized settlements close to the Nile and its delta and along the Mediterranean Sea and Red Sea coasts. These areas comprise less than 10% of the total area of the country. From these aspects, the diverse impacts of climate change and the complex interactions between humans and the environment make the current meteorological network inadequate to properly diagnose climate change and its diverse impacts, particularly at detailed spatial scales (i.e., regional and local scales). Within this context, the current meteorological network is not feasible enough to provide solutions that address appropriately the heterogeneity and dynamics of ecosystems across Egypt. In Egypt, high-density climatic information is desired to assess the possible impacts of recent climate change and variability, which are likely to be accelerated under future greenhouse gas emissions, on a wide spectrum of disciplines (e.g., urban climate, agriculture, water resources, food production, health, biodiversity, energy, etc.) [9–12]. Recalling these data limitations, most of the available studies on climate change and variability in Egypt have employed meteorological information from a limited number of meteorological stations and using coarse temporal resolution (i.e., monthly, seasonal, or annual) (e.g., [1,9,13–17]). For instance, based on long-term (1905–2000) records from 18 observatories, the author of [13] assessed intra-decadal variability of winter minimum temperature over Egypt. Also, the authors of [14] assessed temperature variability over Egypt, deploying a monthly data set of only nine time series covering the period 1971–2000. In contrast, a very limited number of investigations have handled climatic data at fine resolution (i.e., daily or sub-daily), which is mandatory for different hydrometeorological applications (e.g., hydrological Remote Sens. 2019, 11, 2369 3 of 29 modeling, natural hazards assessment and forecasting, etc.). One example is [18], the authors of which recently assessed changes in daily temperature extremes over Egypt during the past four decades (1983–2015). Overall, an inspection of these studies highlights the current limitations of meteorological records in Egypt either spatially or temporarily. These challenges strongly constrain any attempts for a reliable diagnosis of climate change and variability and their possible impacts in the country. With the current advancement in Earth observation in general and thermal remote sensing in particular, data retrieved from many sensors have been made freely available for research community, including, for example, the Advanced Very-High-Resolution Radiometer (AVHRR), Moderate-Resolution Imaging Spectroradiometer (MODIS), Spinning Enhanced Visible and Infra-Red Imager (SEVIRI), Meteosat Visible and Infrared Imager (MVIRI), Landsat 4 and 5 TM, Landsat 7 ETM+, Landsat 8 TIRS, Sentinel 2 and 3, and VIIRS. The space-based meteorological data are generally quasi-global, with high spatial and temporal resolution, opening interesting possibilities to overcome the current limitations of ground meteorological measurements [18–20]. Amongst the different platforms, the MODIS was launched in 1999 on board the Terra satellite and in 2002 on board the Aqua, representing a new avenue to sense the Earth’s terrain, ocean, and the atmosphere. The MODIS sensor has either a descending (Terra) or an ascending node (Aqua) [21]. Both platforms have sun-synchronous near-polar orbits and provide data in 36 spectral bands. These channels range from the blue (0.4 µm) to the thermal portion (14.4 µm) of the electromagnetic spectrum. MODIS data are available at three spatial resolutions: 250 m (channels 1,2), 500 m (channels 3–7), and 1000 m (channels 8–36), provided as a sinusoidal projection [22]. Given their importance in understanding heat fluxes between the surface and the atmosphere, a wide range of studies have employed thermal bands acquired from different space-based products to retrieve land surface temperatures (LSTs) and assess their climatology, anomalies, and trends (e.g., [20,23–25]), or to link climate variability with different hydroclimatic and environmental variables, such as vegetation greening (e.g., [26–28]), drought (e.g., [29–31]), forest fire (e.g., [32,33]), afforestation (e.g., [34]), urbanization (e.g., [35,36]), land use (e.g., [37,38]), agrometeorology (e.g., [39,40]), biodiversity (e.g., [41]), and human health (e.g., [42]). For example, in [25] the authors assessed the spatial and temporal variability of LST over peninsular Spain for the period 1981–2015 using NOAA-AVHRR data and linked LST trends with observed changes in air temperature, solar radiation, and land cover patterns. At the global scale, the authors of [22] assessed the association between annual maximum LST derived from the MODIS product and maximum air temperature observed at more than 10,000 weather stations worldwide. Albeit with their high spatial and temporal resolution, the integration of satellite data has to be validated before their eventual use in the different hydrometeorological applications. This is simply because satellite-based information often has a degree of uncertainty which can originate from different sources that may introduce abnormal radiance measurements from diffuse reflectance. In particular, an accurate estimation of surface emissivity and radiometric temperature from global satellites can be a challenging task due to various atmospheric effects, variable emissivity, thermal characteristics of the ground, and different viewing angles of the sensor [22]. In the Middle East and North Africa (MENA) region, further region-specific challenges can pose further complications to data derived from sensors. In particular, the thin and low cloud which is common in the sub-tropic regions can have similar reflectance ranges to cloud-free pixels. Furthermore, this region is extensively subjected to heavy aerosol loadings, mainly due to frequent sandy and dusty storms, besides the high albedo of bright sandy deserts [43]. These challenges stress the need to assess the quality of remote-sensing-based meteorological measurements, including LST, against surface-based observations. This assessment is important to identify and attribute the agreements/discrepancies between satellite and ground meteorological information, providing feedback to the climate modeling community and other users on the possible sources of these differences. Recently, a wide range of studies have focused on assessing the quality of satellite estimates of different climatic variables, including precipitation (e.g., [18,44,45]), cloudiness (e.g., [46,47]), surface temperature (e.g., [22,25]), soil moisture (e.g., [48,49]), and snow cover (e.g., [50,51]). Remote Sens. 2019, 11, 2369 4 of 29

With this background, the overriding aim of this study was to assess the ability of LST derived from the MODIS Aqua product to reproduce the main characteristics (e.g., climatology, seasonal cycle, anomalies, and statistical distribution) of surface air temperature in Egypt. This work mainly validated daytime and nighttime LST against observational climatic records registered in 34 meteorological observatories spanning the period 2002–2015. This work provides the first comprehensive assessment of the efficiency of remotely sensed meteorological information (LST-based) in Egypt, allowing us to determine whether LST can be employed—with confidence—as a supplementary source to ground measurements in this data-scarce region.

2. Study Area

Egypt is located in northeastern Africa between latitudes of 22◦N and 31◦36’N and longitudes of 25 E and 37 E. With a total area of almost 1 106 km2, Egypt can be divided into four major ◦ ◦ × physiographic entities: the Nile Valley and its delta, the eastern plateau, the western plateau, and Sinai Peninsula (Figure1). Topographically, Egypt is generally gentle, with an average altitude of almost 300 m. The altitude varies between 133 m in the Qattara Depression in the western plateau and − 2600 m in Mount Catherine in southern Sinai (Figure1). Lithologically, the eastern plateau is composed of igneous and metamorphic rock complexes, while sedimentary rock units dominate in the western plateau—mainly sandstone in the south and limestone in the central and northern portions [52]. Sinai has basement complexes of igneous and metamorphic rocks in the central and southern parts, but limestone and sand dunes in the north. The more recent quaternary sediments are mainly located in the delta [53]. Climatologically, Egypt is characterized by mild and rainy winters and hot and dry summers. Apart from a few mountainous sites, the climate is generally classified as arid hot desert (BWh) following Köppen climatic classification [54]. The annual average minimum and maximum temperatures are 15.8 ◦C and 24.6 ◦C, respectively. The summer maximum temperature commonly exceeds 30 ◦C, with frequent heat waves that may exceed 45 ◦C[18]. The winter minimum temperatures infrequently fall below 10 ◦C. Apart from agricultural land in the delta and close to Nile Valley, Egypt is characterized by a lack of vegetation activity with normalized difference vegetation index (NDVI) values, a proxy for green vegetation, generally close to zero [55,56]. This is particularly associated with the dominance of the arid climate, with high temperatures, low precipitation, and high evaporation rates.

3. Materials and Methods

3.1. Observational Data This work employed a complete register of daily maximum and minimum air temperatures from 34 meteorological observatories spanning the period from July 2002 to December 2015 (Table1). Air temperatures were measured at a standard height of 2 m above the ground following the WMO guidelines. Meteorological data were provided by the Egyptian Meteorological Authority (EMA) and subjected to a rigorous procedure to check data quality and homogeneity. A detailed description of this procedure is outlined in [18]. As illustrated in Figure1, the spatial distribution of the meteorological observatories is generally uneven, as most of the stations are located in the delta, along the Nile Valley, and close to the Mediterranean and the Red Sea. Notably, the distribution of the meteorological stations is sparse over non-populated regions (i.e., the eastern and western plateaus and Sinai). This can be seen in Figure2, where almost half of the stations are distributed between the latitudes of 29 ◦N and 31◦36’N. In the same context, roughly 48.5% of the stations are located within a distance of 100 km from the Nile, compared to 57.6% of the stations situated within 100 km of the coasts. Albeit with the limitations of this network, it still represents the densest available network of meteorological data over Egypt, covering the main climatic regions (i.e., the Mediterranean, arid, and semi-arid). Remote Sens. 2019, 11, x FOR PEER REVIEW 5 of 29 RemoteRemote Sens.Sens.2019 2019,,11 11,, 2369 x FOR PEER REVIEW 55 of 2929

Figure 1. Topography of Egypt and the spatial distribution of meteorological observatories employed Figure 1. Topography of Egypt and the spatial distribution of meteorological observatories employed forFigure validating 1. Topography daytime ofand Egypt nighttime and the land spatial surface distributi temperatureon of meteorological (LST). observatories employed for validating daytime and nighttime land surface temperature (LST). for validating daytime and nighttime land surface temperature (LST).

Figure 2. Cumulative distribution of the frequency of meteorological observatories in Egypt as a Figure 2. Cumulative distribution of the frequency of meteorological observatories in Egypt as a function of (a) latitude, (b) longitude, (c) elevation, (d) distance to sea, and (e) distance to the Nile. functionFigure 2. of Cumulative (a) latitude, distribution (b) longitude, of (thec) elevation, frequency ( dof) distancemeteorological to sea, andobservatories (e) distance in toEgypt the Nile. as a The colored area corresponds to the area represented by 50% of the stations. Thefunction colored of (areaa) latitude, corresponds (b) longitude, to the area (c )represented elevation, ( byd) distance50% of the to stations.sea, and (e) distance to the Nile. The colored area corresponds to the area represented by 50% of the stations.

Remote Sens. 2019, 11, 2369 6 of 29

Table 1. Geographical characteristics of the meteorological stations used in this study.

Station Latitude (◦N) Longitude (◦E) Altitude (Meters) Abu Swair 30.6 32.0 26 Alexandria 31.2 30.0 3 Assiut 27.2 31.1 50 Aswan 24.0 32.8 172 Baharia 28.3 28.9 146 Baltim 31.5 31.1 2 Cairo 30.1 31.4 16 Dakhla 25.5 29.0 101 Damanhour 31.1 30.5 1 Edfou 25.0 32.8 84 El Khatatba 30.3 30.9 14 Fayoum 29.3 30.9 16 Habatah 31.1 26.0 211 Helwan 29.7 31.2 49 Hurghada 27.2 33.7 14 Kharga 25.4 30.5 87 Kossier 26.1 34.3 25 Luxor 25.7 32.7 84 Mallawi 27.7 30.8 46 Matrouh 31.3 27.2 30 Minya 28.1 30.7 139 Nuwaibaa 29.0 34.7 222 Ras Benas 24.0 35.5 2 Ras El Nakab 29.6 34.8 633 Ras Sedr 29.6 32.7 16 Safagah 26.8 34.0 189 Salloum 31.5 25.2 1 Shelateen 23.1 35.6 10 Sidi Barrani 31.5 25.9 90 Siwa 29.2 25.5 -19 Sohag 26.6 31.7 62 Suez 29.9 32.5 27 Tahrir 30.7 30.7 12 Tanta 30.8 31.0 15

3.2. MODIS LST Data With the availability of several space-based products (e.g., NOAA-AVHRR, Landsat, Sentinel, SEVIRI, MVIRI, and VIIRS), our preference was to test the performance of the MODIS product as a surrogate for or supplementary source to ground observations in Egypt. This decision was motivated by the fact that MODIS images cover a large geographical extent, where Egypt is completely covered by only three adjacent tiles, namely, h20v05, h20v6, and h21v05. Moreover, the images are available at daily temporal resolution, with high grid intervals ( 1 km). Furthermore, the MODIS LSTs are calibrated ≈ for atmospheric interference in NASA labs. Due to their simplicity and robustness, LST products retrieved from MODIS Terra and Aqua have increasingly been used in different research applications (e.g., [25,28,32,41,57–62]). In this study, daytime and nighttime surface temperatures were extracted from the Aqua LST 8-day composite product (MYD11A2V.6) for the Egyptian territory for the period from July 2002 to June 2015. The data were downloaded from the Land Processes Distributed Active Archive Center of NASA (https://lpdaac.usgs.gov/). In comparison to MODIS Terra, the MODIS Aqua acquires daytime and nighttime data at times much closer to the highest (1:30 p.m.) and lowest (1:30 a.m.) near-surface daily air temperature. Further details about the current status, perspectives, and computational algorithms of satellite-derived LST are documented in [63]. LSTs from MODIS images were retrieved following the generalized split-window algorithm. This algorithm is mostly similar to the split-window method proposed in [64] for AVHRR data and the physics-based day/night algorithm developed in [65]. In general, the generalized split-window method corrects the atmospheric effects based on the differential absorption in adjacent infrared bands. In the MODIS data, this algorithm is commonly applied to retrieve LST from clear-sky pixels Remote Sens. 2019, 11, 2369 7 of 29 using classification-based emissivities in the split-window bands [66]. Further details about the generalized split-window algorithm and temperature/emissivity separation algorithms are outlined in [66]. We retrieved MODIS data for 8-day composites using the brightness temperature values extracted from two adjacent emissivity channels that are located in the atmospheric window, covering the thermal infrared spectrum (TIR): channel 31 (10.78–11.28 µm) and channel 32 (11.77–12.27 µm) [67]. As opposed to other channels of the MODIS, these two channels are virtually unimpacted by changes in the temperature and amount of water vapor in the mid and high troposphere [68]. Moreover, swaths with a valid Level 1B radiance in these channels are acquired under cloud-free conditions. Herein, it is noteworthy that although MODIS Aqua data are available at a daily scale for the product MYD11A1, we decided to deploy the 8-day composite product (MYD11A2). This was simply to avoid the contamination induced by heavy aerosols, which could largely influence the quality of the retrieved data, especially in very dense aerosol regions like the MENA. Furthermore, the reliability of LST retrievals increases with temporal aggregation due to the presence of fewer gaps in the data [69]. Overall, prior to processing LST data, 8-day composites of daytime/nighttime were acquired for the three tiles spanning the study domain. Then, the brightness value (in 16 bits) corresponding to the LST was converted from the Kelvin scale to the Celsius scale for all images and tiles. For validation purposes, the 8-day composites of daytime/nighttime LST were aggregated to monthly, seasonal, and annual scales using a simple arithmetic average. Herein, seasons were defined as winter (DJF), spring (MAM), summer (JJA), and autumn (SON). Likewise, the daily meteorological records of maximum and minimum air temperatures (described in Section 3.1) were aggregated to have a comparable temporal scale to the LST data. This aggregation procedure allows a direct comparison of the two products using the different accuracy metrics.

3.3. MODIS LST Performance There is a degree of uncertainty related to the use of LST, mostly corresponding to atmospheric, angular, and emissivity effects. Accordingly, a comparison of LST against ground-based measurements is needed to define the appropriateness of satellite-based measurements. In this work, the agreement between daytime/nighttime LST retrieved from MODIS Aqua and observational maximum/minimum air temperature (hereafter, Tmax/Tmin) gauged at 34 meteorological observatories was tested. This validation was made for the study period (2002–2015) on seasonal and annual scales using a range of accuracy metrics. Herein, it should be noted that our preference was to apply a point-to-grid validation approach, in which we directly compare LST gridded data with in situ data in the nearest grid box. This was simply to avoid any possible impacts of applying any regridding algorithms to the observational data, particularly with the low station density in some areas (especially in the eastern and western plateaus and Sinai). In this study, we employed four different statistical metrics for validation purposes: the bias, the normalized root-mean-square error (nRMSE), Spearman’s rho correlation coefficient, and the Yule–Kendall (YK) skewness measure. Combining the results from different accuracy metrics is advantageous in that it gives better indications of the agreement between LST and observational data, given that these statistics measure different characteristics of the data (e.g., mean, skewness, asymmetry, etc.). The accuracy metrics were formulated as

XN 1 Bias = N− P O , (1) | i − i| i=1 ! RMSE nRMSE = 100 (2) Xmax X × − min where, s Pn 2 i=1 XOi XPi RMSE = − , (3) n Remote Sens. 2019, 11, 2369 8 of 29

P 6 d2 rho = i , (4) N (N2 1) − (P95 P50) (P50 P5) (P95 P50) (P50 P5) YK = − − − − − − (5) P95 P5 LST − P95 P5 Observation − − where N is the sample size, O is the observed value, p is the corresponding LST value, i is the counter for grid boxes, di is the difference between the ranks of corresponding values of temperature (i.e., the observed and LST values for a particular grid box at a definite time step), and Pi represents the ith percentile of the temperature distribution. The bias indicates whether MODIS LST tends to underestimate or overestimate the observational surface air temperature. On the other hand, Spearman’s rho is a non-parametric statistic that assesses the sign and strength of association between the LST and observed temperature. Recalling that RMSE is sensitive to the presence of outliers in the series, which is a typical characteristic of surface air temperature in hybrid arid regions like Egypt, we used a non-dimensional form of the RMSE in which the RMSE is normalized to the range of the observed data. The YK measure assesses the agreement in asymmetry between the observed and LST data, suggesting a “perfect” symmetry of data distribution when values are close to 0 [70]. Differences between the LST and observed temperature were assessed as a function of the different geologies and landscape formations. The area morphology can be seen as a key factor impacting LST dynamics in Egypt, given the lack of vegetation greening on one hand and the presence of different morphological formations in the country on the other hand. We also assessed the accuracy of LST in reproducing the main spatiotemporal characteristics of in situ temperature data. This included the seasonal cycle, anomalies, temporal variability, and the cold and warm air temperature distributions. Importantly, the slope, which characterizes the amount of change (◦C/year) in the LST from 2002 to 2005, was computed and compared with the observational air temperature. The slope was calculated using the ordinary least squares regression method, with a higher slope suggesting greater LST changes and vice versa. We also compared the level of statistical significance of these changes using the modified Mann–Kendall statistic, as suggested [71]. In comparison to the classical non-parametric Mann–Kendall test, the modified version accounts for the presence of serial autocorrelation in the series, limiting its possible influence on trend significance. The accuracy of LST in reproducing the cold (low) and warm (high) tails of the air temperature distribution was also evaluated. In this research, the percentiles were assessed on a seasonal basis. The cold tail was presented using the 1st, 5th, 10th, and 25th time-varying percentiles of the air temperature distribution, while the warm tail was characterized using the 75th, 90th, 95th, and 99th time-varying percentiles of the temperature distribution. The aim was to assess the accuracy of LST in reproducing all parts of the temperature frequency distribution, not only the average conditions.

4. Results

4.1. Seasonal Cycle of Temperature Figure3 illustrates the seasonal cycles of nighttime and daytime LST compared to those derived from the observed Tmin and Tmax, respectively. The seasonal cycles were averaged for the 34 meteorological observatories spanning the period 2002–2015. As illustrated, nighttime LST exhibited good agreement with the seasonal cycle of Tmin. Overall, the LST tended to underestimate Tmin in the majority of the year. The differences between nighttime LST and Tmin did not exceed 1.5 C, with the highest differences occurring during late autumn and spring (e.g., October [ 1.5 C] ◦ − ◦ and November and January [ 1.4 C]). Exceptionally, the nighttime LST slightly overestimated Tmin − ◦ during the summertime, with values ranging between 0.15 ◦C (June) and 0.39 ◦C (July). On the other hand, the seasonal cycle of daytime LST was generally consistent with Tmax. Nonetheless, there was a remarkable tendency towards overestimating Tmax in all months, with the highest differences found in May (7.5 ◦C), June (7.3 ◦C), and July (6.6 ◦C), and to a lesser extent during winter months (e.g., January [3.7 ◦C] and December [4.4 ◦C]). Remote Sens.Remote2019 Sens., 11 2019, 2369, 11, x FOR PEER REVIEW 9 of 299 of 29

Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 29

Figure 3. Annual cycle of nighttime (minimum) and daytime (maximum) temperatures calculated Figure 3. Annual cycle of nighttime (minimum) and daytime (maximum) temperatures calculated from observationalfrom observational and and LST LST data. data. Figure 3. Annual cycle of nighttime (minimum) and daytime (maximum) temperatures calculated 4.2. Validationfrom Outputs observational and LST data. 4.2. Validation Outputs Figure4.2. FigureValidation4 summarizes 4 summarizes Outputs the ditheff differenerent accuracyt accuracy metrics metrics usedused to asse assessss the the agreement agreement between between Tmin Tmin and nighttimeand nighttimeFigure LST. 4 summarizesLST. The The bias bias suggests the suggests differen a tendency at accuracytendency ofmetrics of the the LSTLST used to to underestimate assess the agreement Tmin Tmin in between inall allseasons seasons Tmin and and annually.annually.and The nighttime other The other indicatorsLST. Theindicators bias (i.e., suggests (i.e nRMSE,., nRMSE, a tendency Spearman’s Spearman’s of the LST rho, rho, to and underestimate and YK) YK) agree agree onTmin on a a betterin better all seasons accuracy accuracy and of of LST in reproducingLSTannually. in reproducing TminThe other during indicatorsTmin spring during (i.e and., nRMSE,spring autumn and Spearman’s comparedautumn rho, compared to and in YK) winter to agree in and winter on summer.a better and accuracysummer. As illustrated, of As it seemsillustrated,LST that in thereproducing it performance seems that Tmin the ofduring performance LST spring at the annualandof LST autumn at scale the compared annual is more scale consistentto inis wintermore withconsistent and LSTsummer. performancewith As LST duringperformanceillustrated, spring and it during seems autumn, springthat ratherthe and performance autumn, than that rather of in LST winterthan at thethat and annual in winter summer. scale and is Forsummer.more example, consistent For example, thewith correlation LST the correlation between Tmin and nighttime LST, as represented by Spearman’s rho, was 0.95 at the betweenperformance Tmin and during nighttime spring LST, and autumn, as represented rather than by that Spearman’s in winter and rho, summer. was 0.95 For at example, the annual the scale, annualcorrelation scale, between compared Tmin to and0.88 nighttimeand 0.90 forLST, spring as re presentedand autumn, by Spearman’srespectively. rho, The was correlation 0.95 at the was compared to 0.88 and 0.90 for spring and autumn, respectively. The correlation was much lower for muchannual lower scale, for compared winter (rhoto 0.88 = 0.56)and 0.90and forsummer spring (rhoand autumn,= 0.58). Figurerespectively. 5 depicts The thecorrelation results wasof the wintervalidationmuch (rho = lower0.56) procedure for and winter summer for (rho Tmax. (rho= 0.56) As= 0.58).andopposed summer Figure to Tm(rho5 depictsin, = the0.58). bias the Figure resultsindicates 5 depicts of that the validationthethe resultsLST tended of procedure the to for Tmax.overestimatevalidation As opposed procedure Tmax in to allfor Tmin, seasons,Tmax. the As with biasopposed values indicates togenerally Tm thatin, the exceeding the bias LST indicates tended5 °C (especially that to the overestimate LST during tended summer Tmaxto in all seasons,andoverestimate spring). with valuesNotably, Tmax in generally theall seasons,lowest exceeding accuracy with values of 5 LS◦ generallyCT (especiallywas mainly exceeding duringevident 5 °C summerduring (especially summertime, and during spring). summer with Notably, an the lowestnRMSEand accuracyspring). value Notably, of of 1.34 LST °C, the was rho lowest mainlycoefficien accuracy evidentt of 0.17of LS during °C,T was and summertime,mainly YK coefficient evident withofduring −0.18 an summertime, °C. nRMSE value with of an 1.34 ◦C, rho coenRMSEfficient value of 0.17 of 1.34C, and°C, rho YK coefficien coefficientt of 0.17 of °C,0.18 andC. YK coefficient of −0.18 °C. ◦ − ◦

Figure 4. Boxplots showing the diﬀerent metrics (Bias, normalized root-mean-square error [nRMSE], Spearman’s rho and Yule–Kendall skewness measure) employed for validating nighttime LST against ground measurements. The central red line of each box shows the mean, while the central black line denotes the median. The bottom and top of boxes represent the 25th and 75th percentiles, respectively, while lower and upper whiskers indicate the 10th and 90th percentile, respectively. Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 29

Figure 4. Boxplots showing the different metrics (Bias, normalized root-mean-square error [nRMSE], Spearman’s rho and Yule–Kendall skewness measure) employed for validating nighttime LST against ground measurements. The central red line of each box shows the mean, while the central black line denotes the median. The bottom and top of boxes represent the 25th and 75th percentiles, respectively, Remote Sens. 2019, 11, 2369 10 of 29 while lower and upper whiskers indicate the 10th and 90th percentile, respectively.

FigureFigure 5. 5.Boxplots Boxplots showing showing the the di differentfferent metrics metrics employed employed for for validating validating daytime daytime LST LST against against ground ground measurements.measurements. The The central central red red line line of of each each box box shows shows the the mean, mean, while while the the central central black black line line denotes denotes the median. The bottom and top of boxes represent the 25th and 75th percentiles, respectively, while the median. The bottom and top of boxes represent the 25th and 75th percentiles, respectively, while lower and upper whiskers indicate the 10th and 90th percentile, respectively. lower and upper whiskers indicate the 10th and 90th percentile, respectively. Figure6 summarizes the association between Tmin and nighttime LST at seasonal and annual Figure 6 summarizes the association between Tmin and nighttime LST at seasonal and annual scales. The results suggest higher correlations during winter (r = 0.62) and spring (r = 0.58), while the scales. The results suggest higher correlations during winter (r = 0.62) and spring (r = 0.58), while the lowest correlation was found in summer (r = 0.40). As illustrated, the ground measurements of lowest correlation was found in summer (r = 0.40). As illustrated, the ground measurements of Tmin Tmin seem to be impacted by both underestimation and overestimation, which compensate each seem to be impacted by both underestimation and overestimation, which compensate each other and other and finally make the observed Tmin less biased with respect to the nighttime LST. Contrarily, finally make the observed Tmin less biased with respect to the nighttime LST. Contrarily, the daytime the daytime LST showed a remarkable overestimation of Tmax in all seasons and annually (Figure7). LST showed a remarkable overestimation of Tmax in all seasons and annually (Figure 7). The The correlations vary considerably from one season to another, with the best agreement found for winter correlations vary considerably from one season to another, with the best agreement found for winter (r (r = 0.73) and autumn (0.60). Conversely, the lowest agreement was found during the summertime = 0.73) and autumn (0.60). Conversely, the lowest agreement was found during the summertime (r = (r = 0.40). The results found for the average conditions, as revealed by the bias, were also confirmed 0.40). The results found for the average conditions, as revealed by the bias, were also confirmed for the for the warm and cold tails of the temperature distributions, albeit with a few differences (Figures S1 warm and cold tails of the temperature distributions, albeit with a few differences (Figures S1 and S2). and S2). As illustrated, the LST tended to underestimate both the lower and upper percentiles of Tmin, As illustrated, the LST tended to underestimate both the lower and upper percentiles of Tmin, particularly during winter and autumn. Exceptionally, LST overestimated the cold tail percentiles particularly during winter and autumn. Exceptionally, LST overestimated the cold tail percentiles of of Tmin during the summertime (Figure S1). For Tmax, the LST showed an overestimation of both Tmin during the summertime (Figure S1). For Tmax, the LST showed an overestimation of both the the lower and upper percentiles, albeit with a greater tendency to overestimate the lower percentiles, lower and upper percentiles, albeit with a greater tendency to overestimate the lower percentiles, especially during summer and autumn (Figure S2). especially during summer and autumn (Figure S2).

Remote Sens. 2019, 11, x FOR PEER REVIEW 11 of 29

Remote Sens. 2019,, 11,, 2369x FOR PEER REVIEW 1111 of of 29

Figure 6. Scatterplots summarizing the association between the observational minimum air Figure 6. Scatterplots summarizing the association between the observational minimum air temperature Figuretemperature 6. Scatterplots (Tmin) and summarizing nighttime LST the at a associatioseasonal scale.n between The 1:1 theline observationalis plotted. minimum air (Tmin) and nighttime LST at a seasonal scale. The 1:1 line is plotted. temperature (Tmin) and nighttime LST at a seasonal scale. The 1:1 line is plotted.

FigureFigure 7.7.Scatterplots Scatterplots summarizing summarizing the association the associatio betweenn between the observational the observational maximum airmaximum temperature air (Tmax) and daytime LST at a seasonal scale. The 1:1 line is plotted. Figuretemperature 7. Scatterplots (Tmax) and summarizing daytime LST atthe a seasonalassociatio scale.n between The 1:1 linethe isobservational plotted. maximum air temperature (Tmax) and daytime LST at a seasonal scale. The 1:1 line is plotted. InIn anan attemptattempt toto exploreexplore thethe linkslinks betweenbetween thethe performanceperformance ofof LSTLST andand lithologicallithological formationsformations in the study area, we plotted the seasonal and annual bias as a function of the main lithological in theIn study an attempt area, we to exploreplotted thethe seasonallinks between and annual the performance bias as a function of LST of and the lithological main lithological formations units units in Egypt. Figure8 illustrates the di ﬀerences (bias) between nighttime LST and Tmin for the inin theEgypt. study Figure area, 8we illustrates plotted the the seasonal differences and (bias) annual between bias as nighttimea function LSTof the and main Tmin lithological for the period units period 2002–2015. The results reveal better performance of the nighttime LST in regions dominated in2002–2015. Egypt. Figure The results8 illustrates reveal the better differences performance (bias) betweenof the nighttime nighttime LST LST inand regions Tmin fordominated the period by by sandstone and quaternary sediment formations (e.g., the delta, southern parts of the western 2002–2015.sandstone andThe quaternary results reveal sediment better formations performance (e.g., of the the delta, nighttime southern LST parts in regions of the western dominated plateau, by plateau, northern Sinai) compared to those with igneous, metamorphic, and limestone formations sandstonenorthern Sinai)and quaternary compared sediment to those formations with igneous, (e.g., themetamorphic, delta, southern and partslimestone of the westernformations plateau, (e.g., (e.g., majority of the eastern plateau, northern parts of the western plateau, and southern Sinai). northern Sinai) compared to those with igneous, metamorphic, and limestone formations (e.g.,

Remote Sens. 2019, 11, x FOR PEER REVIEW 12 of 29 Remote Sens. 2019, 11, 2369 12 of 29 majority of the eastern plateau, northern parts of the western plateau, and southern Sinai). Although these differences were evident in all seasons and annually, they were much more pronounced during Although these diﬀerences were evident in all seasons and annually, they were much more pronounced summer, suggesting higher variations in the response of the different lithological formations to during summer, suggesting higher variations in the response of the diﬀerent lithological formations to nighttime LST during this season. A quick inspection of Figure 8 also indicates that nighttime LST nighttime LST during this season. A quick inspection of Figure8 also indicates that nighttime LST tended to underestimate Tmin for igneous, metamorphic, and limestone formations. This is evident tended to underestimate Tmin for igneous, metamorphic, and limestone formations. This is evident for all seasons and annually. On the other hand, LST slightly overestimated Tmin in regions with for all seasons and annually. On the other hand, LST slightly overestimated Tmin in regions with sandstone and quaternary sediments. sandstone and quaternary sediments.

Figure 8. BoxplotsBoxplots summarizing summarizing the the bias bias (differences (diﬀerences between nightti nighttimeme LST LST and Tmin), for each season and annually, as a functionfunction of the main lithological units in Egypt. The central red line of each box shows the mean, while the central black line denotes the median. The bottom and top of boxes represent thethe 25th25th andand 75th 75th percentiles, percentiles, respectively, respectively, while while lower lower and and upper upper whiskers whiskers indicate indicate the 10th the 10thand 90thand percentile,90th percentile, respectively. respectively. In the legend,In the legend the numbers, the numbers in parentheses in parentheses indicate the indicate total number the total of numbergridded of points gridded belonging points to belonging each lithological to each unitlithological after regridding unit after the regridding bias values the deﬁned bias values at the defined station atlevel the using station the level SPLINE using algorithm. the SPLINE This algorithm. is an exact This interpolation is an exact method interpolation that considers method thethat actual considers data theat the actual gauged data sites. at the gauged sites.

Remote Sens. 2019, 11, 2369 13 of 29 Remote Sens. 2019, 11, x FOR PEER REVIEW 13 of 29

FigureFigure9 illustrates 9 illustrates the the di ffdifferenceserences calculated calculated betweenbetween daytime LST LST and and Tmax Tmax as asa function a function of the of the dominantdominant lithological lithological formation. formation. Notably, Notably, all all geological geological units showed showed anomalous anomalous differences differences between between daytimedaytime LST LST and and Tmax, Tmax, with with a cleara clear overestimation overestimation ofof Tmax (bias (bias is is mostly mostly ab aboveove 5 °C). 5 ◦C). Nonetheless, Nonetheless, therethere were were some some considerable considerable di ffdifferenceserences between between thethe didifferentfferent formations. formations. As As depicted, depicted, the theworst worst performanceperformance of daytime of daytime LST LST was was mainly mainly noted noted for for igneousigneous and metamorphic metamorphic units units in inall allseasons, seasons, while while betterbetter performance performance was was found found in in regions regions withwith sandstone formations formations,, especially especially during during summertime. summertime.

FigureFigure 9. Boxplots 9. Boxplots summarizing summarizing the biasthe bias (diﬀ (differenceserences between between daytime daytime LST LST and and Tmax), Tmax), for eachfor each season andseason annually, andas annually, a function as a offunction the main of the lithological main lithological units inunits Egypt. in Egypt. The The central central red red line line of of each each box showsbox the shows mean, the while mean, the while central the blackcentral line black denotes line denotes the median. the median. The bottom The bottom and top and of boxestop of representboxes represent the 25th and 75th percentiles, respectively, while lower and upper whiskers indicate the the 25th and 75th percentiles, respectively, while lower and upper whiskers indicate the 10th and 90th 10th and 90th percentile, respectively. In the legend, the numbers in parentheses indicate the total percentile, respectively. In the legend, the numbers in parentheses indicate the total number of gridded number of gridded points belonging to each lithological unit after regridding the bias values defined points belonging to each lithological unit after regridding the bias values deﬁned at the station level at the station level using the SPLINE algorithm. This is an exact interpolation method that considers using the SPLINE algorithm. This is an exact interpolation method that considers the actual data at the the actual data at the gauged sites. gauged sites.

Remote Sens. 2019, 11, 2369 14 of 29

Remote Sens. 2019, 11, x FOR PEER REVIEW 14 of 29 4.3. Climatology of Temperature Figure4.3. Climatology 10 illustrates of Temperature the climatology of the observed Tmin and nighttime LST at seasonal and annual scales.Figure As 10 depicted,illustrates the climatology observed Tmin of the showsobserved a clearTmin spatialand nighttime gradient LST from at seasonal south toand north, with theannual lowest scales. temperature As depicted, recordedthe observed in Tmin the delta, shows northwestern a clear spatial gradient region, from and insouth northern to north, Sinai. This picturewith the is lowest evident temperature for all seasons recorded and in the annually, delta, northwestern albeit with region, a stronger and in spatial northern gradient Sinai. This during springpicture and summer is evident compared for all seasons to during and annually, winter albeit and autumn.with a stronger This spatial is a typical gradient characteristic during spring of the and summer compared to during winter and autumn. This is a typical characteristic of the climate in climate in arid and semi-arid regions of the mid-latitudes, where maritime–continental contrasts are arid and semi-arid regions of the mid-latitudes, where maritime–continental contrasts are much much enhanced during solstices (i.e., summer and winter). As shown in Figure 11, daytime LST enhanced during solstices (i.e., summer and winter). As shown in Figure 11, daytime LST did not did notcapture capture well well the spatial the spatial patterns patterns of Tmax, of as Tmax, anomalous as anomalous high temperatures high temperatures (almost >35 °C) (almost were seen>35 ◦C) were seenin the in majority the majority of the of theEgyptian Egyptian territory territory in all in seasons all seasons apart apart from from winter. winter. These These anomalous anomalous temperaturestemperatures were were markedly markedly recorded recorded in in the the barren barren areasareas and and sandy sandy deserts deserts over over the theeastern eastern and and westernwestern plateaus. plateaus. In contrast, In contrast, lower lower temperatures temperatures were were mainly noted noted over over the the delta delta and and northwestern northwestern region,region, especially especially during during winter winter and and autumn, autumn, though though stillstill above the the observed observed Tmax. Tmax.

Figure 10. Spatial distribution of Tmin (left-hand panels) and nighttime LST (central panels) ( C). Figure 10. Spatial distribution of Tmin (left-hand panels) and nighttime LST (central panels) (°C). The ◦ The bias ( C) is plotted in the right-hand panels. The interpolated surfaces were created using the bias (°C)◦ is plotted in the right-hand panels. The interpolated surfaces were created using the SPLINE SPLINE algorithm, which is an exact interpolator scheme that accounts for the real values of the observational data.

Remote Sens. 2019, 11, x FOR PEER REVIEW 15 of 29

algorithm, which is an exact interpolator scheme that accounts for the real values of the observational Remote Sens. 2019, 11, 2369 15 of 29 data.

FigureFigure 11.11. SpatialSpatial distribution distribution of of Tmax Tmax (left-hand (left-hand pane panels)ls) and and daytime daytime LST LST (central (central panels) panels) (°C). ( ◦TheC). Thebias bias(°C) (is◦C) plotted is plotted in the in right- the right-handhand panels. panels. The interpolated The interpolated surfaces surfaces were created were createdusing the using SPLINE the SPLINEalgorithm, algorithm, which is whichan exact is interpolator an exact interpolator scheme that scheme accounts that for accounts the real values for the of real the valuesobservational of the observationaldata. The legends data. of TheTmax legends and daytime of Tmax LST and were daytime not homogenized LST were not to homogenized illustrate the spatial to illustrate gradients the spatialof temperature, gradients particularly of temperature, for daytime particularly LST. for daytime LST. 4.4. Temperature Anomalies 4.4. Temperature Anomalies Figure 12 depicts the anomalies in Tmin and nighttime LST for the whole of Egypt and for four Figure 12 depicts the anomalies in Tmin and nighttime LST for the whole of Egypt and for four selected stations representing diﬀerent lithological settings. The regional series for the whole of Egypt selected stations representing different lithological settings. The regional series for the whole of Egypt was created using the arithmetically averaged records from all stations during the period 2002–2015. was created using the arithmetically averaged records from all stations during the period 2002–2015. Herein, the anomalies were calculated for each month independently, with reference to the mean of Herein, the anomalies were calculated for each month independently, with reference to the mean of the base period 2002–2015. The probability distribution functions (pdfs) corresponding to the Tmin the base period 2002–2015. The probability distribution functions (pdfs) corresponding to the Tmin and nighttime LST are also plotted. We found good agreement between the anomalies of Tmin and and nighttime LST are also plotted. We found good agreement between the anomalies of Tmin and those of nighttime LST for the whole domain (r = 0.66), but with less spatial diﬀerence. In particular, those of nighttime LST for the whole domain (r = 0.66), but with less spatial difference. In particular, irrespective of the location of the meteorological observatory, its lithology, latitude, or dominant land cover, Pearson’s r varied generally between 0.49 (Tanta, the delta) and 0.58 (Kharga, sandy deserts). Remote Sens. 2019, 11, x FOR PEER REVIEW 16 of 29

Remoteirrespective Sens. 2019 of, 11the, 2369 location of the meteorological observatory, its lithology, latitude, or dominant16 land of 29 cover, Pearson’s r varied generally between 0.49 (Tanta, the delta) and 0.58 (Kharga, sandy deserts). For Tmax, there was also generally a good agreementagreement between observed and daytime LST anomalies forfor thethe wholewhole of of Egypt, Egypt, although although with with lower lower correlation correlation compared compared to Tmin to Tmin and nighttimeand nighttime LST (rLST= 0.48 (r =) (Figure0.48) (Figure 13). Also,13). Also, there there was was more more spatial spatial variation variation amongst amongst the the di differentﬀerent stations. stations. InIn particular, thethe lowestlowest correlationcorrelation (r(r == 0.39) was found for Kharga,Kharga, which represents sandysandy deserts.deserts. On the other hand, thethe bestbest correlation correlation was was noted noted for for stations stations located located in agriculturalin agricultural landscape landscape close close to the to Nilethe Nile (i.e., Assiut,(i.e., Assiut, r = 0.55) r = or0.55) in theor in delta the (i.e., delta Tanta, (i.e., rTanta,= 0.41). r = These 0.41). regional These regional diﬀerences differences were also were well also captured well bycaptured the pdfs. by the pdfs.

Figure 12.12. (Left)(Left) Probability Probability distribution distribution function function of of nighttime LST LST and Tmin and (right)(right) theirtheir correspondingcorresponding monthlymonthly anomaliesanomalies forfor selectedselected stationsstations andand thethe wholewhole country.country. AnomaliesAnomalies werewere calculatedcalculated forfor eacheach monthmonth independentlyindependently withwith respectrespect toto thethe basebase periodperiod 2002–2015.2002–2015. TheThe selectedselected stationsstations representrepresent didifferentﬀerent lithologicallithological settingssettings (Tanta:(Tanta: quaternary sediments in the delta, Kharga: sandstonesandstone formationsformations of of the the western western plateau, plateau, Assiut: Assiut: quaternary quaternary sediments sediments close to theclose Nile, to Hurghada:the Nile, basement rock of the eastern plateau). The regional series for the whole country were computed using Hurghada: basement rock of the eastern plateau). The regional series for the whole country were a simple arithmetic average. computed using a simple arithmetic average.

Remote Sens. 2019, 11, x FOR PEER REVIEW 17 of 29

4.5. Time Series Trend Analysis Figure 14 illustrates the slope (°C/year) and the statistical significance of changes in the daytime and nighttime LST from 2002 to 2015, as compared to those of Tmax and Tmin. For Tmin, the results suggest a general agreement with nighttime LST in terms of the sign of the change (i.e., positive/negative), as positive changes dominated in winter and spring, while negative changes were mostly observed in summer and annually. Discrepancy between the Tmin and nighttime LST was mainly noted during autumn, where observational data suggested positive changes in Tmin while LST indicated a mix of positive and negative changes, albeit with greater tendency towards a decline in the nighttime LST. Notably, the results suggest higher spatial variability of Tmin changes, as represented by a higher interquartile range of the slope values. On the other hand, the nighttime LST exhibited lower spatial variability. As illustrated, apart from springtime, changes in the Tmin and Remotenighttime Sens. 2019 LST, 11 were, 2369 statistically non-significant at the 90% level (p < 0.1). 17 of 29

Figure 13. (Left) Probability distribution function of daytime LST and Tmax and (right) their corresponding monthly anomalies for selected stations and the whole country. Anomalies were calculated for each month independently with respect to the base period 2002–2015. The selected stations represent diﬀerent lithological settings (Tanta: quaternary sediments in the delta, Kharga: sandstone formations of the western plateau, Assiut: quaternary sediments close to the Nile, Hurghada: basement rock of the eastern plateau). The regional series for the whole country were computed using a simple arithmetic average.

4.5. Time Series Trend Analysis

Figure 14 illustrates the slope (◦C/year) and the statistical signiﬁcance of changes in the daytime and nighttime LST from 2002 to 2015, as compared to those of Tmax and Tmin. For Tmin, the results suggest a general agreement with nighttime LST in terms of the sign of the change (i.e., positive/negative), as positive changes dominated in winter and spring, while negative changes were mostly observed in Remote Sens. 2019, 11, 2369 18 of 29 Remote Sens. 2019, 11, x FOR PEER REVIEW 18 of 29 summerFigure and 13. annually. (Left) Probability Discrepancy distribution between function the Tmin ofand daytime nighttime LST and LST Tmax was mainlyand (right) noted their during autumn,corresponding where observational monthly anomalies data suggested for selected positive stations changes and the in whole Tmin country. while LST Anomalies indicated were a mix of positivecalculated and for negative each month changes, independently albeit with with greater respect tendency to the base towards period a2002–2015. declinein The the selected nighttime LST. Notably,stations represent the results different suggest lithol higherogical settings spatial (Tanta: variability quaternary of Tmin sediments changes, in the as representeddelta, Kharga: by a highersandstone interquartile formations range ofof thethe slopewestern values. plateau, On theAssiut: other quaternary hand, the sediments nighttime close LST exhibitedto the Nile, lower Hurghada: basement rock of the eastern plateau). The regional series for the whole country were spatial variability. As illustrated, apart from springtime, changes in the Tmin and nighttime LST were computed using a simple arithmetic average. statistically non-signiﬁcant at the 90% level (p < 0.1).

FigureFigure 14.14. BoxplotsBoxplots showingshowing thethe slopeslope andand statisticalstatistical signiﬁcancesignificance ofof changeschanges inin TminTmin (upper)(upper) andand TmaxTmax (lower),(lower), comparedcompared to to thosethose inin nightnight andand daytimedaytime LST,LST, respectively.respectively.Changes Changesand andtheir their statisticalstatistical signiﬁcancesignificance were were assessed assessed for for the the period period 2002–2015. 2002–2015. The The central central black black line ofline each of boxeach shows box shows the mean, the whilemean, the while central the central gray line gray denotes line denotes the median. the median. The bottom The bottom and top and of top boxes of boxes represent represent the 25th the and25th 75thand percentiles,75th percentiles, respectively, respectively, while lowerwhile andlower upper and whiskers upper whiskers indicate theindicate 10th and the 90th10th percentile, and 90th respectively.percentile, respectively. Lower and upperLower dots and depictupper thedots 5th depict and 95ththe 5th percentiles, and 95th respectively.percentiles, respectively.

Similarly,Similarly, thethe resultsresults suggestsuggest aa generalgeneral agreementagreement betweenbetween TmaxTmax andand daytimedaytime LSTLST inin termsterms ofof thethe signsign (direction)(direction) ofof changechange (Figure(Figure 14 14).). AsAs depicted,depicted, changeschanges werewere mainlymainly positivepositive forfor TmaxTmax andand daytimedaytime LSTLST duringduring winterwinter andand spring.spring. OnOn thethe otherother hand,hand, summersummer witnessedwitnessed negativenegative andand positivepositive changeschanges forfor TmaxTmax andand daytimedaytime LST.LST. UnlikeUnlike inin otherother seasons,seasons, therethere werewere moremore didifferencesﬀerences betweenbetween TmaxTmax andand daytimedaytime LSTLST duringduring autumn.autumn. Speciﬁcally,Specifically, TmaxTmax exhibitedexhibited positivepositive changeschanges fromfrom 20022002 toto 2015,2015, whilewhile thethe daytimedaytime LSTLST showedshowed negativenegative changes.changes. Interestingly,Interestingly, thethe LSTLST waswas consistentconsistent withwith groundground measurementsmeasurements atat thethe annualannual scale,scale, suggestingsuggesting aa coolingcooling trend.trend. InIn particular,particular, bothboth nighttimenighttime LST and Tmin showed an amount of change at the annual scale on the order of 0.06 C/year. Similar LST and Tmin showed an amount of change at the annual scale on the order of− −0.06°◦ C/year. Similar to nighttime LST and Tmin, changes in the daytime LST and Tmax were generally non-significant at the 90% level. The only exception corresponds to Tmax in spring. A comparison between the seasonal

Remote Sens. 2019, 11, 2369 19 of 29 to nighttime LST and Tmin, changes in the daytime LST and Tmax were generally non-significant at the 90% level. The only exception corresponds to Tmax in spring. A comparison between the seasonal changes of nighttime and daytime LST suggests stronger changes in the daytime and nighttime LST during cold seasons (i.e., winter and spring) compared to during warm seasons (i.e., summer and autumn). While changes in the daytime and nighttime LST were generally positive during cold seasons, they mostly exhibited negative changes during warm seasons. Apart from springtime, changes were statistically non-significant at the 90% level for all seasons and annually. In terms of the amount of change, the results reveal that—apart from in autumn—there were no significant differences in the amount of change between the nighttime LST and Tmin. Contrarily, the Wilcoxon–Mann–Whitney non-parametric statistic suggests statistically significant differences in the slope between the daytime LST and Tmax (Table2). These di fferences were evident at the 99% significance level (p < 0.01) for all seasons and at the 95% level (p < 0.05) for the annual series.

Table 2. Results of the Wilcoxon–Mann–Whitney non-parametric test to compare the amount of change (slope) in temperature between the LST and ground measurements. Only statistically signiﬁcant diﬀerences between the LST and ground measurements at the 95% level (p < 0.05) are shown in bold.

Statistic Winter Spring Summer Fall Annual Mann–Whitney U 407.0 382.5 464.0 178.5 502.5 Wilcoxon W 968.0 943.5 1025.0 739.5 1063.5 LST/Tmin Z 1.76 2.08 1.03 4.69 0.54 − − − − − Signiﬁcance level 0.08 0.06 0.30 0.00 0.59 Mann–Whitney U 313.000 328.000 61.500 21.000 375.500 Wilcoxon W 874.000 889.000 622.500 582.000 936.500 LST/Tmax Z 2.969 2.778 6.195 6.714 2.169 − − − − − Signiﬁcance level 0.00 0.01 0.00 0.00 0.03

5. Discussion Air temperature is a key climatic variable from the view of various disciplines (e.g., hydrology, agriculture, ecology, ecosystem, health, and energy). As such, an accurate estimation of the spatial and temporal characteristics of air temperature is important for better understanding of land surface–atmosphere interactions and water and energy cycles [4]. Nonetheless, in some data-sparse regions like Egypt, the uneven spatial and temporal distribution of climatic information, combined with data quality limitations, makes it difficult to provide an accurate characterization of air temperature using ground-based observations only. With the advancement in Earth observation techniques, numerous studies have employed remotely sensed information and verified their potential as a proxy for weather ground measurements (e.g., [18,22,72]). Being one of the most important data records provided by NASA and other international data providers, the LST has recently been recognized as an invaluable remote sensing parameter with a wide spectrum of hydroclimatic, ecological, biophysical, and biogeochemical applications (e.g., [22,25,32,33,41,57]). This study provides—for the first time—a comprehensive assessment of the performance of LST retrievals from the MODIS in Egypt. The LST was retrieved using two thermal channels from the MODIS Aqua sensor using the generalized split-window algorithm applied to images of 8-day composites at 1 km grid resolution. The LST records were validated against ground measurements from 34 stations spanning the period between 2002 and 2015. This assessment is highly desired in Egypt, given the lack of meteorological records and their uneven distribution over space and time. Our results stress that the performance of LST is highly driven by seasonality. The different accuracy measures (e.g., bias, normalized RMSE, Spearman’s rho, and YK measure) agree that LST performed well during autumn and spring, while it performed worse during winter and summer. Notably, the daytime LST tended to markedly overestimate the maximum air temperature, with values exceeding 5 ◦C in most seasons. On the other hand, the nighttime LST slightly underestimated the Remote Sens. 2019, 11, 2369 20 of 29 minimum air temperature. This pattern can be understood in the context that daytime air temperature is mainly associated with the heating of the land surface due to remitted thermal radiation driven by solar insolation [73,74], while air acquires its energy during the nighttime solely from the Earth’s own emitted radiation. In summer, the extremely arid climate and the increase in net radiation intake combined together to intensify the warming effect at the surface, especially with the reduction in cloudiness and the decrease in the soil and atmospheric moisture necessary for evaporative cooling. In winter, Egypt has witnessed a statistically significant decrease of rainfall over the past few decades, which is mainly driven by changes in atmospheric circulation in response to greenhouse warming [17,75]. These configurations favor the warming of the land surface during daytime due to a lack of evaporative cooling and the enhancement of aridity. Again, the nighttime LST corresponds to an absence of solar insolation and, accordingly, a lack of solar radiation effect, making surface temperatures much lower than those of the adjacent air. A possible explanation for the seasonal variations in the performance of the LST over Egypt, as revealed in Figures6 and7, can directly be linked to the varying land–atmosphere interactions and feedbacks amongst seasons, especially within the atmospheric boundary layer. These interactions determine sensible heat fluxes and latent heat fluxes, which impact energy balance and exchange between the land surface and the atmosphere. The surface energy balance in general and heat fluxes in particular can largely impact thermal conditions at the surface level and in its surrounding air. In this context, the LST is an important driving force of turbulent heat fluxes and long-wave radiation exchanges [25]. Across the majority of Egypt, where vegetation cover is scarce, the role of the LST in formulating the energy budget is minimized during hyper-arid seasons (i.e., summer and spring), mainly due to higher air temperatures, the lack of vegetation, and the rapid decline in soil moisture. In regions of typical arid climates like Egypt, soil drying can suppress evapotranspiration and induce a reduction in relative humidity and a reinforcement of the atmospheric evaporative demand (AED). All these configurations combine together, contributing to a dramatic increase in LST, especially over bare soil regions like most of the eastern and western plateaus. These regions are characterized by a lack of vegetation and, thus, low vapor pressure deficit and low energy exchange. In their assessment of the coupling between summer air temperature and land surface energy in eastern Asia using measurements from 25 FLUXNET sites, Cho et al. [76] found negative feedback between the air temperature and Brown ratio (i.e., the ratio of sensible heat flux to latent heat flux). This feedback was much stronger over less-vegetated regions. In contrast, the role of the LST in reshaping the air temperature is maximized during less warm seasons (i.e., winter and, to a lesser extent, in autumn), particularly over arable and cultivated lands (e.g., the delta and the Nile Valley). This is mainly attributed to the increase of precipitation and, thus, soil moisture and vegetation greening, which act together to increase energy exchange between the land surface and the atmosphere. This situation induces a decrease in the differences between sensible heat flux and latent heat flux, making the LST much closer to the air temperature. Overall, the anomalous bias found between Tmax and the daytime LST in our domain stresses that some bias correction methods should be applied to the daytime LST. These correction methods can vary from the application of simple additive or multiplicative functions to more complicated techniques (e.g., quantile–quantile mappings). Spatially, our results demonstrate that the nighttime LST captures well the spatial gradient of the minimum air temperature. Contrarily, with the exception of in wintertime, anomalous daytime LSTs (mostly above 30 ◦C) predominate over the majority of the Egyptian territory. This can probably be understood in the context that minimum air temperature is mainly recorded during the nighttime, when temperature is more homogenous over space due to lower surface–atmosphere thermal contrasts. In contrast, maximum air temperature is observed during the daytime, when the spatial variability of the air temperature is much higher. Notably, the eastern plateau showed anomalously high daytime LST, with values exceeding 50 ◦C during summer months. This behavior can be understood in the context that the main lithological settings of this plateau are igneous and metamorphic rocks with dark colors [77]. These dark basement rocks favor a decrease in latent heat flux, mainly due to the low albedo of the surface, enhanced by the absence of vegetation and lack of soil moisture and, accordingly, Remote Sens. 2019, 11, 2369 21 of 29 less evaporative cooling [32]. Conversely, the daytime LST showed less bias in arable lands (e.g., the delta, close to the Nile) and in sandstone and limestone deserts in the western plateau. In sandy deserts, the Brown ratio is reversed compared to in the dark basement rocks of the eastern desert. This is mainly due to a remarkable increase in latent heat fluxes due to the high surface albedo from sandy deserts. In [78] and [79], the authors indicated that the “Charney mechanism” is more enhanced in semi-arid deserts, where the increase of surface albedo induces a significant decrease in surface absorbed solar energy and, accordingly, a considerable negative radiative forcing. On the other hand, although it is characterized by dry climate conditions, with average annual precipitation less than 200 mm, the delta is characterized by a long growing season (almost 11 months), with cropland and sparse shrub vegetation being the dominant land covers. In these arable lands with intensive vegetation cover, the high evapotranspiration rates and the lack of surface resistance largely impact physical land–atmosphere interactions, leading to a greater cooling effect [80,81]. Overall, the relatively good vegetation coverage in the delta can significantly modulate temperature increase close to the land surface and ultimately influence the LST [82,83]. Moreover, the delta and agricultural areas close to the Nile are characterized by an intensive irrigation and drainage network of channels, which has a local cooling impact on the LST. In [59], it was indicated that geology and surface albedo are key determinants of LST spatial variability over Egypt. Trend analysis of changes in MODIS LSTs and near-surface temperatures revealed that—apart from in autumn—changes were generally consistent in terms of the sign of changes. Specifically, changes were mainly positive during winter and spring, negative in summer, and undirected in autumn. This consistency was evident for both daytime and nighttime LSTs. Similarly, LSTs and ground-based temperatures agreed that stronger changes were more pronounced during cold seasons (i.e., winter and spring) compared to warm seasons (i.e., summer and autumn). The magnitude of these changes (slope) did not show statistically significant differences between nighttime LST and in situ measurements, as revealed by the non-parametric Wilcoxon–Mann–Whitney test, while it showed significant differences for daytime LST (Table2). The detected temporal variability of LSTs over Egypt concurred with a global assessment of MODIS LST-based temperature changes in [23], where it was indicated that LST changes from 2001 to 2012 were much stronger during winter and spring across the Northern Hemisphere. In a preliminary assessment of the variability of annual LSTs in Egypt, the LST was found to have increased by 0.3–1.06 ◦C/decade, with more warming in most urbanized sites [59]. In their assessment of LST changes over Greece, the authors of [24] found an overall cooling in the daytime annual LST and warming in the nighttime annual LST from 2000 to 2017. These findings do not agree well with what we observed for Egypt, as our results suggested a reversed pattern with a decline in the nighttime LST at the annual scale ( 0.06 C/year) and no significant changes in the − ◦ daytime LST. Interestingly, both LSTs and ground measurements agreed that summer temperatures in Egypt declined over the past two decades for both daytime and nighttime. This may be explained by the effect of the selected study period boundaries, where the first year of the study period (2002) was amongst the hottest years on record in Egypt, while the last year (2015) was a relatively cool year [18]. Another possible explanation for the LST performance is related to cloudiness anomalies. Amongst the different land–atmosphere coupling forces, cloudiness is a key factor that influences the Earth’s radiation budget, particularly during summertime. Figure 15 reveals an increase in summer cloud cover in Egypt during the past two decades (0.53%/decade). This increase could affect energy and heat transfer throughout insolation, suggesting below-normal temperatures, especially during the daytime. In the same context, we found a decline in the amount of solar radiation (MJ/m2) in summer from 2002 to 2015 ( 0.70 MJ/m2/decade) (Figure 15). The joint influence of the increased cloudiness and decreased − solar radiation may explain the negative changes in summer temperature, as defined by both LSTs and ground measurements. In contrast, cloudiness exhibited a statistically significant decrease during winter and spring (Table3), implying negative feedback with the surface and air temperatures. This is shown in Figure 16, as cloud cover was negatively correlated with LSTs and near-surface temperatures in winter and spring. This suggests that the increase in LSTs and ground temperature measurements Remote Sens. 2019, 11, 2369 22 of 29 during winter and spring was driven by a decrease in cloudiness during these seasons. In contrast, solar radiation exhibited positive changes in winter and spring, which induced temperature rise (Table3, Figure 16). The authors of [ 84] explained the role of cloudiness in controlling radiation fluxes and, thus, surface air temperature. Specifically, decline in the frequency of cloudy days is mostly linked to an increase in sunshine (i.e., lower incoming shortwave radiation compared to outgoing longwave radiation) and, in turn, more warming. In the literature, several works have indirectly assessed the performance of LST by linking it with some key climatic and environmental variables which could have a strong influence on air temperature (e.g., solar radiation) (e.g., [25,28,62]). However, these studies demonstrated that the association between the LST and these variables can vary considerably over space in response to the dominant local characteristics (e.g., land use). In recent decades, several studies have indicated that there is a tendency towards warmer sea surface temperatures (SSTs), especially in the tropical and subtropical oceans (the mid-Atlantic, the tropical Pacific, and the Indian Oceans) (e.g., [85–88]). These SST positive anomalies would induce a decrease in cloudiness in response to warmer ocean waters and accordingly above-normal temperatures over land. For example, the author of [88] found that the SST over the Red Sea has warmed by 0.29 ◦C/decade from 1982 to 2016. In [85], the authors indicated that a warming of 0.5 ◦C in SST over the Indian Ocean is likely to induce a decrease in cloudiness and, thus, more drying in the eastern Mediterranean. In the context of trend detection of LST and near-surface temperatures in Egypt, it is important to stress that the defined changes, for either LSTs or near-surface observations, should be seen in the context of the short interval of the study period (2002–2015). This span is not long enough to give a robust assessment of temporal changesRemote Sens. in 2019 a stable, 11, x FOR climate. PEER REVIEW Nonetheless, this trend also indicates a high interdecadal variability 23 of 29 of observed air temperature and LST in Egypt. Recently, a wide range of long-term station-based smooth topography. This homogeneity guarantees that the ground measurements can largely assessments of air temperature changes in Egypt revealed a strong warming of the maximum and represent the satellite spatial scale, making any attempt to directly validate remotely sensed data minimum air temperatures, and this warming was more strongly evidenced during the past three against station data less complicated. decades (e.g., [13,14,18,89]).

Figure 15. Temporal evolution of the standardized anomalies of cloud cover (left) and solar radiation Figure 15. Temporal evolution of the standardized anomalies of cloud cover (left) and solar radiation (right) in comparison to LSTs and ground-based temperatures during the summer season. In the (right) in comparison to LSTs and ground-based temperatures during the summer season. In the legend, the numbers in parentheses indicate Pearson’s correlation coeﬃcients between each variable legend, the numbers in parentheses indicate Pearson’s correlation coefficients between each variable and cloud cover (left) and solar radiation (right). Only statistically signiﬁcant correlations are given in and cloud cover (left) and solar radiation (right). Only statistically significant correlations are given bold. The magnitudes of change in cloud cover and solar radiation during summertime are included in in bold. The magnitudes of change in cloud cover and solar radiation during summertime are the title of each panel. The plotted series were arithmetically averaged for the whole country. included in the title of each panel. The plotted series were arithmetically averaged for the whole country.

Figure 16. Pearson’s r correlation coefficients calculated between LSTs and ground-based temperatures and cloud cover (left) and solar radiation (right). Correlations were computed on seasonal and annual bases for the period 2002–2015 for cloud cover and 2002–2014 for solar radiation. Prior to calculating the correlation, the series were detrended to avoid the impact of any climate signal present in the data on the strength of the correlation. Monthly cloud cover data were obtained from the global gridded CRU TS4.1 dataset (provided by the Climate Research Unit, East Angelia University, UK) at a spatial resolution of 0.5° grid, while daily solar radiation data were retrieved from Climate Forecast System Reanalysis (CFSR), provided by the National Centers for Environmental Prediction (NCEP). The statistical significance threshold of Pearson’s r is ±0.49.

Remote Sens. 2019, 11, x FOR PEER REVIEW 23 of 29 smooth topography. This homogeneity guarantees that the ground measurements can largely represent the satellite spatial scale, making any attempt to directly validate remotely sensed data against station data less complicated.

Remote Sens. 2019, 11, 2369 23 of 29

Table 3. Seasonal and annual changes in cloud cover (%/decade) and solar radiation (MJ/m2/decade) calculated for the base period 2002–2015 using ordinary least squares regression. Only statistically signiﬁcantFigure 15. Temporal changes at evolution the 95% level of the (p standardized< 0.05) are shown anoma inlies bold. of cloud cover (left) and solar radiation (right) in comparison to LSTs and ground-based temperatures during the summer season. In the Cloud Cover Solar Radiation legend, the numbers in parentheses indicate Pearson’s correlation coefficients between each variable Winter 1.79 2.24 and cloud cover (left) and solar radiation (right).− Only statistically significant correlations are given Spring 0.85 0.12 in bold. The magnitudes of change in cloud− cover and solar radiation during summertime are Summer 0.53 0.70 included in the title of each panel. The plotted series were arithmetically− averaged for the whole Autumn 0.08 0.39 − country. Annual 0.97 1.04 −

Figure 16. Pearson’s r correlation coefficients calculated between LSTs and ground-based temperatures andFigure cloud 16. cover Pearson’s (left) and r solarcorrelation radiation coefficients (right). Correlations calculated were between computed LSTs on seasonaland ground-based and annual basestemperatures for the period and cloud 2002–2015 cover for (left) cloud and cover solar and ra 2002–2014diation (right). for solar Correlations radiation. were Prior tocomputed calculating on theseasonal correlation, and annual the series bases werefor the detrended period 2002–2015 to avoid for the cloud impact cover of anyand climate2002–2014 signal for solar present radiation. in the dataPrior onto calculating the strength the of correlation, the correlation. the series Monthly were detrended cloud cover to avoid data werethe impact obtained of any from climate theglobal signal griddedpresent in CRU the TS4.1 data dataseton the strength (provided of bythe the correla Climatetion. Research Monthly Unit, cloud East cover Angelia data were University, obtained UK) from at a the global gridded CRU TS4.1 dataset (provided by the Climate Research Unit, East Angelia spatial resolution of 0.5◦ grid, while daily solar radiation data were retrieved from Climate Forecast SystemUniversity, Reanalysis UK) at (CFSR),a spatial provided resolution by of the 0.5° National grid, while Centers daily for solar Environmental radiation data Prediction were retrieved (NCEP). Thefrom statistical Climate significance Forecast thresholdSystem Reanalysis of Pearson’s (C rFSR), is 0.49. provided by the National Centers for Environmental Prediction (NCEP). The statistical significance± threshold of Pearson’s r is ±0.49. In accordance with earlier works (e.g., [18,22,25,57,72]), this study indicated that MODIS LST, especially for nighttime, can be trustworthily used for reliable diagnosis of regional climate change in Egypt, with numerous potential applications (e.g., environmental monitoring and assessment, urban climate, water resource management). In MODIS images, each pixel can serve as a thermal meteorological station by itself, providing meteorological information at a very detailed spatial scale (1 km). However, recalling the limited temporal coverage of MODIS LST, which dates back only to 2002, it is important to consider MOSIS LST as a surrogate for or supplementary source to ground measurements. In the same context, it is important to stress that although LSTs and air temperatures are strongly correlated, they have different physical meanings [4]. Following the guidelines of the WMO, the surface air temperature is typically recorded at 2 m above the ground using sensors protected from radiation, allowing for direct comparison between weather stations worldwide. On the other hand, similar to other remotely sensed variables, LST is estimated from sensors using the thermal infrared (TIR) radiation emitted by the land surface [90]. Moreover, LST is measured at the surface level and at temporal scales often different from the standard measurements of near-surface air temperatures. This sometimes makes any attempt to validate LST through a “direct” comparison between ground-based and satellite-based measurements infeasible due to scale mismatch, particularly in regions of complex topography, where the climate is highly variable even over short distances. These limitations motivated several previous works to assess the accuracy of LST by using synthetic data as predictors (e.g., land cover, vegetation greening, surface pressure, solar zenith angle, etc.) (e.g., [22,25]). Nonetheless, it is noteworthy that Egypt does not exhibit such strong spatial Remote Sens. 2019, 11, 2369 24 of 29 variations in terms of air temperature, topography, land cover, and soil characteristics, given its smooth topography. This homogeneity guarantees that the ground measurements can largely represent the satellite spatial scale, making any attempt to directly validate remotely sensed data against station data less complicated.

6. Conclusions In this study, we presented the first comprehensive assessment of the agreement between remotely sensed daytime and nighttime land surface temperatures (LSTs), retrieved from the Aqua/Moderate-Resolution Imaging Spectroradiometer (MODIS) sensor, and their corresponding ground-based measurements. Specifically, we validated LST retrievals against meteorological data from 34 stations spanning the period from 2002 to 2015. The accuracy of the LST in reproducing the climatology, anomalies, and trends in the near-surface air temperature was assessed using a variety of accuracy indicators (e.g., bias, Spearman’s rho correlation, normalized RMSE, and YK statistic), combined with an ordinary linear regression analysis and time-varying percentiles analysis. This work highlights the ability of LST to represent the spatial gradient, anomalies, and trends of the minimum air temperature. In contrast, LST showed less accuracy in capturing the climatology of the maximum air temperature, with a remarkable tendency to overestimate ground measurements. Seasonally, the daytime LST failed to represent temporal changes in the maximum air temperature, especially during warmer seasons (i.e., summer and autumn). Spatially, the performance of the LST is significantly linked to lithology and land use types. In particular, the LST showed better performance in arable lands in the delta and close to the Nile. This can be expected given that the delta is the main agricultural land in the country, where irrigated agriculture is the main land use all around the year. This greening enhances the exchange of incoming solar radiation via evaporation, making the daytime canopy temperature much closer to the air temperature. Also, this agricultural zone induces higher transpiration and an increase in latent heat fluxes, which are accompanied by an increase in evapotranspiration due to lack of surface resistance. All these processes keep the surface temperature much closer to the air temperature during the daytime. In contrast, the LST exhibited worse performance in regions of dark basement rocks (e.g., southern portions of the eastern plateau). Under the conditions of bare soil, low surface albedo, and less vegetation, the LST showed high uncertainty in these regions, with large deviations from the ground observations. This work stresses that MODIS LST can potentially be used as a good proxy for observed maximum and minimum air temperatures in Egypt, albeit with a better agreement between meteorological data and the MODIS-based nighttime LST. Overall, remotely sensed data could compensate for the paucity of the point network of meteorological stations, recalling that meteorological stations are not uniformly distributed across the country. While the current meteorological network cannot adequately reflect the “real” climatic conditions, especially at the regional and local scales, the LST is provided at high spatial (1 km) and temporal (daily) resolution, allowing for better understanding of land surface processes and energy balance changes at detailed spatial scales. As such, the spatially detailed LST could contribute to a more reliable diagnosis of the role of different climatic and environmental variables in regional climate variations in Egypt (e.g., elevation, lithology, continental and maritime influences, vegetation, etc.). Furthermore, MODIS LST can offer potential possibilities to evaluate the impacts of anthropogenic activities on local and regional climate change (e.g., land use changes, urbanization). In the absence of sufficient climatic records, LSTs can also be employed for the diagnosis of surface energy budgets and hydrological modeling, especially in a region with drastic water stress and scarcity like Egypt.

Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/11/20/2369/s1, Figure S1: Time-varying percentiles of Tmin distribution (left-hand panels) compared to those of nighttime LST (right-hand panels), Figure S2: Time-varying percentiles of Tmax distribution (left-hand panels) compared to those of daytime LST (right-hand panels). Remote Sens. 2019, 11, 2369 25 of 29

Author Contributions: Conceptualization, M.E.H. and A.M.E.K.; methodology, A.M.E.K.; software, M.E.H. and A.M.E.K.; validation, A.M.E.K.; formal analysis, A.M.E.K.; investigation, all authors; resources, S.M.R.; M.E.H.; data curation, M.E.H.; writing—original draft preparation, A.M.E.K.; writing—review and editing, S.M.R.; M.E.H.; visualization, A.M.E.K.; supervision, S.M.R..; project administration, M.E.H.; funding acquisition, A.M.E.K.; M.E.H. Funding: The research reported in this publication was supported by the HM Trust Fund (Strategic Project # SR/ART/GEOG/17/01) financed by Sultan Qaboos University, Oman. Acknowledgments: The authors are grateful to the Egyptian Meteorological Authority for providing the temperature data used in this study. We also would like to thank the ORNL DAAC for providing MODIS/VIIRS Land data. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Fontaine, B.; Janicot, S.; Monerie, P.-A. Recent changes in air temperature, heat waves occurrences, and atmospheric circulation in Northern Africa. J. Geophys. Res. Atmos. 2013, 118, 8536–8552. [CrossRef] 2. Donat, M.G.; Lowry, A.L.; Alexander, L.V.; O’Gorman, P.A.; Maher, N. More extreme precipitation in the world’s dry and wet regions. Nat. Clim. Chang. 2016, 6, 508. [CrossRef] 3. Seneviratne, S.I.; Stöckli, R. The Role of Land-Atmosphere Interactions for Climate Variability in Europe. In Climate Variability and Extremes during the Past 100 Years; Brönnimann, S., Luterbacher, J., Ewen, T., Diaz, H.F., Stolarski, R.S., Neu, U., Eds.; Springer: Dordrecht, The Netherlands, 2008; pp. 179–193. 4. Jin, M.; Dickinson, R.E. Land surface skin temperature climatology: Benefitting from the strengths of satellite observations. Environ. Res. Lett. 2010, 5, 041002. [CrossRef] 5. Peterson, T.C.; Easterling, D.R.; Karl, T.R.; Groisman, P.; Nicholls, N.; Plummer, N.; Torok, S.; Auer, I.; Boehm, R.; Gullett, D.; et al. Homogeneity adjustments of in situ atmospheric climate data: A review. Int. J. Climatol. 1998, 18, 1493–1517. [CrossRef] 6. Burke, M.; Dykema, J.; Lobell, D.B.; Miguel, E.S.S. Climate and Civil War: Is the Relationship Robust? NBER Work Pap 16440; National Bureau of Economic Research: Cambridge, MA, USA. 7. Smith, D.; Vivekananda, J. Climate Change, Conflict, and Fragility: Getting the Institutions Right. In Climate Change, Human Security and Violent Conflict: Challenges for Societal Stability; Scheffran, J., Brzoska, M., Brauch, H.G., Link, P.M., Schilling, J., Eds.; Springer: Berlin/Heidelberg, Germany, 2012; pp. 77–90. 8. Link, P.M.; Brücher, T.; Claussen, M.; Link, J.S.A.; Scheffran, J. The Nexus of Climate Change, Land Use, and Conflict: Complex Human–Environment Interactions in Northern Africa. Bull. Am. Meteorol. Soc. 2015, 96, 1561–1564. [CrossRef] 9. Paeth, H.; Born, K.; Girmes, R.; Podzun, R.; Jacob, D. Regional Climate Change in Tropical and Northern Africa due to Greenhouse Forcing and Land Use Changes. J. Clim. 2009, 22, 114–132. [CrossRef] 10. Hertig, E.; Seubert, S.; Jacobeit, J. Temperature extremes in the Mediterranean area: Trends in the past and assessments for the future. Nat. Hazards Earth Syst. Sci. 2010, 10, 2039–2050. [CrossRef] 11. Lelieveld, J.; Hadjinicolaou, P.; Kostopoulou, E.; Giannakopoulos, C.; Pozzer, A.; Tanarhte, M.; Tyrlis, E. Model projected heat extremes and air pollution in the eastern Mediterranean and Middle East in the twenty-first century. Reg. Environ. Chang. 2014, 14, 1937–1949. [CrossRef] 12. El Kenawy, A.M.; McCabe, M.F. Future projections of synoptic weather types over the Arabian Peninsula during the twenty-first century using an ensemble of CMIP5 models. Theor. Appl. Climatol. 2017, 130, 173–189. [CrossRef] 13. Hasanean, H.M. Wintertime surface temperature in Egypt in relation to the associated atmospheric circulation. Int. J. Climatol. 2004, 24, 985–999. [CrossRef] 14. Domroes, M.; El-Tantawi, A. Recent temporal and spatial temperature changes in Egypt. Int. J. Climatol. 2005, 25, 51–63. [CrossRef] 15. Hasanean, H.M.; Basset, H.A. Variability of summer temperature over Egypt. Int. J. Climatol. 2006, 26, 1619–1634. [CrossRef] 16. El Kenawy, A.; López-Moreno, J.I.; Vicente-Serrano, S.M.; Abd-Elaal, M. Temperature variability along the Mediterranean Coast and its links to Large Scale Atmospheric Circulations (1957–2006). Bull. Soc. Geogr. d’Egypte 2010, 83, 12–26. Remote Sens. 2019, 11, 2369 26 of 29

17. Donat, M.G.; Peterson, T.C.; Brunet, M.; King, A.D.; Almazroui, M.; Kolli, R.K.; Boucherf, D.; Al-Mulla, A.Y.; Nour, A.Y.; Aly, A.A.; et al. Changes in extreme temperature and precipitation in the Arab region: Long-term trends and variability related to ENSO and NAO. Int. J. Climatol. 2014, 34, 581–592. [CrossRef] 18. El Kenawy, A.M.; Lopez-Moreno, J.I.; McCabe, M.F.; Robaa, S.M.; Domínguez-Castro, F.; Peña-Gallardo, M.; Trigo, R.M.; Hereher, M.E.; Al-Awadhi, T.; Vicente-Serrano, S.M. Daily temperature extremes over Egypt: Spatial patterns, temporal trends, and driving forces. Atmos. Res. 2019, 226, 219–239. [CrossRef] 19. El Kenawy, A.M.; McCabe, M.F. A multi-decadal assessment of the performance of gauge- and model-based rainfall products over Saudi Arabia: Climatology, anomalies and trends. Int. J. Climatol. 2016, 36, 656–674. [CrossRef] 20. Rasul, A.; Balzter, H.; Smith, C. Spatial variation of the daytime Surface Urban Cool Island during the dry season in Erbil, Iraqi Kurdistan, from Landsat 8. Urban Clim. 2015, 14, 176–186. [CrossRef] 21. Justice, C.O.; Townshend, J.R.G.; Vermote, E.F.; Masuoka, E.; Wolfe, R.E.; Saleous, N.; Roy, D.P.; Morisette, J.T. An overview of MODIS Land data processing and product status. Remote Sens. Environ. 2002, 83, 3–15. [CrossRef] 22. Mildrexler, D.J.; Zhao, M.; Running, S.W. A global comparison between station air temperatures and MODIS land surface temperatures reveals the cooling role of forests. J. Geophys. Res. Biogeosci. 2011, 116.[CrossRef] 23. Mao, K.B.; Ma, Y.; Tan, X.L.; Shen, X.Y.; Liu, G.; Li, Z.L.; Chen, J.M.; Xia, L. Global surface temperature change analysis based on MODIS data in recent twelve years. Adv. Space Res. 2017, 59, 503–512. [CrossRef] 24. Eleftheriou, D.; Kiachidis, K.; Kalmintzis, G.; Kalea, A.; Bantasis, C.; Koumadoraki, P.; Spathara, M.E.; Tsolaki, A.; Tzampazidou, M.I.; Gemitzi, A. Determination of annual and seasonal daytime and nighttime trends of MODIS LST over Greece—Climate change implications. Sci. Total Environ. 2018, 616–617, 937–947. [CrossRef][PubMed] 25. Khorchani, M.; Vicente-Serrano, S.M.; Azorin-Molina, C.; Garcia, M.; Martin-Hernandez, N.; Peña-Gallardo, M.; el Kenawy, A.; Domínguez-Castro, F. Trends in LST over the peninsular Spain as derived from the AVHRR imagery data. Glob. Planet. Chang. 2018, 166, 75–93. [CrossRef] 26. Zhang, Y.; Odeh, I.O.A.; Han, C. Bi-temporal characterization of land surface temperature in relation to impervious surface area, NDVI and NDBI, using a sub-pixel image analysis. Int. J. Appl. Earth Obs. Geoinf. 2009, 11, 256–264. [CrossRef] 27. He, G.; Zhao, Y.; Wang, J.; Wang, Q.; Zhu, Y. Impact of large-scale vegetation restoration project on summer land surface temperature on the Loess Plateau, China. J. Arid Land 2018, 10, 892–904. [CrossRef] 28. Gomis-Cebolla, J.; Jimenez, J.C.; Sobrino, J.A. LST retrieval algorithm adapted to the Amazon evergreen forests using MODIS data. Remote Sens. Environ. 2018, 204, 401–411. [CrossRef] 29. Kogan, F.N. Global Drought Watch from Space. Bull. Am. Meteorol. Soc. 1997, 78, 621–636. [CrossRef] 30. Karnieli, A.; Agam, N.; Pinker, R.T.; Anderson, M.; Imhoﬀ, M.L.; Gutman, G.G.; Panov, N.; Goldberg, A. Use of NDVI and Land Surface Temperature for Drought Assessment: Merits and Limitations. J. Clim. 2010, 23, 618–633. [CrossRef] 31. AghaKouchak, A.; Farahmand, A.; Melton, F.S.; Teixeira, J.; Anderson, M.C.; Wardlow, B.D.; Hain, C. Remote sensing of drought: Progress, challenges and opportunities. Rev. Geophys. 2015, 53, 452–480. [CrossRef] 32. Quintano, C.; Fernández-Manso, A.; Calvo, L.; Marcos, E.; Valbuena, L. Land surface temperature as potential indicator of burn severity in forest Mediterranean ecosystems. Int. J. Appl. Earth Obs. Geoinf. 2015, 36, 1–12. [CrossRef] 33. Quintano, C.; Fernandez-Manso, A.; Roberts, D.A. Burn severity mapping from Landsat MESMA fraction images and Land Surface Temperature. Remote Sens. Environ. 2017, 190, 83–95. [CrossRef] 34. Peng, S.-S.; Piao, S.; Zeng, Z.; Ciais, P.; Zhou, L.; Li, L.Z.X.; Myneni, R.B.; Yin, Y.; Zeng, H. Aﬀorestation in China cools local land surface temperature. Proc. Natl. Acad. Sci. USA 2014, 111, 2915–2919. [CrossRef] [PubMed] 35. Fu, P.; Weng, Q. A time series analysis of urbanization induced land use and land cover change and its impact on land surface temperature with Landsat imagery. Remote Sens. Environ. 2016, 175, 205–214. [CrossRef] 36. Zhang, Y.; Balzter, H.; Liu, B.; Chen, Y. Analyzing the Impacts of Urbanization and Seasonal Variation on Land Surface Temperature Based on Subpixel Fractional Covers Using Landsat Images. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2017, 10, 1344–1356. [CrossRef] Remote Sens. 2019, 11, 2369 27 of 29

37. Amiri, R.; Weng, Q.; Alimohammadi, A.; Alavipanah, S.K. Spatial–temporal dynamics of land surface temperature in relation to fractional vegetation cover and land use/cover in the Tabriz urban area, Iran. Remote Sens. Environ. 2009, 113, 2606–2617. [CrossRef] 38. Jafari, R.; Hasheminasab, S. Assessing the effects of dam building on land degradation in central Iran with Landsat LST and LULC time series. Environ. Monit. Assess. 2017, 189, 74. [CrossRef] 39. Rhee, J.; Im, J.; Carbone, G.J. Monitoring agricultural drought for arid and humid regions using multi-sensor remote sensing data. Remote Sens. Environ. 2010, 114, 2875–2887. [CrossRef] 40. Yi, Y.; Yang, D.; Chen, D.; Huang, J. Retrieving crop physiological parameters and assessing water deficiency using MODIS data during the winter wheat growing period. Can. J. Remote Sens. 2007, 33, 189–202. [CrossRef] 41. Albright, T.P.; Pidgeon, A.M.; Rittenhouse, C.D.; Clayton, M.K.; Flather, C.H.; Culbert, P.D.; Radeloff, V. Heat waves measured with MODIS land surface temperature data predict changes in avian community structure. Remote Sens. Environ. 2011, 115, 245–254. [CrossRef] 42. Orimoloye, I.R.; Mazinyo, S.P.; Nel, W.; Kalumba, A.M. Spatiotemporal monitoring of land surface temperature and estimated radiation using remote sensing: Human health implications for East London, South Africa. Environ. Earth Sci. 2018, 77, 77. [CrossRef] 43. Jish Prakash, P.; Stenchikov, G.; Kalenderski, S.; Osipov, S.; Bangalath, H. The impact of dust storms on the Arabian Peninsula and the Red Sea. Atmos. Chem. Phys. 2015, 15, 199–222. [CrossRef] 44. Almazroui, M. Calibration of TRMM rainfall climatology over Saudi Arabia during 1998–2009. Atmos. Res. 2011, 99, 400–414. [CrossRef] 45. El Kenawy, A.M.; Lopez-Moreno, J.I.; McCabe, M.F.; Vicente-Serrano, S.M. Evaluation of the TMPA-3B42 precipitation product using a high-density rain gauge network over complex terrain in northeastern Iberia. Glob. Planet. Chang. 2015, 133, 188–200. [CrossRef] 46. Sun, L.; Mi, X.; Wei, J.; Wang, J.; Tian, X.; Yu, H.; Gan, P. A cloud detection algorithm-generating method for remote sensing data at visible to short-wave infrared wavelengths. ISPRS J. Photogramm. Remote Sens. 2017, 124, 70–88. [CrossRef] 47. Tzallas, V.; Hatzianastassiou, N.; Benas, N.; Meirink, J.F.; Matsoukas, C.; Stackhouse, P.; Vardavas, I. Evaluation of CLARA-A2 and ISCCP-H Cloud Cover Climate Data Records over Europe with ECA & D Ground-Based Measurements. Remote Sens. 2019, 11, 212. 48. Peng, J.; Loew, A.; Merlin, O.; Verhoest, N.E.C. A review of spatial downscaling of satellite remotely sensed soil moisture. Rev. Geophys. 2017, 55, 341–366. [CrossRef] 49. Fontanet, M.; Fernàndez-Garcia, D.; Ferrer, F. The value of satellite remote sensing soil moisture data and the DISPATCH algorithm in irrigation fields. Hydrol. Earth Syst. Sci. 2018, 22, 5889–5900. [CrossRef] 50. Zhang, Y.; Yan, S.; Lu, Y. Snow Cover Monitoring Using MODIS Data in Liaoning Province, Northeastern China. Remote Sens. 2010, 2, 777–793. [CrossRef] 51. Naderpour, R.; Schwank, M. Snow Wetness Retrieved from L-Band Radiometry. Remote Sens. 2018, 10, 359. [CrossRef] 52. El Shazly, E.M. The Geology of the Egyptian Region. In The Ocean Basins and Margins: Volume 4A the Eastern Mediterranean; Nairn, A.E.M., Kanes, W.H., Stehli, F.G., Eds.; Springer: Boston, MA, USA, 1977; pp. 379–444. 53. Stanley, D.J.; Warne, A.G. Nile Delta: Recent Geological Evolution and Human Impact. Science 1993, 260, 628–634. [CrossRef] 54. Peel, M.C.; Finlayson, B.L.; McMahon, T.A. Updated world map of the Köppen-Geiger climate classification. Hydrol. Earth Syst. Sci. 2007, 11, 1633–1644. [CrossRef] 55. El-Shirbeny, M.; Ali, A.; Saleh, N. Crop Water Requirements in Egypt Using Remote Sensing Techniques. J. Agric. Chem. Environ. 2014, 3, 57–65. [CrossRef] 56. Kimura, R.; Kato, H.; Iwasaki, E. Cultivation Features Using Meteorological and Satellite Data from 2001 to 2010 in Dakhla Oasis, Egypt. J. Water Resour. Prot. 2015, 7, 209–218. [CrossRef] 57. Wan, Z.; Wang, P.; Li, X. Using MODIS Land Surface Temperature and Normalized Difference Vegetation Index products for monitoring drought in the southern Great Plains, USA. Int. J. Remote Sens. 2004, 25, 61–72. [CrossRef] 58. Hereher, M.E. Recent trends of temperature and precipitation proxies in Saudi Arabia: Implications for climate change. Arab. J. Geosci. 2016, 9, 575. [CrossRef] Remote Sens. 2019, 11, 2369 28 of 29

59. Hereher, M.E. Time series trends of land surface temperatures in Egypt: A signal for global warming. Environ. Earth Sci. 2016, 75, 1218. [CrossRef] 60. Hereher, M.E. Estimation of monthly surface air temperatures from MODIS LST time series data: Application to the deserts in the Sultanate of Oman. Environ. Monit. Assess. 2019, 191, 592. [CrossRef] 61. Hereher, M.E. Effects of land use/cover change on regional land surface temperatures: Severe warming from drying Toshka lakes, the Western Desert of Egypt. Nat. Hazards 2017, 88, 1789–1803. [CrossRef] 62. Deng, Y.; Wang, S.; Bai, X.; Tian, Y.; Wu, L.; Xiao, J.; Chen, F.; Qian, Q. Relationship among land surface temperature and LUCC, NDVI in typical karst area. Sci. Rep. 2018, 8, 641. [CrossRef] 63. Li, Z.-L.; Tang, B.-H.; Wu, H.; Ren, H.; Yan, G.; Wan, Z.; Trigo, I.F.; Sobrino, J.A. Satellite-derived land surface temperature: Current status and perspectives. Remote Sens. Environ. 2013, 131, 14–37. [CrossRef] 64. Becker, F.; Li, Z.-L. Towards a local split window method over land surfaces. Int. J. Remote Sens. 1990, 11, 369–393. [CrossRef] 65. Wan, Z.; Li, Z. A physics-based algorithm for retrieving land-surface emissivity and temperature from EOS/MODIS data. IEEE Trans. Geosci. Remote Sens. 1997, 35, 980–996. 66. Snyder, W.C.; Wan, Z.; Zhang, Y.; Feng, Y.Z. Classification-based emissivity for land surface temperature measurement from space. Int. J. Remote Sens. 1998, 19, 2753–2774. [CrossRef] 67. Wan, Z.; Dozier, J. A generalized split-window algorithm for retrieving land-surface temperature from space. IEEE Trans. Geosci. Remote Sens. 1996, 34, 892–905. 68. Qin, Z.; Karnieli, A.; Berliner, P. A mono-window algorithm for retrieving land surface temperature from Landsat TM data and its application to the Israel-Egypt border region. Int. J. Remote Sens. 2001, 22, 3719–3746. [CrossRef] 69. Li, X.; Zhou, Y.; Asrar, G.R.; Zhu, Z. Creating a seamless 1km resolution daily land surface temperature dataset for urban and surrounding areas in the conterminous United States. Remote Sens. Environ. 2018, 206, 84–97. [CrossRef] 70. Ferro, C.A.; Hannachi, A.; Stephenson, D.B. Simple Nonparametric Techniques for Exploring Changing Probability Distributions of Weather. J. Clim. 2005, 18, 4344–4354. [CrossRef] 71. Hamed, K.H.; Ramachandra Rao, A. A modified Mann-Kendall trend test for autocorrelated data. J. Hydrol. 1998, 204, 182–196. [CrossRef] 72. Luintel, N.; Ma, W.; Ma, Y.; Wang, B.; Subba, S. Spatial and temporal variation of daytime and nighttime MODIS land surface temperature across Nepal. Atmos. Ocean. Sci. Lett. 2019, 12, 305–312. [CrossRef] 73. Vancutsem, C.; Ceccato, P.; Dinku, T.; Connor, S.J. Evaluation of MODIS land surface temperature data to estimate air temperature in different ecosystems over Africa. Remote Sens. Environ. 2010, 114, 449–465. [CrossRef] 74. Benali, A.; Carvalho, A.C.; Nunes, J.P.; Carvalhais, N.; Santos, A. Estimating air surface temperature in Portugal using MODIS LST data. Remote Sens. Environ. 2012, 124, 108–121. [CrossRef] 75. El Kenawy, A.M.; McCabe, M.F.; Vicente-Serrano, S.M.; Robaa, S.M.; Lopez-Moreno, J.I. Recent changes in continentality and aridity conditions over the Middle East and North Africa region, and their association with circulation patterns. Clim. Res. 2016, 69, 25–43. [CrossRef] 76. Cho, J.; Oki, T.; Yeh, P.J.F.; Kim, W.; Kanae, S.; Otsuki, K. On the relationship between the Bowen ratio and the near-surface air temperature. Theor. Appl. Climatol. 2011, 108, 135–145. [CrossRef] 77. Abu El-Liel, I.; Soliman, N.; Bekhiet, M.; Elhebiry, M.S. Lithological mapping in the eastern desert of Egypt, Wadi Um Gheig area, using LANDSAT enhanced thematic mapper (ETM+). Al Azhar Bull. Sci. 2014, 25, 1–8. [CrossRef] 78. Dirmeyer, P.A.; Shukla, J. The effect on regional and global climate of expansion of the world’s deserts. Q. J. R. Meteorol. Soc. 1996, 122, 451–482. [CrossRef] 79. Zeng, N.; Neelin, J.D.; Lau, K.-M.; Tucker, C.J. Enhancement of Interdecadal Climate Variability in the Sahel by Vegetation Interaction. Science 1999, 286, 1537–1540. [CrossRef] 80. Marland, G.; Pielke, R.A.; Apps, M.; Avissar, R.; Betts, R.A.; Davis, K.J.; Frumhoff, P.C.; Jackson, S.T.; Joyce, L.A.; Kauppi, P.; et al. The climatic impacts of land surface change and carbon management, and the implications for climate-change mitigation policy. Clim. Policy 2003, 3, 149–157. [CrossRef] 81. Mu, Q.; Jones, L.A.; Kimball, J.S.; McDonald, K.C.; Running, S.W. Satellite assessment of land surface evapotranspiration for the Pan-Arctic domain. Water Resour. Res. 2009, 45, W09420. [CrossRef] Remote Sens. 2019, 11, 2369 29 of 29

82. Yuan, F.; Bauer, M.E. Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery. Remote Sens. Environ. 2007, 106, 375–386. [CrossRef] 83. Zeng, N.; Yoon, J. Expansion of the world’s deserts due to vegetation-albedo feedback under global warming. Geophys. Res. Lett. 2009, 36.[CrossRef] 84. Dai, A.; Genio, A.D.D.; Fung, I.Y. Clouds, precipitation and temperature range. Nature 1997, 386, 665–666. [CrossRef] 85. Hoerling, M.; Eischeid, J.; Perlwitz, J.; Quan, X.; Zhang, T.; Pegion, P. On the increased frequency of mediterraneanMediterranean drought. J. Clim. 2012, 25, 2146–2161. [CrossRef] 86. L’Heureux, M.L.; Collins, D.C.; Hu, Z.-Z. Linear trends in sea surface temperature of the tropical Pacific Ocean and implications for the El Niño-Southern Oscillation. Clim. Dyn. 2013, 40, 1223–1236. [CrossRef] 87. Noori, R.; Tian, F.; Berndtsson, R.; Abbasi, M.R.; Naseh, M.V.; Modabberi, A.; Soltani, A.; Kløve, B. Recent and future trends in sea surface temperature across the Persian Gulf and Gulf of Oman. PLoS ONE 2019, 14, e0212790. [CrossRef][PubMed] 88. Shaltout, M. Recent sea surface temperature trends and future scenarios for the Red Sea. Oceanologia 2019, 61, 484–504. [CrossRef] 89. Nashwan, M.S.; Shahid, S.; Abd Rahim, N. Unidirectional trends in annual and seasonal climate and extremes in Egypt. Theor. Appl. Climatol. 2019, 136, 457–473. [CrossRef] 90. Wan, Z.; Li, Z.L. Radiance-based validation of the V5 MODIS land-surface temperature product. Int. J. Remote Sens. 2008, 29, 5373–5395. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).