Diurnal Variation Characteristics of Precipitation in Summer Associated with Diverse Underlying Surfaces in the Arid Region of Eastern Xinjiang, Northwest China

Highlights

What are the main findings?

• High-resolution WRF-NJU simulation data reliably captured the diurnal variation characteristics of precipitation (DVCP) in eastern Xinjiang, consistent with multiple observational and reanalysis datasets.
• The DVCP shows pronounced spatiotemporal differences across basins and mountains, with distinct precipitation peaks at different elevations and underlying surface types.

What is the implication of the main finding?

• The results highlight the crucial role of complex topography and land surface conditions in shaping precipitation diurnal cycles, providing new insights into precipitation mechanisms in arid regions.
• These findings support improved agricultural water management and more rational allocation of water resources, while also informing disaster risk reduction in arid and semi-arid regions.

Abstract

Investigating the diurnal variation characteristics of precipitation (DVCP) in Xinjiang, an arid region of Northwest China, is essential for improving water resource management and disaster mitigation strategies. This study examines the DVCP associated with diverse underlying surfaces in Eastern Xinjiang (EX)—one of the most arid regions in China—during summer (June–August) from 2015 to 2019, using hourly simulation data from the real-time forecasting system of Nanjing University (WRF_NJU). Evaluation against automatic weather station (AWS) observations indicates that WRF_NJU outperforms reanalysis (ERA5), satellite (CMORPH), and MESWEP datasets, demonstrating its reliability for regional precipitation analysis. Further investigation reveals that in the Turpan-Hami Basin (THB), below 1000 m above sea level (ASL), peaks in precipitation amount (PA), intensity (PI), and frequency (PF) occur at 06 local solar time (LST), whereas in mountainous areas above 3000 m ASL, these peaks are delayed until 13 LST. Analysis of the coefficient of variation (CV) shows that the most pronounced differences in DVCP between mountainous and basin regions are associated with PF and PI. Specifically, regions with high CV for PF are concentrated in the central to northern parts of the THB, while high CV for PI is found in the eastern Mid-Tianshan Mountains (MTM) and East Tianshan Mountains (ETM). Moreover, significant differences in DVCP are observed across land surface types: PA peaks over grasslands, forests, and water bodies occur around noon, whereas over impervious surfaces, croplands, and barren areas, they occur during the early morning hours.

1. Introduction

Precipitation is a critical component of the hydrological cycle, exerting significant controls on hydro-meteorological processes—particularly in arid regions where water scarcity amplifies ecological vulnerability [1]. Against the backdrop of intensifying global climate change, the frequency and magnitude of heavy precipitation events are projected to increase substantially (IPCC 2021) [2]. These changes may trigger a series of potential weather disasters, such as torrential rain, high winds, hails, and sometimes tornadoes, posing a serious threat to human life, property, and the economy [3,4,5,6,7,8,9]. Therefore, understanding the spatiotemporal characteristics of precipitation amount (PA), intensity (PI), frequency (PF) is essential for improving meteorological risk management, optimizing water resource utilization, and advancing cloud water resource assessment in arid zones [10,11].

The DVCP is one of important characteristics of weather and climate, reflecting the response of the land–atmosphere system due to diurnal variation characteristics (DVC) of solar radiation forcing received by a particular region due to the rotation of the Earth [12,13,14,15]. Extensive studies on the DVCP have been conducted on the DVCP which was beneficial to gain a deeper understanding of the climate system and improve precipitation forecasting [13]. Most of the traditional studies focused on accumulated or time-averaged precipitation, whereas recent work has shifted to changes in PA, PF, and PI and to how topography and underlying-surface characteristics affect the DVCP [14,15,16,17,18,19].

Critically, underlying surface properties—including both natural terrain and human-modified landscapes—fundamentally modulate DVCP. Topography, as a key component of underlying surfaces, regulates convective triggering timing through thermal contrasts induced by elevation gradients and mountain-valley circulations, thereby controlling peak precipitation occurrence [18]. The DVC of extreme hourly precipitation are generally consistent with those of total precipitation, but these characteristics show significant spatial differences under different large-scale circulations and underlying surface characteristics [19]. For instance, urbanization-induced heat islands amplify heavy precipitation risks in cities, while mountain-valley circulations shift precipitation peaks downstream [17,18,19,20,21].

Some previous studies have shown that the height inhomogeneity of local mountain topography drives the formation of thermal circulation. This circulation, induced by differential ground heating, subsequently shapes the DVCP [22]. In high mountain regions, convective precipitation systems usually form during the day. They then move downstream in a regular pattern and produce a nighttime precipitation peak in downstream areas [23]. Land use changes, especially urbanization, also exert a clear influence on the DVCP [24,25]. Therefore, investigating the DVCP and its drivers serves two purposes. It deepens our understanding of local climate and provides a benchmark for validating parameterizations in numerical models [17]. Many studies have found that both PA and PF of heavy precipitation events have increased significantly in urban areas. This increase heightens the risk of flooding [26,27,28,29,30,31]. Using ground-based rain-gauge data, Shepherd and Burian [32] demonstrated that the urban heat-island (UHI) effect in Houston, USA, causes anomalous changes in urban precipitation. Meanwhile, research in India shows that the probability of heavy precipitation in urban areas is much higher than in non-urban areas. The upward trend of extreme precipitation events in cities is also becoming more evident [33].

Early studies on the DVCP based on observational precipitation data often faced problems of insufficient reliability, limiting in-depth analysis of precipitation changes [34]. With the implementation of the Tropical Rainfall Measuring Mission (TRMM) data, scientists obtained reliable global-scale satellite data, significantly enhancing the understanding of the DVCP. Among many datasets, the WRF (Weather Research and Forecasting) simulation data, with its superior tempo-spatial resolution, extensive coverage of many 3D meteorological elements, and high-quality data, has become an ideal data for studying the DVCP, because the WRF simulation data has the ability to effectively avoid difficulties commonly encountered by satellite and radar observations, such as sensitivity to clouds and beam blockage caused by topography [35].

The DVCP and its physical mechanisms in China show significant regional and seasonal differences. Han Et Al. [36] revealed the DVCP of summer precipitation in North China using precipitation data observed by the automatic weather stations (AWSs) from 2008 to 2014 and the CMORPH precipitation data. They found that most areas in North China show a distinct bimodal feature in the PA and PF, with significant regional differences. Moreover, in the western part of the Taihang Mountains in North China, the daily peak of PA and PF usually occurs in the evening, while in the plains and coastal areas to the east of the Taihang Mountains, the daily peak generally occurs in the morning. The DVCP in summer over the middle-to-lower reaches of the Yangtze River has two peaks, occurring during the afternoon and early morning period [37,38]. The afternoon peak is often caused by short-lived (less than 3 h) convective activity, while the nighttime to early morning peak is related to the activity of long-lived (more than 6 h) convective systems, with large-scale forcing and the topography of the Tibetan Plateau playing an important modulating role [39]. The daily precipitation peak in the coastal areas of South China occurs in the early morning, with strong wind days mainly related to orographic lift and land–sea friction differences, while weak wind days are mainly related to land–sea breezes [40]. The daily precipitation peak in the inland mountainous areas of South China occurs in the afternoon, closely related to orographic thermal effects and land–sea breezes [41]. Some studies have also shown that there are two daily precipitation peaks in the Tibetan Plateau, occurring in the middle of the night and in the afternoon, while the precipitation peak in the adjacent areas to the east of the Tibetan Plateau mainly occurs in the middle of the night [42,43].

Generally speaking, most studies have been concentrated in the relatively developed regions of central, eastern, northern, and southern China, with fewer studies in the western regions [44,45,46,47,48,49,50]. The overall climatic characteristics of precipitation in the Tianshan mountainous area are strongly affected by topography [51], yet the extreme aridity and complex underlying surfaces (oasis-desert transitions) in Eastern Xinjiang (EX, Figure 1a) likely drive unique DVCP and related mechanisms that are probably not similar with that of the humid regions. However, the unique DVCP affected by the extreme aridity and complex underlying surfaces in the EX region remain poorly quantified. In addition, a systematic quantification of the unique DVCP, particularly the characteristics related to the elevations and specific land surface types, is still lacking. To address this critical gap, this study leverages high-resolution WRF_NJU simulations to investigate the summer DVCP in the EX region by addressing the following research questions (RQs):

RQ1: What are the spatial characteristics of the PA, PF, and PI in the EX region, and which regions exhibit the most pronounced differences between mountains and basins?

RQ2: How do the diurnal peaks of PA, PF, and PI differ between the low-lying Turpan-Hami Basin (<1000 m ASL) and the high-altitude mountainous areas (>3000 m ASL)?

RQ3: How does the timing of precipitation peaks vary across major underlying surface types (barren areas, croplands, grasslands, water bodies) in the EX region?

By answering these questions, this study aims to elucidate the DVCP related to the topography and land surface processes on precipitation regimes in one of China’s most arid regions. This study may help further understand regional climate dynamics and support flood prevention strategies.

The remainder of the present paper is organized as follows. The data and methods utilized in this work are depicted in Section 2, and the results, discussions and conclusions are described in Section 3, Section 4 and Section 5, respectively.

2. Overview of the Study Area

The EX region comprises two major regions to the north and south of the East-Tianshan Mountains (ETM), and there is the Naomao Lake Basin to the north of the ETM. The other part is the oasis belt to the south of the ETM, consisting of the Turpan-Hami Basin (Figure 1d). Located in the easternmost part of Xinjiang, at the end of the Tianshan Mountains (TM) and far from the ocean, the EX region has a temperate continental arid and semi-arid climate. The terrain is complex, and precipitation distribution within the year is highly uneven. Particularly during the flood season, the non-uniform distribution of precipitation in time and space can easily lead to localized drought and flood disasters, severely restricting local economic development and the standard of living. The TM run through the EX region. The EX region’s elevation can be categorized into seven classes at 500 m intervals. Since the last category (3000–4878 m) has a very small area, it was combined into one class (as shown in Figure 1a) for subsequent research. The EX region mainly includes Yiwu County, Yizhou District, Xinxiang City, Balikun Kazak Autonomous County, Shanshan County, Gaochang District, and Tuokexun County (Figure 1b). The main land cover types in eastern Xinjiang are forests, shrubs, wetlands, snow and ice, grasslands, bare land, croplands, water bodies, and impervious surfaces. Centered around cities like Turpan and Hami, an urban agglomeration has formed in the eastern part of the TM. The TM divide the city into northern and southern parts. The northern part consists of forests, grasslands, snow-capped mountains, and glaciers, while the southern part is an oasis surrounded by Gobi deserts (Figure 1c).

3. Data and Methods

3.1. Data

Our present study used precipitation data (hourly) from 121 automatic weather stations (AWSs) during summer (from June to August) from 2015 to 2019, ERA5-land data, WRF_NJU data with a 4 km resolution, CMORPH precipitation data with a 0.25° resolution, and MSWEP (Multi-Source Weighted-Ensemble Precipitation) data with a 3hourly and 0.1° resolution. The AWS data were derived from the CMA (China Meteorological Administration). Furthermore, the ERA5-land data were obtained from the ECMWF. The CMORPH data, produced by the NOAA Climate Prediction Center, integrates multi-platform satellite observations to create a high-resolution global precipitation product [42]. The MSWEP precipitation data, with its global reach, combines data from ground-based observations, satellite measurements, and reanalysis datasets to produce high-quality product of precipitation [52]. The CMORPH and MSWEP precipitation data offer broad coverage and timeliness, effectively compensating for the limitations of ground-based data from rain-gauges and meteorological radar observations. Furthermore, the China Land Cover Dataset (CLCD) [53] with a spatial resolution of 30 m is used in this study to present the land use types over the study area. The CLCD was generated on the Google Earth Engine (GEE) platform, and it offers annual land cover information and its changes across China from 1990 to 2023. This dataset was created by interpreting time-series satellite imagery alongside references from Google Earth and Google Maps. Through processing 335,709 Landsat images on GEE, multiple temporal metrics were developed and input into a random forest (RF) classifier to generate the land cover classifications. Given that the WRF_NJU dataset used in this study covers the period 2015–2019, the land use data from the central year of 2017 was selected for illustrating the land use characteristics of the study area to ensure maximal representativeness.

The WRF_NJU forecasting system, operating at a 4 km horizontal resolution, has produced real-time summer forecasts for China twice per day since 2013 [54]. Its computational domain consists of 1409 × 1081 grid points with a spatial interval of 4 km and 51 layers in the vertical direction. The principal physical parameterizations employed include the Morrison double-moment microphysics scheme [55]; the CAM radiation schemes for both shortwave and longwave processes [56]; the Pleim–Xiu land-surface and surface-layer schemes [57]; and the Asymmetrical Convective Model, Version 2, planetary boundary layer scheme [58]. To remove or reduce model bias, this study used the 12–36 h forecast results as the precipitation data for the next day, excluding the initial 12 h spin-up period.

3.2. Methods

The calculation formula of precipitation amount (PA):

$$PA=\sum _{d=1}^{n}p(h,d)/n$$

(1)

where the p(h,d) indicates the hourly PA at time h (range from 0 to 23) on the d day. The n is the number of days with the p(h,d) ≥ 0.01 mm.

The calculation formula of precipitation frequency (PF):

$$PF=\sum _{d=1}^{n}pn(h,d)/n\times 100\%$$

(2)

where the pn(h,d) represents the number of precipitation occurrences (≥0.01 mm) at time h on the d day. When the pn(h,d) = 1, it indicates that a precipitation of 0.01 mm is counted as an occurrence of precipitation at time h, and PA < 0.01 is counted as 0 occurrences.

The calculation formula of precipitation intensity (PI):

$$PI={\displaystyle \frac{PA}{PF}}$$

(3)

The PI indicates the total precipitation amount at a specific time divided by the total number of precipitation occurrences at that time.

The calculation formula of coefficient of variation (CV):

$$CV={\displaystyle \frac{\sigma }{\mu }}\times 100\%$$

(4)

The $\sigma$ and $\mu$ represent the standard deviation and mean value of the variable to be calculated (PA). The CV was usually used to assess the degree of dispersion (from temporal and spatial perspective) of a variable.

According to the geographical location of EX, the time divisions are shown in

Table 1. The time slot labels and their associated time intervals employed in this study for analyzing the DVCP in this study.

Time Slot Name	Time Range (LST = UTC + 6)
Midnight	2300–0100
Early morning	0200–0400
Dawn	0500–0700
Morning	0800–1000
Noon	1100–1300
Afternoon	1400–1600
Nightfall	1700–1900
Evening	2000–2200

below.

4. Results

4.1. Evaluation of the Results of the WRF_NJU Simulation

The spatial distribution of the summer precipitation derived from the data of WRF_NJU also indicated the unevenness of precipitation in this region. The annual average precipitation in mountainous areas is above 100 mm, while in the basins, it is below 50 mm (Figure 2).

Using WRF_NJU data, AWS data, CMORPH data, ERA5-Land data, and MSWEP data from June to August during 2015–2019, we analyzed the spatial distribution of PA at 0000, 0600, 1200, and 1800 LST in EX. The results show that at 0000 LST (Figure 3a–e), AWS data (Figure 3a) indicate that precipitation increases gradually from the southern plains to the northern mountainous areas and then decreases towards the northern basins of the EX region. Heavy precipitation is mainly distributed in the eastern part of the Mid-Tianshan Mountains (MTM) and the ETM (indicated by the black ellipse), with a PA of 0.2 mm at the middle of the heavy precipitation area. Lower PA (0.001–0.025 mm) are found in the Turpan-Hami Basin (indicated by the red ellipse), and almost no precipitation occurs in the Naomao Lake Basin in the northeastern portion of the EX region (indicated by the red ellipse). The PA distribution derived from WRF_NJU (Figure 3b) shows some deviations in the eastern MTM and the ETM, with a PA of 0.175 mm at the heavy precipitation center. However, the overall trend of PA distribution is roughly similar to that shown by the data of AWSs, but it is about 0.5° to the south of the observed data. This is inconsistent with the conclusions drawn by our previous study [5] in their study of atypical blocking processes caused by the merging of sea breeze fronts and gust fronts. They found that the simulated northwestern convective system developed about 1.5 h later than the observed system and was about 0.8° to the south of the observed system.

Despite some deviations in time and location, the overall pattern of WRF simulation was reproduced. However, the PA distribution derived from ERA5-Land data (Figure 3c) shows stronger PI in the mountainous areas, but the spatial distribution trend is consistent with WRF_NJU data. The PA distribution derived from CMORPH data (Figure 3d) shows that the overall precipitation is relatively weak, but the precipitation in the mountainous areas is stronger than that in the basins, which is inconsistent with the spatial distribution of the data of WRF_NJU. Moreover, the lower PA values in the Turpan-Hami Basin and Naomao Lake Basin are very similar to those derived from WRF_NJU data. The PA distribution derived from data of MSWEP (Figure 3e) indicates that the PA at the heavy precipitation center is 0.1 mm, which is about 0.075 mm lower than that derived from WRF_NJU data. The banded precipitation pattern in the eastern part of the MTM and the ETM is consistent with that derived from WRF_NJU data.

Overall, the PA distributions from these five datasets are quite similar, with a west-northwest to east-southeast oriented precipitation belt in the eastern part of the MTM and the ETM, and almost the same PA intensity levels. The lowest value of the PA (0.02 mm) occurs in the Naomao Lake Basin in the northern part of the EX region and the Turpan-Hami Basin in the southern part. At 0600 LST, the PA distributions derived from AWS and WRF_NJU data (Figure 3f,g) show that the basins are filled with precipitation, especially in the Naomao Lake Basin, and the spatial range of 0.1 mm PA in the mountainous areas expands in the observations. The PA derived from ERA5-Land data increases in both the mountainous and basin areas, with an expanded spatial coverage, which is generally consistent with the PA distribution derived from WRF_NJU data (Figure 3h). The PA distribution derived from CMORPH data (Figure 3i) shows that stronger precipitation is distributed in the eastern MTM and the ETM, and the precipitation belt distribution is inconsistent with that derived from WRF_NJU data. The PA values derived from MSWEP data show no distinct precipitation features in the mountainous areas, but the precipitation over the basins still increases, which is generally consistent with the PA distribution derived from WRF_NJU data (Figure 3j). These products show similar spatial patterns of precipitation features, but there are some differences in PA values.

At 1200 LST, the daily peak of PA occurs in the eastern MTM and the ETM, while precipitation in the basins is significantly weakened. In all five datasets (Figure 3k–o), the PA distribution in the mountainous areas is quite similar. Analysis of the PA distribution derived from WRF_NJU data reveals that the precipitation on plains < 2000 m ASL (above sea level) is consistent with AWS data, while the PA in the mountainous areas reflected by AWS data is weaker than that described by WRF_NJU data. This may be because the interpolation distribution pattern of PA in high-altitude terrain areas revealed by AWS data cannot accurately represent the true PA values of the relatively sparsely distributed AWSs. In comparison, the PA values in the mountainous areas depicted by the ERA5-Land data are relatively stronger than those derived from data of WRF_NJU, which is in accord with the previous PA evaluations [59]. The PA derived from CMORPH data shows no significant improvement, with stronger precipitation in the mountainous areas than in the basins.

The PA distribution derived from MSWEP data (Figure 3) shows that stronger precipitation with a PA of 0.3 mm is distributed in the mountainous areas, which is basically consistent with WRF_NJU data, although the overall PA values are weaker, but the distribution trend of the precipitation belt is consistent. At 1800 LST in the nightfall, the PA in the eastern MTM remains strong according to data of AWSs, while the PA in the ETM weakens, and precipitation in the basins also decreases, which is generally consistent with the PA distributions derived from WRF_NJU, CMORPH, and MSWEP data. However, the overall PA values derived from CMORPH and MSWEP data are weaker [59], while the PA derived from ERA5-Land data is still overestimated.

The PA values in the mountainous areas are overestimated by about 0.05 mm. Under the influence of complex underlying surfaces, reanalysis data show a very limited ability to represent the fine-scale precipitation changes systematically. This is consistent with the hourly precipitation feature evaluations by Chen Et Al. [60]. Most of the precipitation in the mountainous areas weakens and disappears in the following 6 h, while precipitation in the basins increases. It can be concluded based on the above analysis that the PA distribution shows a consistent trend across the five datasets. The WRF_NJU data successfully captured the spatial distribution of the PA, and related DVCP in the EX region. Similarly, some previous studies [59,61] investigated the DVCP in the Ili River Valley region based on the same data from WRF_NJU and also stated that the overall precipitation characteristics of data from WRF_NJU are basically consistent with observations and reanalysis data.

In addition, to quantitatively evaluate the performance of the different datasets (WRF_NJU, ERA5, CMORPH, MSWEP) in the EX region, the scatter plots and relevant basic statistical metrics between the average daily precipitation observed by automatic weather stations (AWS) and these datasets are analyzed below to assess their correspondence. As shown in Figure 4, the WRF_NJU data exhibit the highest correlation, with relatively good results of Pearson’s correlation coefficient and coefficient of determination (r = 0.78, R² = 0.61), along with a root mean square error (RMSE) of 0.36 mm. MSWEP and ERA5 show slightly lower correspondence: r = 0.69 and 0.65, and R² = 0.48 and 0.42, respectively, with an RMSE values of 0.34 mm and 1.03 mm. In contrast, CMORPH demonstrates the weakest agreement (r = 0.09, R² = 0.01) and a root mean square error (RMSE) of 0.36 mm. For the sake of clearer and more intuitive comparison, these statistical metrics are summarized in

Table 2. Pearson’s correlation coefficient (r), R², root mean square error (RMSE), obtained by linear regression calculation between average daily precipitation observed by automatic weather stations (AWS) and four other datasets (WRF_NJU, ERA5, CMORPH, MSWEP, respectively) in the EX region during June–August of 2015–2019.

	r	R2	RMSE
WRF_NJU	0.78	0.61	0.36
ERA5	0.65	0.42	1.03
CMORPH	0.09	0.01	0.36
MSWEP	0.69	0.48	0.34

below.

In addition, another issue warrants detailed explanation: the performance of numerical models in simulating precipitation is generally lower than that for other meteorological variables, such as temperature. Numerous previous studies [62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78] have shown that, because the mechanisms governing precipitation—particularly those related to terrain-induced convective cloud formation—are highly complex, numerical simulations often exhibit timing and location biases in precipitation. Therefore, although the values of r and R² for the WRF_NJU data are not particularly high, they nonetheless represent reasonably good performance by meteorological standards. Given that WRF_NJU demonstrates the best overall performance among the available datasets and offers the highest spatial resolution (4 km), it is selected for further analysis of DVCP in the EX region in the following.

4.2. Spatial Distribution of DVCP

The spatial distribution of average PA in the EX region (Figure 5) shows that the PA is generally higher in mountainous areas than in plains. From 0000 to 0600 LST, precipitation in the Turpan-Hami Basin and Naomao Lake Basin gradually increases, from 0.025 mm to 0.05 mm. Stronger rainfall bands appear in the eastern MTM and the ETM, with no significant changes in PA. From 0900 to 1200 LST, both the extent and intensity of the precipitation bands in the mountainous areas increase. The peak precipitation in the ETM occurs around 1200 LST, reaching 0.4 mm. Precipitation in the eastern part of the MTM also increases gradually, reaching 0.6 mm. The distribution of the highest precipitation aligns with the orientation of the mountains, while precipitation in the basins gradually weakens. At 1500 LST, the intensity and extent of precipitation in the eastern part of the MTM increase, while precipitation in the ETM begins to weaken significantly. Precipitation in the basins remains weak. From 1800 LST to 0000 LST, precipitation in both the mountainous areas and the basins decreases markedly.

The spatial distribution of the PA (Figure 6) shows that the pattern of PF has a similar DVCP to that of the PA. At 0000 LST, higher PF is observed in the eastern part of the MTM and the western part of the ETM. The Turpan Basin has a higher precipitation frequency (10–15%) than the Hami Basin (5–12.5%), while the Naomao Lake Basin has relatively lower PF. From 0300 to 0600 LST, PF in the eastern MTM gradually decreases, from 10% to around 20%. In the ETM, PF increases gradually from 10% to 20%, and its spatial extent expands. The Turpan-Hami Basin and the Naomao Lake Basin also see an increase in PF, with the area of 12.5% PF expansion. From 0900 to 1500 LST, PF in the eastern part of the MTM and the ETM increases gradually, reaching a maximum of 40%. In contrast, precipitation in the basins weakens, with the minimum frequency approaching 0%. From 1800 to 2100 LST, PF in the mountainous areas gradually decreases, while in the basins, PF increases gradually.

As depicted in the spatial distribution of the PI (Figure 7), at 0000 LST (Figure 7a), the PI in the eastern MTM and along the ETM is stronger than that in the central and northern parts of the EX region. The PI ranges from 0.5 to 1 mm·h⁻¹ in Yiwu County, Balikun Autonomous County, the northern part of Yizhou District, the northern part of Gaochang District, and the northern border of Shanshan County. In the remaining basins, the PI is between 0.05 and 0.5 mm·h⁻¹.

At 0300 LST, the PI in the central part of the EX region increases, from 0.1 to 0.2 mm·h⁻¹ to 0.2–0.5 mm·h⁻¹. At 0600 LST (Figure 7c), the PI in the northeastern part of Yiwu County and the western part of Balikun County increases, from 0.4 mm·h⁻¹ to 0.5–0.75 mm·h⁻¹. Meanwhile, the area with PI above 0.5 mm·h⁻¹ in the higher-altitude mountainous regions decreases. At 0900 LST (Figure 7d), the PI in the central and northern parts of the EX region gradually weakens. At 1200 LST, PI increases in the eastern MTM and the ETM, reaching 1.5 mm·h⁻¹. In contrast, PI in the central and northern basins of the EX region weakens. At 1500 LST, a strong precipitation center appears in the northwestern part of Balikun County, with a PI of 1.5 mm·h⁻¹. PI in the remaining mountainous areas weakens, while the basins in the central and northern parts of the EX region still have relatively weak PI. From 1800 to 2100 LST, PI in the eastern MTM and the ETM gradually weakens, while PI in the basin areas increases.

To further investigate the DVC of the peak precipitation times in the EX region, we analyzed the timing and spatial distribution of summer precipitation peaks (as shown in Figure 8). Areas near the western edge of the EX region (Toksun County) exhibit a precipitation peak from afternoon to nightfall (~1400–1900). In contrast, a “Y”-shaped distribution of precipitation peaks occurs around midnight over the central and southern parts of Gaochang District and Shanshan County. Additionally, the northern parts of Gaochang District, Shanshan County, and Yizhou District, together with the western part of the Barkol Kazakh Autonomous County, and the northern slope of the ETM, show a “Y”-shaped distribution of precipitation peaks around noon to afternoon (~1300–1600 LST). Meanwhile, the surrounding areas, including the central part of Barkol Kazakh Autonomous County, Yiwu County, and Yizhou District, display a “U”-shaped distribution of precipitation peaks in the early morning through late morning (~0300–1000 LST).

In the mountainous areas, precipitation is more frequent during noon and the afternoon, while in the basin areas, precipitation is more concentrated in the early morning and morning hours. There is a gradual color transition from the western mountainous areas to the central basin areas and then to the central mountainous areas, indicating a temporal shift in the appearance of peak daily precipitation. This suggests that there is a propagation phenomenon of peak precipitation in the diurnal variation across the region.

To better investigate the diurnal differences and interactions of precipitation characteristics, we further analyze the diurnal trend (Figure 9a) and linear regression (Figure 9b,c) of the regional average precipitation characteristics over the entire EX region from June to August during 2015–2019. Figure 9a shows that the peak PA occurs at 0600 LST, with a smaller secondary peak between 1300 and 1400 LST, and a trough between 1900 and 2000 LST. The peak in PF is at 0600 LST, while the trough is at 1800 LST. The DVC of PF and amount are quite similar, both showing lower values in the afternoon to nighttime compared to the early morning to noon period. The high value of PI is at 0600 LST, and the low value is at 2000 LST.

The scatter plots show that the data points are relatively evenly distributed around the regression lines, indicating a linear relationship. The coefficient of determination (R²) between PF and amount is 0.73 (Figure 9b), suggesting a good fit of the regression model. The R² between PI and amount is 0.69 (Figure 9c), indicating that compared to PI, the PF has a stronger association with the PA.

4.3. Coefficient of Variation in PF

The distribution of the CV for PF (Figure 10a) shows that higher CV values are observed in the central and northern basin areas of the study region. This indicates significant diurnal variability in PF in these areas, which are more prone to hazardous weather conditions such as droughts. Regarding the CV of PI (Figure 10b), higher CV values are found in the eastern part of the MTM and the ETM. These areas experience greater diurnal differences in PI and are more susceptible to hazardous weather events such as heavy rainfall.

4.4. DVCP at Different Elevations

Precipitation characteristics in Xinjiang are closely related to topography. We analyzed the variation in summer precipitation in the ETM with elevation. The relationship between PA and elevation (Figure 11a) indicates that the maximum values of PA, PF, and PI all increase with elevation, peaking in high-altitude areas. Among the six different elevation ranges, the low-elevation area (0–500 m) has a precipitation peak at 0600 LST, with a peak value of 0.02 mm. The trough occurs at 1300 LST, with a value of 0.08 mm. For the 500–1000 m elevation range, the precipitation peak is also at 0600 LST, with a value of 0.025 mm. The troughs occur at 1300 LST and 1800–1900 LST, with values of 0.11 mm. In the 1000–1500 m elevation range, the precipitation peak is at 0600 LST, with a value of 0.029 mm, and the trough is at 1900 LST, with a value of 0.014 mm. For the 1500–2000 m and 2000–2500 m elevation ranges, the precipitation peaks both occur at 1300 LST, with values of 0.069 mm and 0.11 mm, respectively. The troughs occur at 2000 LST, with values of 0.027 mm and 0.049 mm, respectively. In the high-altitude areas (2500–3000 m and 3000–4878 m), the precipitation peaks both occur at 1300 LST, with values of 0.154 mm and 0.194 mm, respectively. The troughs occur at 2200 LST, with values of 0.049 mm and 0.064 mm, respectively.

The diurnal variation in PF with different elevations in the EX region is shown in Figure 11b. The PF increases with elevation across seven different elevation ranges, with the highest values occurring in high-altitude areas and the lowest values in low-altitude areas. At 0600 LST, precipitation is more frequent at elevations below 1500 m (with a frequency of about 10%), while the least frequent precipitation occurs in the afternoon between 1400 and 1600 LST (with a frequency of about less than 7%). For the 1500–2000 m elevation range, the peak PF occurs at 1400 LST, with a value of 13.34%, and the trough occurs at 1900 LST, with a value of 8.24%. For the 2000–2500 m, 2500–3000 m, and 3000–4878 m elevation ranges, the peaks in PF all occur at 1300 LST, with values of 19.28%, 24.19%, and 27.64%, respectively. The troughs occur at 2000 LST, with values of 9.31%, 10.51%, and 12.37%, respectively. Overall, PF is lower in low-altitude areas than in high-altitude areas.

There are some differences in the DVC of the PI at different elevations (as shown in Figure 11c). In the 0–500 m elevation range, the peak of PI occurs at 0800 LST, with a maximum value of 0.20 mm·h⁻¹. The low-value period occurs at 1300 LST, with a value of 0.14 mm·h⁻¹. For the 500–1000 m elevation range, the high-value moment for PI is at 0600 LST, with a value of 0.24 mm·h⁻¹, and the low-value moment is at 1300 LST, with a value of 0.17 mm·h⁻¹. In the 1000–1500 m elevation range, the high-value moment for PI is at 0600 LST, with a value of 0.28 mm·h⁻¹, and the low-value moment is at 2000 LST, with a value of 0.19 mm·h⁻¹. At higher elevations, the hourly differences in PI are larger, and the peak characteristics of the PI curve are not obvious. For example, in the 1500–2000 m elevation range, higher PI occur at 0100–0200 LST and 1400 LST, with values of 0.42 mm·h⁻¹, 0.43 mm·h⁻¹, and 0.42 mm·h⁻¹, respectively.

The minimum PI occurs at 1800 LST, with a value of 0.33 mm·h⁻¹. In the 2000–2500 m elevation range, higher PI occurs at 0200 LST, 1200 LST, and 2100 LST, with values of 0.51 mm·h⁻¹, 0.54 mm·h⁻¹, and 0.43 mm·h⁻¹, respectively. The minimum PI occurs at 1900 LST, with a value of 0.39 mm·h⁻¹. In the 2500–3000 m elevation range, higher PI occurs at 0200 LST, 1100 LST, and 2100 LST, with values of 0.59 mm·h⁻¹, 0.66 mm·h⁻¹, and 0.52 mm·h⁻¹, respectively. The minimum PI occurs at 1800 LST, with a value of 0.45 mm·h⁻¹. In the high-altitude area (3000–4878 m), the peak of PI occurs at 1100 LST, with a value of 0.79 mm·h⁻¹. Secondary peaks occur at 0200 LST and 2100 LST, with values of 0.68 mm·h⁻¹ and 0.55 mm·h⁻¹, respectively. The minimum PI occurs at 1800 LST, with a value of 0.47 mm·h⁻¹.

In order to further investigate the DVCP in regions at different elevations, we conducted an analysis based on cross-sections along different elevation profiles (Figure 12). For example, the cross-section line segment c1d1 is taken in the western EX region (Figure 12a). Along this section, the PA decreases gradually from the mountainous areas to the basin. In the mountains, the maximum precipitation is 0.18 mm. In the basin areas, the PA is less than 0.05 mm. The peak times in the mountainous areas all occur at 1200 LST. The precipitation peak in the foothill areas to the basin occurs around 2100–0000 LST.

As shown in Figure 12b, the cross-section c2d2 still shows a pattern of higher values in the middle and lower values at both ends. At the northern end of the section, which is in the Balikun County basin, the precipitation peak occurs between 0300 and 0600 LST. In the mountainous areas above an elevation of about 3100 m, the precipitation peak occurs at 1500 LST. In the mountainous areas at an elevation of about 2700 m, the precipitation peak occurs at 1200 LST. At the southern part of the section, the precipitation peak in the basin and mountain-edge areas occurs between 0600 and 0900 LST, with a peak value around 0.04 mm.

Along the line segment c3d3 in the eastern portion of the EX region (Figure 12c), the precipitation peaks in the basin areas at both ends do not exceed 0.06 mm. At the northern end (the area at c3), the peak occurs between 0300 and 0900 LST. In the northeastern basin and adjacent areas, the peak of precipitation occurs between 0600 and 0900 LST. In the terminal areas of the section, which are at an elevation of 2200–2400 m, the precipitation peak occurs at 1200 LST.

4.5. DVCP over Different Land Use Types

Significant differences exist in DVCP over different land use types at uniform elevations. Hence, the DVCP over these different surface types are also investigated in the following. As shown in Figure 13, there are substantial differences in precipitation among the six underlying surface types, in the following order: forest > grassland > water body > cropland > barren > impervious surface. These differences can be primarily attributed to contrasting land–atmosphere interactions, which are modulated by surface properties that influence the surface energy balance and moisture flux. Specifically, forested areas and grasslands exhibit lower albedo than barren land and impervious surfaces. This leads to greater absorption of net radiation, which is partitioned more into latent heat flux (evapotranspiration) rather than sensible heat flux. The intense evapotranspiration not only enhances local moisture availability but also promotes the development of convective boundary layers, thereby increasing the potential for convective precipitation. In contrast, barren and impervious surfaces have higher albedo and lower moisture availability. The absorbed energy is primarily dissipated as sensible heat flux, creating a hot and dry boundary layer that suppresses rather than promotes convective activity. Although water bodies have a high evaporation capability similar to forests, their limited spatial extent in the EX region restricts their overall impact on mesoscale moisture convergence. This is likely why precipitation amounts over water bodies are slightly lower than those over forests and grasslands, despite their high evaporation rates.

Since the land use and land cover (LULC) in the Turpan-Hami region are dominated by deserts and bare land, urban heat islands are mainly concentrated in these arid and rugged land areas, forming a distinct desert heat island effect. In contrast, urban built-up areas generally have lower temperatures than the surrounding desert regions, creating a pronounced urban cool island effect. The summer precipitation in the EX region varies diurnally with different land surface types (Figure 13a). At noon (1200 LST), the forest has the highest PA. This is because the transpiration from trees and the moisture accumulated by vegetation in certain periods may be released at this time, promoting local convection and precipitation. In contrast, at 2000 LST in the evening, precipitation is relatively rare (with a maximum precipitation rate of 0.15 mm·h⁻¹). The second highest PA is over grasslands, with a peak at 1300 LST (maximum precipitation rate of 0.10 mm·h⁻¹). The peak precipitation over water bodies occurs at 1400 LST (maximum precipitation rate of 0.10 mm·h⁻¹), with the lowest value at night (2000 LST). The peak precipitation over croplands occurs at 0700 LST, with a peak value of about 0.03 mm·h⁻¹. The peak precipitation over barren areas occurs at 0600 LST, with a peak value of about 0.03 mm·h⁻¹. The peak precipitation over impervious surfaces occurs at 0400 LST, with a maximum precipitation rate of 0.021 mm·h⁻¹. Another peak occurs between 1700 and 1900 LST, with a precipitation rate of 0.02 mm·h⁻¹. In the Turpan–Hami region, the LULC are dominated by desert and barren area. The Urban heat island effect is primarily concentrated over these arid and topographically rugged land surfaces, giving rise to a pronounced “desert heat island” effect. In contrast, urban built-up areas in this region typically exhibit lower temperatures than the surrounding desert, resulting in a distinct “urban cool island” effect.

Similarly, differences in PF were compared (as shown in Figure 13b). Forests, grasslands, and water bodies have higher precipitation frequencies than other categories, such as croplands, barren areas, and impervious surfaces, and their peak times occur later. The peak PF over the grasslands, forests, and water bodies occurs during 1300–1400 LST, while the peak over croplands, barren areas, and impervious surfaces occurs between 0000 and 0600 LST.

The DVC of PI (PI) also differs among different land surface types (as shown in Figure 13c). The largest differences in PF (PF) over forests occur at 0200 and 1100 LST. Similarly, grasslands show distinctly high values during this period, but the high values of PF over grasslands are significantly weaker than those over forests. The highest PF over water bodies occurs around 1300 LST. Impervious areas, the croplands, and unused lands have the weakest PI values, and their PF peaks occur early, all between 0000 and 0600 LST.

As shown in Figure 14a, in the cross-section a1b1, the desert area of the Kumtag Desert has a precipitation peak at 0600 LST (90°E, 42.8°N) with a maximum PA of 0.02 mm. The desert area in the northeastern part of Yizhou District, Hami (91.5°E, 42.81°N) has a precipitation peak at 0300 LST with a maximum PA of 0.03 mm. The urban area of Hami City (93.5°E, 42.82°N) experiences a precipitation peak between 0300 and 0600 LST.

In the cross-section a2b2, the urban area of Xinxiang City has a precipitation peak at 1500 LST. The cropland in the southern part of Hami Mountain has a precipitation peak at 0600 LST. In the cross-section a3b3 (Figure 14c), the small urban area in Balikun County at the northwest end has a precipitation peak of 0.05 mm at 0600 LST. The barren at the terminal end of the cross-section has a precipitation peak around 0600 LST, and the cropland at the far left end also has a precipitation peak around 0600 LST. Overall, grasslands, forests, and water bodies have peak PAs (PA) during noon, while impervious areas, croplands, and barren areas reach their PA peaks during the early morning to morning period.

5. Discussion

5.1. Scope and Data Rationale

This study investigates the DVCP associated with diverse underlying surfaces in EX region—one of the most arid regions in China—during summer (June–August) from 2015 to 2019. Hourly, 4 km simulation data are obtained from the real-time forecasting system of Nanjing University (WRF_NJU). Evaluation against automatic weather station (AWS) observations indicates that WRF_NJU outperforms other three datasets (ERA5, MSWEP, CMORPH), confirming its reliability and accuracy for regional precipitation analysis.

The EX region’s complex topography, harsh arid environment, and sparse population result in extremely limited coverage of conventional and AWS observations. Consequently, the use of high-resolution WRF_NJU simulations provides a robust and spatially continuous depiction of DVCP in this data-scarce area. The model successfully reproduces the spatio-temporal evolution and diurnal peak timing of precipitation amount (PA), frequency (PF), and intensity (PI), thereby deepening our understanding of the DVCP over heterogeneous land surfaces in arid regions.

This study focuses on the diurnal cycle of summer precipitation—a choice guided by the regional precipitation characteristics, natural constraints (e.g., moisture transport, thermodynamic conditions, and precipitation type), and societal needs (e.g., water resource management and disaster prevention/mitigation) in the EX region. This emphasis does not imply that other seasons are unimportant; rather, in extremely arid regions like EX, summer precipitation holds primary scientific and socioeconomic importance because of its representative physical mechanisms and considerable societal impacts. In future work, we will investigate snow accumulation and melt processes in the EX region, as well as their roles in glacier mass balance and spring snowmelt-induced flooding, thereby enriching our understanding of seasonal precipitation characteristics in this region.

5.2. Key Characteristics and Regional Implications

The results reveal that PA generally peaks between morning and noon across EX—distinct from the afternoon maxima observed in many inland regions. This atypical pattern reflects the influence of complex terrain and mesoscale circulations. Clear spatial heterogeneity is found: precipitation in mountainous areas peaks around 12 LST, while basin regions exhibit earlier peaks near 06 LST. PI shows delayed peaks in some subregions, confirming substantial spatial variability. A pronounced desert–oasis contrast and significant coefficient of variation (CV) differences between mountainous and basin areas further highlight the impact of surface heterogeneity. Urban influences are minimal, except near Urumqi. These findings hold practical implications for flood and drought mitigation, irrigation planning, and hydrological modeling. Understanding the regional DVCP can improve flash-flood early warnings, optimize water resource allocation, and refine model representations of runoff, soil moisture, and groundwater recharge.

5.3. Cross-Sectional Evidence for Topographic and Surface-Type Modulation

The elevation and land-type cross-sectional analyses provide clear evidence that both topography and surface heterogeneity exert strong controls on the DVCP in EX. Along elevation transects (e.g., c1d1 and c2d2), precipitation in high-elevation mountainous areas consistently peaks in the early afternoon (1200–1500 LST), confirming the dominant role of topographically triggered convection. In contrast, basin and foothill regions show distinct nocturnal-to-morning peaks (2100–0900 LST), reflecting the delayed effects of mountain-generated cold-air outflows and the nighttime intensification of low-level jets.

Beyond topographic influences, surface-type cross-sections reveal systematic diurnal differences among land-cover categories. Natural surfaces such as grasslands, forests, and water bodies tend to peak near noon, consistent with locally driven thermodynamic convection. Conversely, impervious, cropland, and barren surfaces—including the Kumtag Desert—exhibit earlier peaks between 0300 and 0900 LST, associated with enhanced nocturnal radiative cooling and shallow boundary-layer processes.

These contrasts highlight the combined effects of surface energy partitioning, thermal inertia, and mesoscale circulations in shaping the spatial heterogeneity of precipitation timing. Together, the cross-sectional results provide quantitative and mechanistic support for the broader conclusion that topographic forcing and land-surface properties jointly regulate the DVCP across the EX region.

5.4. Possible Physical Interpretation and Mechanisms

The widespread morning-to-noon PA maximum is interpreted as primarily resulting from topographic forcing. Mesoscale circulations—such as mountain–valley breezes and their interactions with large-scale flow—likely trigger nocturnal and early-morning convection, compensating for the weak daytime instability typical of arid environments. This mechanism coherently explains the prevalence of morning peaks and regional differences in PI timing.

Large-scale systems modulate, but do not dominate, the diurnal cycle. Mid-latitude trough–ridge systems in the westerlies influence precipitation potential on multi-day timescales, while nocturnal low-level jets strengthen nighttime moisture transport, favoring early-morning rainfall. Central Asian low-vortex systems enhance afternoon-to-nighttime convection through topographic thermal forcing, while a westward-extended Western Pacific Subtropical High (WPSH) transports moist air from the Bay of Bengal and South China Sea into Xinjiang. When this moisture reaches the eastern Xinjiang mountains in the early morning, daytime surface heating efficiently triggers convection.

Future studies should classify precipitation events by synoptic regime and conduct composite DVCP analyses for each type to isolate topographic signals from broader circulation effects.

5.5. Model Evaluation and Uncertainties

Averaging five consecutive summers effectively reduces day-to-day variability, yielding a climatologically representative DVCP pattern—the first multi-year numerical benchmark for EX. Spin-up treatment was applied by discarding forecast hours 1–12, minimizing initialization shocks in key boundary-layer and hydrometeor fields that influence DVCP amplitude and phase. Using 12–36 h forecasts provides a balance between physical equilibrium and synoptic realism; future work will test alternative lead-time windows (e.g., 6–30 h, 18–42 h) to evaluate sensitivity.

Uncertainties remain due to initial and boundary conditions (IC/BC) and parameterizations. DVCP sensitivity to microphysics, cumulus, and land-surface parameterizations is significant in arid, high-elevation regions with strong land–atmosphere coupling. Although WRF_NJU simulations show good agreement with AWS, ERA5, and MSWEP datasets, systematic biases may persist, particularly in representing fine-scale terrain–convection interactions. Future ensemble experiments perturbing IC/BC and testing alternative physics suites will help quantify model robustness.

5.6. Role of Cloud Microphysics

Accurate representation of cloud microphysical processes is essential for simulating both precipitation intensity and frequency. Previous research demonstrated that one-moment versus two-moment microphysics schemes markedly affect the evolution of squall-line stratiform precipitation [79]. Guo Et Al. [80] found that implementing a two-moment scheme with prognostic precipitation improves simulations of precipitation intensity and diurnal variability in regions dominated by deep convection. Observational studies [81] further indicate that precipitation efficiency and raindrop size distributions vary among weather systems, modulated by microphysical characteristics. These results emphasize that realistic microphysics representation is critical for improving predictive skill in both extreme and climatological precipitation.

Moreover, new variable-resolution modeling frameworks [82] provide opportunities to resolve convective life cycles and their interactions with microphysics more effectively, advancing understanding of precipitation scaling across temporal and spatial scales.

5.7. Limitations and Future Directions

This study characterizes DVCP contrasts across land-surface types but does not fully reveal how specific surfaces (e.g., deserts, croplands, urban areas, and water bodies) modulate the cycle. Understanding these mechanisms remains a key challenge in arid zones with distinctive surface-energy budgets. Future work will quantify the influence of individual surface types and clarify their physical roles in shaping DVCP.

Additionally, the linear regression analyses between PA and PF/PI provide an initial depiction of their relationships but may overlook nonlinear or conditional dependencies. Sensitivity tests using percentile-based thresholds (e.g., 90th or 95th percentiles) can verify whether PF–PA and PI–PA relationships persist under high-precipitation conditions. Such percentile-based approaches have been widely used to distinguish general from extreme-event behavior [83]. Moreover, cloud microphysical processes—such as variations in condensation nuclei concentration, liquid water content, and droplet coalescence efficiency—may affect precipitation efficiency and confound these relationships, especially in arid regions where mesoscale variability is strong. Future research combining observational data with cloud-resolving model simulations will help capture these nonlinear responses and enhance understanding of precipitation variability in EX.

6. Conclusions

This work evaluated hourly and 4 km horizontal resolution WRF-NJU model simulation data. Based on high-resolution simulation data, the DVCP in the EX were revealed, and the following main conclusions were drawn:

(1)

The data from WRF_NJU accurately captured the DVCP in the EX region. The PA from the other four datasets is generally similar to the WRF-NJU data. However, because of the sparse distribution of meteorological stations in the EX region, accurately reflecting the precipitation over complex terrain (mountainous areas) is very difficult. At the same time, precipitation derived from the ERA5 data shows overestimation and hardly reveals the fine-scale variations in the precipitation. CMORPH data has weaker assessment capabilities for mountainous precipitation, while MESWEP data has assessment capabilities that are basically consistent with WRF_NJU data.

(2)

The DVCP in the EX region show significant spatiotemporal differences. Precipitation peaks mainly occur from early morning to noon, with a gradual decrease in the afternoon. Moreover, the PA in the basin areas is significantly lower than that in the mountainous areas. The temporal-spatial distribution characteristics of PA, PF, and PI are closely related to elevation.

(3)

There are clear differences in the DVCP at different elevations. Below an elevation of 1500 m, and the peak of PA occurs at about 0600 LST, while in the higher altitude (mountainous) areas > 1500 m, the PA reaches its peak at 1300 LST in the afternoon hours, with the PF indicating a similar variation trend. The peak of PI in the mountains occurs at 0000 LST, 1400 LST, and 2100 LST, while in the basins, the PI peaks around 0600–0800 LST.

(4)

There are obvious differences in the DVCP between mountainous and basin areas. The mountains are more influenced by topography than by land surface type. Additionally, the strength of the desert-oasis effect is mainly influenced by different land surface types. The central and northern basins of the EX region show higher CV values for PF, indicating obvious differences in the DVCP in these areas and suggesting a higher likelihood of meteorological disasters.

(5)

The timing of precipitation peaks varies among different underlying surfaces. Grasslands, forests, and water bodies have precipitation peaks between 1200 LST and 1400 LST, while impervious areas, the croplands, and barren areas have peaks at 0600 LST. The peaks of PF for grasslands, forests, and water bodies occur at 0700 LST. However, the PF peaks for impervious surfaces, barren areas, and croplands occur between 1700 and 1900 LST. The peaks of PI for barren areas and forests occur at 1100 LST, while the peaks of PI for impervious areas, water bodies, croplands, and grasslands occur at 1700 LST.