Sensitivity of temperature and precipitation characteristics to land use classification over the Taklimakan Desert and surrounding area

The Taklimakan Desert (TD) is the largest desert in China, which has a great influence on the weather and climate in the surrounding oases. To further study the impact of the unique underlying surface and artificial green space of TD on the weather and climate of the desert and its surrounding oases, we selected the winter and summer months to carry out a continuous simulation experiment using a high-resolution numerical model. The changes of land use in TD (the land use category denoted as “barren or sparsely vegetated” was modified to “shrub land”) have direct impacts on the surface heat flux, which affect the stability of the lower atmosphere and water vapor conditions, thus affected precipitation. Both sensible heat flux and latent heat flux in the desert increased in winter and summer after changing the land use type, while in areas with unchanged land use around the desert have a slight weakening in the net radiation flux. The daily temperatures increased as a whole, especially in the hinterland and south of the desert, and the warming amplitude in winter and summer was basically same. In the Kunlun Mountains, the western desert, and the oasis area to the west of the desert, the lower atmosphere became warmer and wetter, which led to a significant increase in monthly cumulative precipitation in July, with an increase in some areas greater than 36 mm which is significant for a desert area.


Introduction
Land surface has complex properties, diverse forms, and uneven distribution (Shukla and Mintz 1982;Betts et al. 1996).Driven by the atmospheric circulation and solar radiation, there is an uninterrupted multi spatiotemporal interaction between the land surface and the atmosphere, including the exchange of energy, momentum, and matter and radiation transfer.The underlying surface has unique properties in surface albedo, emissivity, and roughness, which affect the transport of latent heat, sensible heat, and radiation from the land surface to the atmosphere (Kitoh et al. 1988; Sud and Molod 1988;Jin and Miller 2011), thus having an important impact on regional precipitation, atmospheric structure, and climate conditions (Xue andShukla 1993, 1996;Berg et al. 2013;Dyer and Rigby 2020).
Deserts have unique characteristics, such as extreme drought, high surface albedo, low heat capacity, sandy underlying surface, frequent dust weather, and are a strong sensible heat source.These characteristics not only play an important role in global and regional energy balance and climate change (Miller and Tegen 1998;Yang et al. 2011) but are also extremely sensitive to human activities and climate change.The ecological environment and human activities affect the earth-atmosphere energy balance, cause changes in the surface type and atmospheric composition, and change the balance of surface heat and radiation in desert areas, thus affecting regional weather and climate characteristics (Charney 1975;Sud and Smith 1985;Nicholson and Tucker 1998;Yang et al. 2005;Huo et al. 2016).The Taklimakan Desert is the largest desert in China, and more than 95% of the desert is covered by bare shifting sand.An accurate understanding of its land surface characteristics and the impact on surface radiation and energy is of great significance to understand regional and global climate change (Gong et al. 2003;Zhang et al. 2003;Zhao et al. 2003Zhao et al. , 2006;;Yang et al. 2020).In recent years, the construction of an ecological protection barrier along the Taklimakan Desert highway and the installation of an artificial green cover on a soil base in the hinterland of Taklimakan Desert have changed the nature of the regional underlying surface, the characteristic parameters of surface, and changed the near surface energy distribution pattern (Abbas et al. 2021).It has formed a local green microclimate which is different from the extreme climate of the desert (Zhou et al. 2017;Jin et al. 2017;Mao et al. 2017;Cao et al. 2020).Although the vegetation is sparse and mostly consists of sandy shrubs with salt tolerance, it also has an impact on the local climate.It can be seen that the influence of the whole desert on the climate of surrounding oases is important and significant.At present, research on the influence of Taklimakan Desert on the climate of surrounding oasis mainly involves the difference in thermal, temperature, and vapor and so on between desert and oasis and the changes in soil properties and energy characterization parameters of the local artificial green climate based on observation data (Su et al. 2010;Gui et al. 2010;Zhou et al. 2020).However, there is still a lack of simulation research on the effects of special underlying surface of the Taklimakan Desert on the climate of surrounding oasis by sensitivity experiments.
In order to further understand that the thermal effect of the Taklimakan desert has an important influence on the climate of surrounding oases, because of the special underlying surface, the objective of this paper was to further study the impact of the unique underlying surface of Taklimakan Desert on the weather and climate of the desert and its surrounding oases using a mesoscale high-resolution numerical weather modeling framework.Based on a continuous experiment in which the underlying surface of the desert is gradually changed to artificial green space, the analysis focused on the comparison of surface energy and regional weather and climate elements such as temperature and precipitation.

Data and methods
The WRF-ARW (Advanced Research Weather Research and Forecasting) v3.8.1 was used to simulate daily meteorological condition over the Taklimakan Desert using a 9-km spatial resolution (Fig. 1) with 41 vertical sigma levels in the seasonal-scale simulations (January 1 to 31 and July 1 to 31, 2017, representing winter month and summer month respectively).
In order to study the influence of the underlying surface and artificial green space of the Taklimakan Desert on the climate of surrounding oases, in 77-88 °E and 36.5-41°N areas, all grids where the land use category was denoted as "barren or sparsely vegetated" were modified to "shrub land" (Fig. 1b, replaced 19 with 8).This reflects the land use category of artificial green space and the transitional zone between desert and oasis (Jin  . 2017).From this point forward, the simulation done using the original land use definitions is referred to as the "control," while the simulation using the modified land use definitions is referred to as the "modified." The primary differences between the "barren or sparsely vegetated" and "shrub land" categories are the albedo, surface emissivity, leaf area index, and surface roughness (Niu et al. 2011) (Table 1).The decrease in albedo and the increase in surface emissivity with "shrub land" will modify the amount of thermal energy absorbed at the surface, leading to variations in the surface heat fluxes.The increase in surface roughness with "shrub land" will inflect near-surface wind speed by increasing friction, which can modify surface fluxes by decreasing the transport of heat and water from the surface into the lower atmosphere.Additionally, the increase in leaf area index with "shrub land" will act to surface temperature and soil humidity.

Comparison between model and observations
We compared the results of the control simulation with the observations for 2-m temperature using the mean absolute error (MAE) and the root mean square error (RMSE).The simulated 2-m temperature values showed a general warm bias compared with the associated observations, with an average MAE of 2.5 °C in January and 3.2 °C in July and a RMSE of 3.1 °C in January and 3.9 °C in July.The bias was relatively consistent across all stations in and surrounding the Taklimakan Desert, with a standard deviation of MAE and RMSE less than 1.0 (Table 2).
In this paper, soil temperature observations were divided into eight layers in the vertical direction: 5 cm, 10 cm, 15 cm, 20 cm, 40 cm, 80 cm, 160 cm, and 320 cm.Soil temperatures at 5 to 20 cm corresponded to shallow ground temperature, and soil temperatures at 40 cm to 320 cm correspond to deep ground temperature.The model's land surface scheme used is the Noah land surface model, with soil temperature and moisture in four layers at 10 cm, 30 cm, 60 cm, and 100 cm, that is, the effective depth of each layer was 5 cm, 25 cm, 80 cm, and 150 cm respectively.The simulated soil temperature values of the first layer (5 cm) and the third layer (70 cm) were compared with observed soil values at 5 cm and 80 cm, representing the shallow ground temperature and deep ground temperature, respectively.In the comparison, the simulated 5-cm soil temperature showed a general weak cold bias in January and a general warm bias in July, with an average MAE of 2.4 °C in January and 3.2 °C in July and a RMSE of 2.9 °C in January and 4.1 °C in July.Similarly, the soil temperature bias at 2 m was relatively consistent across all stations in and surrounding the Taklimakan Desert, with a standard deviation of MAE and RMSE less than 1.0 (Table 2).
In recent years, an artificial green space was established near the Tazhong station (ID: 51747) located in the hinterland of the Taklimakan Desert.The regional underlying surface has changed, but the land use category of this area was still denoted as "barren or sparsely vegetated" in the soil categories database used in the model, so the simulated values of 2-m temperature and soil temperature at the station were significantly different from the actual observations, especially in winter months (January).
To analyze the biases in 2-m temperature and soil temperature between the WRF simulations and the observations, the time series of 2-m temperature and 5-cm and 70-cm soil temperature are shown for Bachu (ID: 51716) and Cele (ID: 51826) which are respectively located at the western edge of the Taklimakan Desert and in the Taklimakan Desert (Fig. 2).The land use category of Bachu is "shrub land" and the land use category of Cele is "barren or sparsely vegetated" in WRF simulations.In the desert and surrounding oasis area, the simulated 2-m temperature was mainly positive deviation compared with the observations in January and July.The diurnal range simulated in winter was basically consistent with the observations; the maximum or minimum temperature was slightly higher.The diurnal range simulated in summer was smaller than the observations, the minimum temperature was generally higher than the observations, and the maximum temperature in the desert was significantly higher.In winter, for shallow soil temperature, the daily range in desert area was obvious, but the diurnal range of simulation results is smaller, and the minimum temperature was high as a whole, while the diurnal range in surrounding oasis area was basically consistent with the observations, and the minimum temperature was slightly higher.For the deep soil temperature, the simulated value in the desert area was lower than the observations.In summer, the diurnal range of shallow soil temperature in desert area and nearby oasis was basically consistent with the actual, and the maximum temperature was higher as a whole, and the simulated value of deep soil was consistent with the actual.

Influence on surface energy fluxes
We compared the average daily accumulated sensible heat fluxes (Fig. 3) and latent heat fluxes (Fig. 4) in January and July in the control run and after replacing land use type in the model.The sensible heat flux in the desert hinterland is the main contributor to net radiation flux, while the latent heat flux was usually weak.Due to the difference in solar altitude and overall energy, the net radiant flux in winter was significantly smaller than in summer as a whole.In January, the accumulated sensible heat flux of the desert and its surrounding areas was 0.5-1.0MJ m −2 , which is greater than 2.0 MJ m −2 in the southern margin of the desert (Fig. 3a1).
In July, the cumulative sensible heat flux of the desert and its surrounding areas was above 6.0 MJ m −2 on the whole and greater than 12.0 MJ m −2 in some areas of the desert hinterland (Fig. 3b1).The accumulated latent heat flux in the desert and its surrounding areas was about 0.5 MJ m −2 in January, which was larger in the Kunlun Mountains, at about 2.0 MJ m −2 (Fig. 4a1).The accumulated latent heat flux in the desert hinterland in July was similar to that in winter, while, on the western side of the desert, the accumulated latent heat flux was about 2.0 MJ m −2 , while at the edge of the desert and the transition zone, it was greater than 5.0 MJ m −2 , and in mountainous and high altitude areas greater again (Fig. 4b1).
Replacing "barren or sparsely vegetated" land use type in the model with "shrub land" led to slightly increased mean daily accumulated latent heat fluxes (< 0.5 MJ m −2 ) and sensible heat fluxes (< 0.2 MJ m −2 ) in January (Figs.3a2 and Fig. 4a2), mainly obvious in the southern and southeastern edges of the desert.In July, there was a noticeable increase in the mean daily accumulated sensible heat fluxes (Fig. 3b2), more than 2.0 MJ m −2 overall, and more than 3.0 MJ m −2 in the desert hinterland and southeast.The pattern for sensible heat flux was slightly different from the pattern for latent heat flux (Fig. 4b2).In the transition zone between desert and oases, differences in sensible heat fluxes were more obvious than that in the desert hinterland, with an increase greater than 1.5 MJ m −2 and an increase in the desert hinterland under 0.4 MJ m −2 .Replacing land use type in the desert led to a slight weakening in the net radiation flux in areas with unchanged land use around the desert, less than 0.2 MJ m −2 .
To better understand the impact of land use type replacement on net radiation flux, we performed a comparative analysis of the diurnal variations of sensible heat flux and latent heat flux in representative stations.These stations represented respectively: the area in the southeast and northwest of the desert where the land use category was changed from "barren or sparsely vegetated" to "shrub land," the oases in the northwest edge of the desert where the land use type was not replaced, the desert at the eastern edge where the land use type was not replaced, and the oases to the north of the desert.These stations' locations are seen in Fig. 3.The land use type of these stations in control run and modified run is shown in Table 3. Figures 5 and 6 show that the diurnal variation trends of sensible heat flux and latent heat flux in the artificial green space and desert area were consistent before and after land use change.The diurnal variations followed an inverted U-shaped quadratic curve, lowest in the morning and evening and largest at noon.When the land use type of the Taklimakan Desert changed, it will have influence the surface energy flux of surrounding oasis and desert areas to a certain extent.However, as the land use type of these stations did not changed in modified run, the amplitudes of variation in sensible heat flux and latent heat flux are small (represented by stations 51716 and 51777).Both of sensible heat flux and latent heat flux increased after replacing the land use type in January (Fig. 5) and July (Fig. 6), the increase being mainly visible around noon (represented by station 51826).In the afternoon when the net radiation flux is large, both latent heat flux and sensible heat flux were greater in July than in January, in both the control and the experiment.There was an obvious seasonal variation of sensible heat flux, with a maximum sensible heat flux in January of 30 W m −2 and a maximum sensible heat flux in July of 150 W m −2 .After replacing the underlying land use, there was no obvious difference in the increase in latent heat flux between January and July, with an increase in January about 3 W m −2 and 3-5 W m −2 in July.There was a more obvious difference in the increase in sensible heat flux between January and July, with the increase in January about 10-15 W m −2 and 20-40 W m −2 in July.However, compared with  the control run, the increased amplitude of sensible heat flux in the experiment was about 50% in January and about 30% in July.

Two-meter temperature above the surface
After replacing the land use type of the underlying surface by large-area shrubs, the 2-m temperature above the surface (referred to as 2-m temperature) in Taklimakan Desert generally increased in both winter (January) and summer (July) (Fig. 7).After replacing the underlying land use, there was no obvious difference in the increase of latent heat flux between January and July, but the increased amplitude of sensible heat flux in January was larger than in July (mentioned in Section 3.2), so the 2-m temperature increase in January (Fig. 7a2) was greater than that in July (Fig. 7b2).In January, most average temperatures in desert areas increased above 0.6 °C, with a maximum increase of above 1.0 °C in the hinterland and the eastern part of the desert.In July, the monthly average temperature in the desert area increased by more than 0.4 °C, with the highest increase in the hinterland of the desert of about 0.8 °C.In the area surrounding the Taklimakan Desert where the land use type of the underlying surface is grassland or dry land farmland, monthly average temperature increased slightly in January.In the same areas, the monthly average temperature decreased by of about 0.2 °C in July.In the hinterland of the Taklimakan Desert and its surrounding areas, six representative stations (mentioned in Section 3.2) were selected to analyze the diurnal variation of 2-m temperatures in January and July (Figs.8 and  9).The variations of daily average temperature before and after replacing the land use of the underlying surface were unchanged.After replacement, the daily temperature increased as a whole, especially in the hinterland and south of the desert, where the temperature increased by about 1.5 °C because of a significant increase in the net radiation flux.In the areas where the land use type was not replaced, the daily average temperature of the oasis in the north of the desert was unchanged, and the daily average temperature in the east of the desert increased by less than 0.5 °C.The warming amplitude in January and July remained basically the same, because replacing the land use type did not lead to significant differences in the seasonal variations of net radiation flux.After replacing the land use of the underlying surface, the daily temperature amplitude in winter decreased.For example, there was an obvious cooling on January 10, and the 24h temperature change value of the control experiment was greater than that of the sensitivity experiment.The temperature variability in summer was more obvious.For example, there was a weak cooling in mid-July, and the cooling rate and heating rate in the sensitivity experiment were greater than in the control experiment.

Precipitation
In this paper, the cumulative precipitation from 12:00 UTC (20:00 BJT) to 12:00 UTC the next day reported by the model from 00:00 UTC (08:00 BJT) every day was defined as the daily precipitation, and the monthly cumulative precipitation was obtained by accumulating the daily precipitations.There was less precipitation in winter (January) in Southern Xinjiang, the monthly cumulative precipitation of multiple stations was less than 3 mm, with individual monthly cumulative precipitation reaching more than 12 mm, while replacing the land use type of the underlying surface in the Taklimakan Desert area had no significant impact on precipitation (figures omitted).Under the influence of terrain and major impact systems (Zhang and Deng 1987), the highest precipitation areas were concentrated in the western mountains of Southern Xinjiang, the Tianshan Mountains, and the Kunlun Mountains in July, and the cumulative precipitation results in desert and surrounding areas were generally less (Fig. 10a1).
After replacing the land use type of the underlying surface, the boundary layer height of the Taklimakan Desert  3 increases as a whole (more than 100 m), which in the eastern part of the desert increases more than 200 m.Meanwhile, the boundary layer height in the Tianshan Mountains and surrounding oases in the north of the desert decreased slightly (Fig. 10b).In modified simulations, the average 2-m dew point temperature in the west of the desert and the oases to the west of the desert increased with the maximum increase of 0.5 °C, and the dew point temperature also increased in Kunlun mountains.However, in the eastern part of the desert and in the Tianshan Mountains, the dew point temperature decreased (Fig. 10c).Combined with the 2-m temperature changes (mentioned in Section 3.3.1),it can be seen that in the Kunlun Mountains, the western desert, and the oasis area to the west of the desert, the lower atmosphere became warmer and wetter under the influence of the change of land use type, which lead to a significant increase in precipitation, with the increase greater than 36 mm, which was a significant increase for a desert area.Meanwhile, in the Tianshan Mountains and the eastern part of the desert, the lower atmosphere became warmer and drier, which corresponded to the decrease of precipitation (Fig. 10a2).Hence, one can see that the changes of land use should have direct impacts on the surface heat flux, which would affect the stability of the lower atmosphere and water vapor conditions, thus affected precipitation.

Conclusions
As the largest desert in China and the second largest mobile desert in the world, the Taklimakan Desert has a great influence on the weather and climate in the surrounding oases.Planting an artificial green space in the hinterland of the Taklimakan Desert changed the nature of the regional underlying surface, changed the near surface energy distribution pattern, and formed a local green microclimate different from the extreme desert climate.To further study the impact of the unique underlying surface and artificial green space of Taklimakan Desert on the weather and climate of the desert and its surrounding oases, we selected the winter months (January 2017) and summer month (July 2017) to carry out a continuous simulation experiment using a high-resolution numerical model.In a model sensitivity simulation, the land use category denoted as "barren or sparsely vegetated" was modified to "shrub land," reflecting the land use category of the artificial green space installed in the transitional zone between desert and oasis.The main results of the comparative analysis are summarized as follows: (1) The net radiation flux was sensitive to changes in land use.Both sensible heat flux and latent heat flux in the desert increased in January and July after changing the land use type, the increase seen mainly around noon.The increase in net radiation fluxes, especially sensible heat fluxes, was low in January and higher in July.However, compared with the control run, the increase in amplitude of sensible heat flux was similar in January and July.Meanwhile, artificially greening some desert areas led to a slight weakening in the net radiation flux in areas with unchanged land use around the desert.
(2) After replacing the land use type of the underlying surface, average daily temperature variations remained the same and daily temperatures increased as a whole, especially in the hinterland and south of the desert because of a significant increase in the net radiation flux in the area.The temperature amplitude between January and July remained basically the same, because there were no significant differences between the seasonal variations in net radiation flux.The amplitude of daily temperature change was slow in January but faster in July.
(3) The changes of land use in desert had direct impacts on the surface heat flux, which affected the stability of the lower atmosphere and water vapor conditions, thus affected precipitation.In the Kunlun Mountains, the western desert, and the oasis area to the west of the desert, the lower atmosphere became warmer and wetter under the influence of the change of land use type, which led to a significant increase in monthly cumulative precipitation in July, with an increase in some areas greater than 36 mm which is significant for a desert area.
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http:// creat iveco mmons.org/ licen ses/ by/4.0/.

Fig. 1
Fig. 1 Land use category in the control simulation (a) and the modified simulation (b)

Fig. 3 Fig. 4
Fig. 3 Daily average accumulated upward sensible heat flux in control and associated change in heat flux relative to the modified (modified minus control) in January (a1, a2) and July (b1, b2)

Fig. 5
Fig. 5 Average diurnal variations of upward latent heat flux and upward sensible heat flux in the control and the modified simulation and difference between the control and modified simulation (modified minus control; bar) in January.The local time in study area is Beijing time (BJT)

Fig. 6 Fig. 8
Fig. 6 Average diurnal variations upward latent heat flux and upward sensible heat flux in the control and the modified simulation and difference between the control and modified simulation (modified minus control; bar) in July

Fig. 9
Fig. 9 Daily average 2-m temperature in the control and modified simulation, and difference between the control and modified simulation (modified minus control, bar) in July.Stations locations are shown in Fig. 3

Table 1
Characteristics of land use type based on USGS 24-category land use data (described in the VEGPARM.TBL file in the WRF run directory)

Table 3
Land use type of representative stations in control and modified simulations