Increased diurnal temperature range in global drylands in more recent decades

The diurnal temperature range (DTR) has generally decreased at global scale according to the IPCC reports and previous publications. Here, we examined the change in DTR within the past four decades and found that DTR decreased in wet areas, but it increased in dry areas. This is because the daily minimum air temperature (Tmin) increased at a slower rate than the daily maximum air temperature (Tmax) in the dry areas, while it increased more rapidly in the wet areas. The changes in cloud cover, water vapour and moisture flux were observed to have contributed to the observed change in DTR. The change in moisture flux largely contributed to this change through the indirect effect on the change in water vapour across the aridity gradient. The significant impact of moisture flux is linked to the differential change in moisture flux between the dry and wet zones.


| INTRODUCTION
The recent IPCC report states an increase in the average global temperature from the 19th century to the current century, particularly in the past four decades, with a temperature increase of 0.99 C within the past two decades (IPCC, 2021).The increase in global mean temperature is characterized by a greater increase in night-time temperature, with most of the warming occurring in the night period compared to the daytime (Easterling et al., 1997;Guan et al., 2022;Karl et al., 1993).This uneven warming has caused a change in the global diurnal temperature range (DTR) trend, which is therefore recognized as an essential indicator of climate change (Shen et al., 2017).The trend and variation of DTR have significant implications on both human health (Jang & Chun, 2021;Lei et al., 2020;Wang et al., 2020b) and agricultural productivity (Hernandez-Barrera et al., 2017;Lobell, 2007;Rahman et al., 2017;Sunoj et al., 2020).According to Cheng et al. (2014), mortality and mobility are linked to DTR, especially through the effect on cardiovascular and respiratory systems of the body.Hernandez-Barrera et al. (2017) showed that DTR has a major impact on wheat yield compared to the independent influence of maximum and minimum temperatures.Differential warming in the day and night-time have also been observed to alter water budget and specie vulnerability to heat-related conditions (Cox et al., 2020).This warming was shown to have differing effect on plant diversity as night-time warming reduced plant diversity while daytime warming increased plant diversity (Barton & Schmitz, 2018).The effect of day and night-time warming on root dynamics has also been reported to vary at the leaf to ecosystem level, indicating the significant impact of daytime warming on root lifespan (Bai et al., 2012).Varying effects of warming have also been revealed to occur with different biomes exhibiting different responses to the warming (Wen et al., 2018).
The observed decrease in global DTR trend has been linked to a couple of factors like changes in cloud cover, soil moisture, greenhouse gases as well as aerosols (Dai et al., 1999).At the regional level, while the observed decrease in DTR is closely associated with cloud cover and precipitation as key factors (Dike et al., 2019;He et al., 2015;Jhajharia & Singh, 2011;Lauritsen & Rogers, 2012;Xia, 2013), other factors have also been reported to influence the change in DTR, such as urbanization, farming practices (irrigation) and alteration of vegetation cover (Ren & Zhou, 2014;Wang et al., 2014).
In recent decades, satellite observations have revealed enhanced evapotranspiration over the global land areas which is linked to elevated warming (Pascolini-Campbell et al., 2021).The processes of land and atmosphere are tightly coupled so that changes in evapotranspiration, a crucial component of the moisture cycle, impact the planetary boundary layer with corresponding impact on precipitation changes and cloud formation (Santanello et al., 2018).With the increase in temperature in recent years, more moisture loss has occurred which could further induce more cloud formation and increase in atmospheric water vapour (Devaraju et al., 2018;Green et al., 2017).More cloud cover impacts solar insolation during the day, causing less warming and also exerts feedback effect for more warming during the night, which may thus sustain the declining DTR trend.More so is the greening of the earth observed over the globe, as reported in recent studies (Piao et al., 2020;Zhu et al., 2016).Greening of the globe also enhances transpiration which contributes to the moisture cycle and impacts precipitation (Piao et al., 2020;Zeng et al., 2018), particularly in humid areas.Drier areas are, however, constrained with respect to increased evapotranspiration due to limited moisture.Although greening has occurred in the drylands (Fensholt et al., 2012;Gonsamo et al., 2021), the moisture-limited nature of the drylands causes plants to regulate their stomata to prevent loss of water to the atmosphere with the resultant effect of reduced transpiration (Grossiord et al., 2020;Osakabe et al., 2014).Thus, due to the coupled process of the land and atmosphere, dry and wet areas may experience a differential DTR trend.This raises the question of if DTR has been affected by the recent changes in the enhanced moisture loss from the land surface.Therefore, in this study, we examine the change in DTR across the aridity zones in more recent decades as it relates to moisture changes.The Aridity Index (AI) is an extensively used indicator for global and regional aridity (Ghazanfari et al., 2013;Huang et al., 2016;Nastos et al., 2013).AI is a climate index suitable for assessing drought occurrence and changes in aridity trends, as well as the division of climate regimes (Nastos et al., 2013).It is a useful approach for identifying global and regional drylands.It is the ratio that defines the available rainfall in relation to the evaporative demand of the atmosphere (Feng & Fu, 2013;Ghazanfari et al., 2013).For this study, the AI dataset (Zomer et al., 2007(Zomer et al., , 2008) ) was obtained from the Consultative Group for International Agriculture Research (CGIAR) Consortium for Spatial Information.The aridity index was used to classify the global land into different zones accordingly: dryland (AI < 0.65) and non-dryland (AI > 0.65) (Figure 1).
The temperature dataset used in this study which includes minimum air temperature (T min ), maximum air temperature (T max ) and diurnal temperature range (DTR), were obtained from two sources.First is the Climate Research Unit (CRU TS v. 4.05) temperature data which is an extensively used gridded climate data for both global and regional climate studies (Harris et al., 2020).The gridded data product is interpolated from a wide network of ground observation stations which are homogenized and quality-controlled.The other temperature data is the reanalysis data product, ERA5, a product of the European Centre for Medium-Range Weather Forecasts (ECMWF).ERA5 is the latest climate reanalysis dataset that followed the ERA-Interim data (Hans et al., 2019), and its suitability for climatic studies has been demonstrated in various studies (Gleixner et al., 2020;Graham et al., 2019;Tetzner et al., 2019;Zhu et al., 2021).Cloud cover data were obtained from the Climate Research Unit (CRU TS v. 4.05) and EUMESAT Satellite Application Facility on Climate Monitoring (CM SAF).For the moisture flux and total column water vapour, the ERA5 reanalysis data products were used.Radiation data, which includes shortwave and longwave radiation flux, were also obtained from the ERA5 reanalysis.For this study, soil moisture data were obtained from multiple data sources: European Space Agency's Climate Change Initiative (ESA CCI SM v04.7),Global Land Evaporation Amsterdam Model (GLEAM v3.5) and Global Land Surface Data Assimilation System (GLDAS), with a spatial resolution of 0.25 × 0.25 .Additional soil moisture data was obtained from the Climate Prediction Center (CPC SM), which has a spatial resolution of 0.5 × 0.5 .
The bare land areas which mostly characterize the hyperarid were masked out due to the lower coverage of station observation over the area.The CRU data used had total coverage of 83% for grid cells with more than three stations (Figure 2).The regions below latitude 25 N and above 25 S which had the least number of ground observations were not considered.Areas with less than two stations per grid were masked out to avoid inconsistency in the trend, as the data over these areas are derived by interpolation (Harris et al., 2020) disagreement.They concluded that the selection of postprocessing methods such as gridding and interpolation significantly impacts the limited data areas.
To ensure consistency in the spatial representation of the datasets due to the varying resolutions, the datasets were gridded to 0.5 spatial resolution.For the trend analysis, the Mann-Kendall trend test was used to detect the change in the climatic variable over time.In addition to the Mann-Kendall test, the Theil-Sen slope estimator was also used to quantify the magnitude of change for the climatic variables over time.The trend significance was assessed at both 95% and 99% confidence levels using a two-tailed Student's t test.To examine the relationship and contribution of various factors to the change in temperature, simple linear regression was used.Standard Normal Homogeneity test (SNHT) was used to examine the temperature time series for breakpoints that could have occurred as a result of inhomogeneity in the observation dataset.

| Diurnal temperature range across aridity zones
In the current study, we found that the DTR over the global whole land area had an insignificant changing trend in recent decades from 1979 to 2020 (Figure 3).However, when we split the global land area into two groups, dryland versus non-dryland, we found that the DTR had significantly increased in global drylands but significantly decreased in the non-drylands.Geographically, we found that DTR increased over Europe, Australia and the southern part of Africa but mainly decreased at the high latitudes.The zonal average trends show decreasing DTR trend over the North Hemisphere, while it increased over the Southern Hemisphere.This is consistent with Guan et al. (2022) who reported a decreasing DTR trend over the Northern Hemisphere between 1902 and 2011.They reported a decrease in DTR across the aridity gradient over the Northern Hemisphere, but positive trend was observed in the dry region of the northwestern hemisphere, which agrees with the results presented in this study.
Although both datasets show different magnitudes of change, the pattern of change was similar along the aridity gradient.This was evident as we examined the changes in DTR along the global moisture gradient and found that it decreased across the aridity zones, from dry to wet (Figure 4).Along the aridity gradient, T min increased by 0.2-0.39CÁdecade −1 and T max increased by 0.22-0.34CÁdecade −1 for the CRU dataset, while T min changed between 0.2 and 0.49 CÁdecade −1 and T max between 0.33 and 0.4 CÁdecade −1 for the ERA5 dataset.In the areas where the aridity index (AI) was lower than 0.5, T max increased more rapidly than T min , which resulted in a significant increase in DTR during the past four decades except the extremely dry areas with AI < 0.1 for both datasets.With further increase in moisture condition, AI > 0.5, the DTR started declining due to a greater increase in T min than T max .We found that the DTR significantly decreased in the wet areas where AI > 1.0 for the CRU dataset, but not for the ERA5 dataset.For both datasets, the greatest decrease in DTR was found in the areas where AI was around 1.5.We also found the increase in DTR at the dry areas to be consistent for both datasets when assessed for breakpoints in the time series data across the aridity index gradient (Table S1, Supporting Information).We examined the changes in cloud cover, moisture flux and atmospheric water vapour along the moisture gradient as shown in Figure 5. Result showed that cloud cover decreased mostly at the dry regions, with the most decrease around 0.2 < AI < 0.5 for both cloud cover datasets (CC CRU and FCC CMSAF ).The changes in total column water vapour (TCWV) revealed little change in the dry regions, especially 0.2 < AI < 0.5, while it increased the most in the non-dry regions.The average moisture flux (MF) trend along the moisture gradient revealed a contrasting change at the dry and non-dry regions.Moisture flux decreased in regions where AI < 0.6 but increased in the regions where AI > 0.65.We also examined the correlation between the changes in these parameters and the change in maximum and minimum temperatures along the moisture gradient (Figure 6).Maximum temperature (T max ) had no significant correlation with cloud cover, while minimum temperature had a significant correlation ).Over the Southern Hemisphere where DTR increased, moisture flux decreased significantly (−0.87 mmÁdecade −1 ), while it increased significantly over the Northern Hemisphere, particularly at the non-dry regions (0.58 mmÁdecade −1 ).The average trend of atmosphere water vapour increased significantly over the Northern Hemisphere, particularly in the non-dry regions (0.16 kgÁm −2 Ádecade −1 ).
The DTR ERA5 trend over the non-dry region of the Southern Hemisphere however shows a significant increase (0.13 CÁdecade −1 ) while no change was observed from DTR CRU .This contrast is assumed to be due to the total land cover of the Southern Hemisphere which is predominantly dominated by the dry regions as revealed by the spatial plot of aridity index distribution in Figure 1.The non-dry region where DTR ERA5 increased and DTR CRU did not change is central Africa, where there is data quality issue.
The impact of atmospheric water vapour and cloud cover on DTR change is associated with their influence on downward longwave and shortwave radiation.Downward longwave and shortwave radiation are influenced by changes in atmospheric water vapour, and cloud cover.We examined the correlation between these parameters as shown in Figure 8.The correlation between water vapour and downward longwave radiation (Figure 8) showed significant positive correlation for both dryland and non-dryland (0.77 and 0.86, p < 0.01, respectively), indicating an increase in downward longwave radiation flux as atmospheric water vapour increases, with a more significant effect at the non-dryland.Cloud cover had a negative correlation with downward shortwave radiation (dryland: 0.72 and 0.78, p < 0.01; non-dry: 0.35 and 0.69, p < 0.05 and 0.01, respectively), indicating an increase in downward shortwave radiation as cloud cover decreases.

| DISCUSSION
Earlier studies show a decline in DTR since the 1950s (Easterling et al., 1997;Vose et al., 2005;Zhou et al., 2009).Vose et al. (2005) identified a significant decrease in DTR of about −0.066 CÁdecade −1 between 1950 and 2004 across the global land areas.An earlier study of DTR change between 1950 and 2004 showed a greater decrease in DTR trend over the drier areas compared to the more humid areas (Zhou et al., 2009), which opposes the findings of the current study.The decrease in DTR in the second half of the 20th century was linked to the strong effect of the increase in clouds and precipitation (Dai et al., 1999).It was also suggested that the large-scale effect of elevated greenhouse gases and aerosols contributed largely to the observed decrease in DTR (Zhou et al., 2009).The study pointed to the coincidental decrease in precipitation and clouds with DTR decrease, concluding that other mechanisms could have also been involved in the DTR decrease.Several datasets revealing DTR trends since the 1980s have however shown that there are more areas where DTR trend is increasing recently compared to the previous periods when there was general agreement in the globally decreasing trend (Thorne et al., 2016).Over Western and Eastern Europe for example, the DTR changed from a decreasing trend to an increasing trend since the 1970s and 1980s, respectively (Makowski et al., 2008).It is reported that the period of decreasing DTR trends has stopped as of the early 1980s, indicating a reversal (Sun et al., 2019;Thorne et al., 2016;Trenberth et al., 2007).This explains the insignificant change in DTR over the global land area as a whole reported in this study.According to Huang et al. (2023), the raw CMIP6 models have also failed to capture this reversal in the DTR trend.Focusing on the recent decades with emphasis on the moisture gradient, our study revealed a greater increase in DTR at the dry areas compared to the decrease at the wetter areas, indicating the contribution of arid regions to the recent reversal in DTR trend.
The conditions associated with the differential change across the dry/wet areas were examined in terms of moisture change.The changes in cloud cover, moisture flux (defined here as the net rate of moisture exchange between the land surface and the atmosphere), and atmospheric water vapour were observed to have played significant roles in DTR change along the aridity gradient.The impact of cloud cover on DTR is linked to the incoming solar radiation through reflection at the top of the atmosphere.As cloud cover increases, the incoming solar radiation reaching the surface reduces, which causes less warming.The reduced warming due to increased cloud cover has been shown to impact DTR as a result of less increase in maximum temperature (Lauritsen & Rogers, 2012;Wang et al., 2014).Results presented here however show that the effect of cloud was from the change in T min along the aridity gradient rather than change in T max.The effect of cloud on T min is due to the feedback as clouds are considered to trap heat within the atmosphere, therefore causing more warming at night (Lindvall & Svensson, 2015), leading to a further reduction in DTR.
We found that the moisture flux also impacted the change in DTR along the aridity gradient.The moisture flux effect is linked to the recent elevated temperature, which favours enhanced evaporation (Pascolini-Campbell et al., 2021) and, therefore, the release of energy during the surface to atmosphere moisture transfer.This creates feedback in which increased temperature induce enhanced evaporation and evaporation further feedback to the climate, cooling the temperature (Alkama & Cescatti, 2016;Li et al., 2015;Shen et al., 2015).However, the potential evapotranspiration of dryland is substantially higher than the actual evapotranspiration (Huang et al., 2016) and is expected to continually increase with elevated warming (Feng & Fu, 2013).The drylands are also characterized with little vegetation, mostly grassland and shrublands, which have smaller leaf area index than the more highly vegetated wet areas; therefore, there is less transpiration effect.Although precipitation increases over the drylands, enhanced potential evapotranspiration associated with elevated warming may cancel out the effect of precipitation, which leads to drier conditions (Sherwood & Fu, 2014), therefore more warming (Daramola & Xu, 2022).This implies the further moisture-limited conditions of the drylands.Moisture flux along the aridity gradient is assumed to have a direct and indirect effect on DTR though differential changes in T max and T min .The direct effect is linked to the cooling impact on temperature during the day, while the indirect effect is through the associated changes in cloud cover and atmospheric water vapour (Figure S1).With changes in atmospheric water vapour occurring, a change in downward longwave radiation is expected.Atmospheric water vapour is regarded as a greenhouse gas (GHG), with significant contribution to the global warming induced by increased GHGs concentration.Apart from the other anthropogenically induced GHGs, water vapour plays a key role in the climate feedback (Gordon et al., 2013;Schmidt et al., 2010).Atmospheric water vapour is considered the largest contributor to the natural GHG effect, accounting for about 50% of the absorbed longwave radiation (Schmidt et al., 2010), which makes it a major source of climate feedback.
The differential warming between maximum and minimum temperature leading to change in DTR has been observed to impact terrestrial ecosystem through plant's response to day and night-time warming (Bai et al., 2012;Peng et al., 2013;Wan et al., 2009), thus impacting the terrestrial ecosystem carbon budget (Su et al., 2015;Wen et al., 2018).Plant photosynthesis occurs in the daytime, while respiration is continuous during the day and night (Peng et al., 2013).Experimental results have shown that night-time warming causes increased plant respiration (Xia et al., 2010).However, night-time warming has also been revealed to enhance the next-day photosynthesis due to higher carbohydrate utilization at night, which led to overcompensation of the increase in carbon loss from increased respiration, suggesting a negative feedback effect on the climate through carbon sequestration (Wan et al., 2009).Night-time warming was shown to increase the ecosystem's carbon efficiency by more than 50% (Wang et al., 2020a).Increased respiration from night-time warming may also have implications on overall biomass accumulation (Kläring & Schmidt, 2017;Sunoj et al., 2016Sunoj et al., , 2020)).For example, maize growth was negatively affected by higher night-time warming while the reverse was the case for greater daytime warming (Sunoj et al., 2016).Similarly, increased respiration from increased night-time warming limited the growth of sorghum due to the utilization of carbohydrate for damage repair instead of plant growth (Sunoj et al., 2020).Also, greater increase in daytime temperature was found to enhance the growth of tomato through increased photosynthesis and improved nutrient uptake (Yang et al., 2016).These point to how crucial the changes in day and night-time temperature is to food production especially in drylands where 44% of the cultivated systems and $50% of global livestock are covered by the drylands (Safriel et al., 2005).
The changes in maximum and minimum temperatures have also been shown to have varying impacts on vegetation greening as well as carbon exchange at wet and dry regions (Peng et al., 2013).For example, increased maximum temperature in the drylands may limit photosynthesis, compared to the effect in boreal regions where temperature is a limiting factor, and increased maximum temperature may cause increased vegetation (Peng et al., 2013).For cold and dry regions, more daytime warming may also aggravate drought conditions, whereas more night-time warming may contribute to reduced frost damage (Shen et al., 2016).Our finding that T max increased at a greater rate than T min in global drylands may relatively impact plant interactions.Asymmetric warming with associated moisture changes can impact the interactions between grasses and Crassulacean Acid Metabolism (CAM) plants.For example, rainfall variability may favour grasses over CAM plants; however, in asymmetric warming conditions, CAM growth is enhanced due to the mortality of grasses as a result of the drought stress induced by limited deep soil water for grasses (Huang et al., 2020).
The focus of differential warming has been on the greater nocturnal warming with the associated decrease in DTR (Lindvall & Svensson, 2015;Zhou et al., 2009).The results presented in this study, however, show more daytime compared to the night-time warming in the dry regions.The characteristics of drylands makes it very important in the global carbon budget.Global drylands contribute largely to the interannual variability and change in global carbon (Ahlström et al., 2015;Poulter et al., 2014), and account for about 32% of the global carbon stored in the soil (Plaza et al., 2018).Changes in surface moisture and elevated warming are consequential to the soil organic carbon as drier soils store less carbon (Hanan et al., 2021).The lower increase in night-time warming and insignificant change in the long-term trend of soil moisture (Figure S2) raises the question of how these may impact the coupled processes of carbon assimilation and loss in the drylands.Drying conditions in the dry areas of the northeastern hemisphere is suggested to have impacted the differential warming of the region (Guan et al., 2022) due to an increase in the carbon release.

| CONCLUSION
Research on diurnal temperature range is an important study in understanding climate feedback effect in our changing world.While earlier studies reported a general decrease in the DTR across the globe, our study revealed the recent change in the DTR trend in more recent decades.We examined the recent changes in diurnal temperature range across the moisture gradient and showed that moisture flux plays an important role in changing day and night-time warming along the moisture gradient.It is important to consider moisture changes along the aridity gradient in the assessment of differential warming.Other factors such as cloud cover and atmospheric water vapour also influenced this observed change in the diurnal temperature range.Our study indicates that while other recent studies have also reported the reversal of the DTR trend, the increase in DTR trend since the 1980s has primarily occurred in the arid regions.
It is important to note that our study presented here focuses on reporting the DTR phenomenon and possible correlations with other variables.Considering the complex nature of land and atmosphere interactions, there are several other processes that may be involved in the changes associated with day and night-time warming.The processes further need to be explored.

F
I G U R E 1 Global aridity index classification [Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 2 Distribution of average number of stations for CRU TS between 1979 and 2020 [Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 3 The spatial trend and average DTR trend at the global (G), dryland (D), non-dryland (ND); zonal plot of change in DTR (a) CRU TS and (b) ERA5 between 1979 and 2020 (*significant at 0.05, **significant at 0.01) [Colour figure can be viewed at wileyonlinelibrary.com] Thorne et al. (2016) used multiple datasets to assess the DTR trend from 1950 and observed more consistency in the DTR trend prior to 1979, while post-1979 showed more F I G U R E 4 Diurnal temperature range (DTR) change across aridity index (a) CRU TS and (b) ERA5 (*significant at 0.05, **significant at 0.01).Aridity Index is the ratio of precipitation to potential evapotranspiration (P/PET); lower AI represent dry areas, and high AI represents wet areas [Colour figure can be viewed at wileyonlinelibrary.com] E 5 Average change in (a-b) cloud cover, (c) total column water vapour (TCWV) and (d) moisture flux (MF) along the Aridity Index [Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 6 Correlations between (a) T max and cloud cover (FCC and CC), (b) T max and moisture flux (MF) and total water vapour (TCWV), (c) T min and cloud cover (FCC and CC), (d) T min and moisture flux (MF) and total water vapour (TCWV), along the Aridity Index as grouped [Colour figure can be viewed at wileyonlinelibrary.com]3.2| Changes in cloud cover, moisture flux and atmospheric water vapour

F
I G U R E 7 Average change in (a) DTR, (b-c) cloud cover, (d) moisture flux (MF) and (e) total column water vapour (TCWV), over the Northern Hemisphere (NH), Northern Hemisphere dryland (NH-Dry), Northern Hemisphere non-dryland (NH-Non), Southern Hemisphere (SH), Southern Hemisphere dryland (SH-Dry), Southern Hemisphere non-dryland (SH-Non) [Colour figure can be viewed at wileyonlinelibrary.com] between 0.44 and 0.59 with cloud cover.Moisture flux (MF), however, had a significant correlation with T min but not with T max .The changes in cloud cover, moisture flux and atmospheric water vapour play essential roles in the changes in DTR.We show this in Figure 7 which reveals the average change in these parameters over the Northern and Southern Hemispheres.DTR trend increased significantly over the Southern Hemisphere (0.09 CÁdecade −1 [DTR CRU ], 0.2 CÁdecade −1 [DTR ERA5 ]), but no significant change was observed over the Northern Hemisphere except in the non-dry regions where DTR decreased (−0.04 CÁdecade −1 [DTR CRU ], −0.12 CÁdecade −1 [DTR ERA5 ]

F
I G U R E 8 Correlation between the anomalies of (a) downward longwave radiation (DLW) and total water vapour (TCWV), (b) downward shortwave radiation (DSW) and cloud cover (FCC and CC), for dryland and non-dryland (*significant at 0.05, **significant at 0.01) [Colour figure can be viewed at wileyonlinelibrary.com]