Intra-seasonal contrasting trends in clouds due to warming induced circulation changes

Quantification of long term changes in cloud distribution and properties is critical for the proper assessment of future climate. We show contrasting trends in cloud properties and cloud radiative effects over Northwest Indian Ocean (NWIO) in south Asian summer monsoon. Cloud top height (CTH) decreases in June (− 69 ± 3 myr−1) and July (− 44 ± 3 myr−1), whereas it increases in August (106 ± 2 myr−1) and September (37 ± 1 myr−1). These contrasting trends are investigated to be due to the changes in upper tropospheric winds and atmospheric circulation pattern. Strengthening of upper tropospheric easterlies and changes in vertical wind dampen the vertical development of clouds in June and July. In contrast, weakening of upper tropospheric winds over NWIO and strengthening of updraft favour the vertical growth of clouds in August. Further, changes in horizontal winds at 450–350 hPa and strengthening of Indian Ocean Walker cell favour the westward spread of high level clouds, contributing to the increase in CTH over NWIO in August. Decrease of cloud cover and altitude in June and July and increase of the same in subsequent months would affect the monsoon rainfall over the Indian region. Proper representation of these intra-seasonal contrasting trends of clouds in climate models is important for the better prediction of regional weather.

Positive values show the regions, where CTH is increasing and negative values indicate those, where CTH is decreasing. Linear trend of CTH in each pixel and its statistical significance are estimated using the trend analysis technique detailed in the method section. All the pixels, except those with black dots, show the regions where estimated trends are statistically significant at 95% confidence level. The map is generated using MATLAB 2020a, www. mathw orks. com.
Westward spread of high altitude clouds. Similar to the seasonal mean spatial distribution, clouds are seen at higher altitudes over the eastern regions of NIO, compared to those over the western regions, in the month of August. (see Supplementary Fig. S4). MODIS-Terra measurements during 2000 to 2017 show that the mean CTH in August over NWIO is 6.46 ± 0.03 km and that over NEIO is 9.48 ± 0.01 km. This difference in cloud altitudes between the western and eastern regions of NIO is corroborated by the results from www.nature.com/scientificreports/ MODIS-Aqua measurements during 2002 to 2017, which show mean CTH of 6.42 ± 0.03 km and 9.29 ± 0.01 km in August over the NWIO and NEIO respectively. Mean CTH over NWIO in August is observed to be increasing, from ~ 5.5 km to more than 7 km over the years (see Fig. 2b). MODIS-Terra measurements show that the increase in CTH over NWIO during 2000 to 2017 is 34%, with respect to the mean value during the period, whereas that over NEIO is merely 9%. MODIS-Aqua measurements also show consistent results of 29% increase in CTH over NWIO and 7% increase over NEIO during 2002 to 2017. Further, mean CTH in August, averaged over the latitude band between 5° S and 15° N, shows westward spread of high level clouds over NIO during 2000 to 2017 (see Fig. 3). Mean CF averaged over the same latitude band also shows the westward spread of clouds over this region (see Supplementary Fig. S5). High level clouds are found mainly over the regions at east of 65°E in the initial years, whereas they spread westwards and seen up to the regions of 60°E in the final years of the study period. Further, we identified the mean longitude, within the latitude band between 5° S and 15° N, with clouds at east of it at higher altitudes (CTH > 7 km, corresponding to ~ 440 hPa) and those at west of it at lower altitudes. This mean longitude is represented as LON 7km , hereafter in this manuscript. LON 7 km is observed to be drifting westward over the years, from ~ 66° E in 2000 to ~ 60° E in 2017 (see Supplementary Fig. S6). Shift of LON 7km shows that clouds with CTH above 7 km are mainly seen over the regions at east of 66° E in 2000 and their westward spread leads to the presence of high level clouds even over the western regions, up to ~ 60° E, in 2017. This westward spread of high level clouds, over the years, leads to the increasing trends in CTH and CF over the NWIO. In general, northward migration of monsoon cloud-bands from the equatorial Indian Ocean trough region to inland plays a major role in the monsoon rainfall over India. Hence the decrease of CF and CTH in June and July over NWIO and increase of the same in subsequent months would affect the precipitation over the Indian region.
Physical mechanisms. Influence of upper tropospheric winds on cloud distribution. Role of higher altitude winds in the westward spread of high level clouds is examined by estimating trends of horizontal wind speed at the altitude levels corresponding to 450 to 350 hPa (see Fig. 4). Strengthening of winds is observed in August over the regions at west of 75° E, extending westward up to 40° E, within the latitude band between 13° S and 3° S. Within the latitude band between 3° S and 5° N, increase in wind speed is seen over the regions at east of 75° E. It is interesting to notice the weakening of winds over the regions of NWIO, where increase in CTH is observed. Over the southern hemispheric part, easterlies are observed to be turning over the region between 40° E and 60° E and becoming southwesterlies towards the regions of increase in CTH. This leads to sweeping of more clouds from the regions of increase in wind speed, over the southern hemispheric part, to the regions where CTH is found to be increasing. Easterlies over the northern hemispheric part (between equator and 5° N) also push more clouds to the west, from the regions where they are strengthening. Cloud distribution over the south Asian summer monsoon region is highly influenced by the Tropical Easterly Jet (TEJ) stream 34,35 . These strong easterly winds sweep clouds at higher altitudes and hence inhibit vertical In general, CTH is higher over the eastern regions (as indicated by red color), compared to that over the western parts. Higher level clouds are observed to be spreading westward, over the years, as indicated by the black arrows. www.nature.com/scientificreports/ growth of clouds over this region beyond an altitude limit (~ 300 hPa) 34 . In order to examine the changes in upper tropospheric winds, trends of wind speed at 200 hPa in the summer monsoon months are examined (see Supplementary Fig. S7). Estimated trends show strengthening of upper tropospheric winds over the NIO at south of the core regions of TEJ, in June and July. This indicates increase in hindrance to vertical growth of clouds over this region in these months, which leads to the negative trends in CTH. However, decreasing trends in wind speed are seen over the NWIO in August, which favour further vertical development of clouds over this region. Indian Ocean Walker circulation cell is strengthening in summer monsoon months, as indicated by the enhancement in updraft over the regions at east of 80° E and strengthening of downdraft over the regions at west (see Fig. 5). These changes in vertical winds favour intensification of upper level easterlies over this region, as the Indian Ocean Walker cell is characterised by upper level easterlies and lower level westerlies along with the updraft and downdraft over the eastern and western regions respectively. However, the weakening of updrafts (or the strengthening of downdrafts) over the regions between 55° E and 80° E in June and July prevents further westward spread of clouds over to the regions at west of 55° E. Compared to these months, strengthening of downdraft in August is observed further west, over the region between 40° E and 50° E. This favours further westward spread of high level clouds in August, leading to the increase in CTH over NWIO. Strengthening of   Fig. S8c). Changes in the strength of the Indian Ocean Walker cell are estimated by considering the difference in mean vertical wind, over the regions 45° E-55° E (descending limb) and 85° E-95° E (ascending limb), at the altitude levels from 600 to 400 hPa for the latitude band between 10° S and equator. Increasing trend of difference in vertical wind between west and east indicates strengthening of the Walker circulation cell over this region (see Supplementary Fig. S9). This strengthening of the Walker circulation cell over the Indian Ocean is reported to be in response to the warming climate 36,37 . This leads to strengthening of surface level westerlies over the equatorial Indian Ocean and enhancement in upward  www.nature.com/scientificreports/ air motion over its eastern parts 37 . This enhancement in convection over the eastern parts of the equatorial Indian Ocean (see Fig. 5) and associated strengthening of easterly winds at higher altitudes lead to westward spread of high level clouds. Strengthening of convection over the eastern regions causes increase in CTH over NEIO also. However, intensification of upper tropospheric easterlies ( Supplementary Fig. S7) inhibits further vertical development of clouds over this region. Warming of the tropical Indian Ocean is reported to be primarily due to the increase in concentrations of greenhouse gases and associated enhancement in downward longwave radiation 38 . This, along with the results presented here, indicates the effect of anthropogenic climate warming on atmospheric circulation and hence on cloud distribution over the south Asian summer monsoon region. Convection activity over the eastern tropical Indian Ocean weakens during El Niño events, whereas it strengthens during La Niña. These variations in strength of convection and associated changes in upper level easterlies lead to modifications in the magnitude of observed trends in CTH over the NWIO (see Fig. 2). Westward spread of high level clouds in August reduces the difference in CTH, existed among the summer monsoon months, over NWIO. Mean CTH values over NWIO from MODIS-Terra during 2000 to 2009 and MODIS-Aqua during 2002 to 2009, are compared with those during the latter period from 2010 to 2017. Mean CTH in August during the former period is considerably lower than that in the other two months. Mean CTH in August during 2000 to 2009 (2002 to 2009) is 5.94 ± 0.03 km (5.99 ± 0.04 km) from Terra (Aqua) measurements, whereas it is 7.71 ± 0.03 km (7.74 ± 0.04 km) and 7.39 ± 0.03 km (7.43 ± 0.04 km) in June and July respectively. However, mean CTH is almost same (~ 7 km) in all the three months during the latter period. Compared to the former period, latter period shows decrease in CTH in June and July.

Changes in convection and
Climate conditions over the equatorial Indian Ocean are affected by the Pacific Decadal Oscillation (PDO), which is a dominant pattern of natural climate variability in the Pacific Ocean at decadal to multi-decadal time scales 39,40 . PDO is reported to affect the circulation pattern, including the Indian Ocean Walker cell and upper tropospheric winds, over this region 41 . While cold phase of the PDO supports the normal pattern of Indian Ocean Walker cell with convection over the eastern regions of equatorial Indian Ocean, warm phase leads to a contrasting pattern with subsidence over the same region 41 . During the current study period, PDO shifted between its warm and cold phases (see Supplementary Fig. S10a). In order to examine the effect of PDO, the trends of CTH in August are examined separately for the warm and cold phases of PDO. Though the rate of increase in CTH is less in the warm phase, compared to that in the cold phase, positive trend in CTH is observed irrespective of the phase changes of PDO (see Supplementary Fig. S10b). Thus, the increasing trend in CTH over the NWIO is persistent, with only changes in the magnitude of rate of increase, in spite of the alterations in climate conditions due to PDO.

Impact on cloud top temperature
The intra-seasonal contrasting changes in clouds lead to increase in CTT over NWIO in the first half of the season and decrease of CTT in the second half (see Supplementary Fig. S11

Discussion
Increase in global temperature has significant effect on the hydrological cycle and cloud distribution, especially over the tropical regions. Cloud feedback in response to the global temperature increase is determined by the net effect of its shortwave and longwave feedbacks. Alterations in cloud cover and altitude affect both of these feedbacks. Our study shows changes in atmospheric circulation pattern and their effects on cloud distribution over the south Asian summer monsoon region. Alterations in atmospheric circulation pattern leading to contrasting trends in cloud properties, even within the same season, show the importance of proper incorporation of these changes in climate models. Intensification of upper tropospheric easterlies and weakening of updraft (strengthening of downdraft) increase hindrance to the vertical development of clouds and hence decrease CTH over NWIO in the first half of the south Asian summer monsoon period. In contrast, weakening of the upper tropospheric easterlies over NWIO and strengthening of updraft provide favourable condition for the vertical developments of clouds and hence increase CTH in the second half of the season. Along with this, changes in horizontal winds at 450 to 350 hPa (~ 7 km to 9 km) altitude levels lead to spread of clouds at these levels to the regions of NWIO.
Warming induced intensification of the Indian Ocean Walker cell and associated strengthening of upper level easterlies favour the westward spread of high level clouds, leading to the increase in CTH and CF over the NWIO in August. However, weakening of updraft (strengthening of downdraft) over the regions between 55° E and 80° E prevents further westward spread of high level clouds in the first half of the season. Strengthening of downdraft in August occurs further west, over the regions between 40° E and 50° E, which causes further westward spread of high level clouds compared to that in the previous months. This enhancement in westward movement of clouds leads to increase in amount of high level clouds and hence increase in mean CTH and CF over NWIO in the second half of the season, whereas the hindrance to this westward movement of clouds causes decrease in amount of high level clouds and hence decrease in mean CTH and CF over NWIO in the first half. Thus the changes in atmospheric circulation lead to contrasting trends in cloud height and cloud cover over NWIO in south Asian summer monsoon, with decrease in CTH (− 69 ± 3 myr −1 in June and − 44 ± 3 myr −1 in July) and CF (− 0.44 ± 0.02% yr −1 in June and − 0.35 ± 0.01% yr −1 in July) in the first half of the season and increase in CTH (106 ± 2 myr −1 in August and 37 ± 1 myr −1 in September) and CF (0.3 ± 0.02% yr −1 in August and 0.03 ± 0.01% yr −1 in September) in the second half. Impact of these intra-seasonal contrasting changes of clouds on the south Asian summer monsoon rainfall needs to be examined. NWIO experiences a net warming in the first half of the season as CRE LW dominates CRE SW , whereas the region experiences a net cooling in the second half as CRE SW dominates CRE LW . However, the seasonal mean CRE NET is observed to be negative, indicating a net cooling over the NWIO. The alterations in cloud properties lead to contrasting changes in both CRE SW and CRE LW in the first and second half of the season. While decrease in cloud cover changes CRE SW by 15.33Wm −2 and decrease in CTH changes CRE LW by − 17.98 Wm −2 in the first half, increase in CF and CTH lead to a change of − 9.89 Wm −2 and 9.57 Wm −2 in CRE SW and CRE LW respectively in the second half. While changes in clouds lead to a stronger cooling trend in the first half of the season, contrasting changes in the second half weaken the seasonal mean cooling trend over NWIO. Alterations in atmospheric circulation pattern and associated modifications in cloud distribution are more likely to occur over several other regions in a warming climate scenario. Global scale analysis needs to be carried out to estimate the net effect of these changes on the cloud feedback. As cloud cover is a crucial factor for the sensitivity of general circulation models 24,42 , capability of proper representations of these intra-seasonal contrasting trends in clouds is decisive for the realistic predictions of regional weather phenomena such as south Asian summer monsoon and associated rainfall.

MODIS cloud parameters.
Trends of cloud properties are estimated using collection 6.1(C6.1), L3 data of monthly mean CTH, CF and CTT at a spatial resolution of 1° × 1°, from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard both Terra and Aqua satellites. MODIS retrieves cloud top pressure (CTP), using CO 2 slicing technique and an alternate method with brightness temperature at 11μm 43,44 . CO 2 slicing method makes use of the variation in radiation absorption within the broad 15 μm CO 2 absorption band 44 . CO 2 slicing technique is mostly suitable for mid level to high level clouds, whereas the latter one is more effective in the presence of low level clouds with CTP greater than 700hPa 43 . Retrieved CTP is converted to CTH and CTT, using gridded temperature profiles from National Centers for Environmental Prediction Global Data Assimilation System 45,46 . Thus, CTH from MODIS is the geopotential height at the retrieved cloud top pressure level. Collection 6 (C6) MODIS cloud top products have improved, compared to its previous versions, in several aspects due to the refinements in algorithm and calibration 47,48 . The calibration approaches employed in the generation of C6 products remove the calibration trends in the data 49

Trend analysis.
Trend analysis is carried out by estimating linear trends of the parameters and corresponding uncertainty in each pixel 51,52 . Significance of the estimated trends is assessed using the values of absolute trend relative to corresponding uncertainty. In order to estimate the trends, a linear model is considered as Y t = μ + ωX t + N t , t = 1, 2, 3, …., n, where Y t , ω, X t , n and N t , are the parameter, trend, time in year, number of years and noise respectively. Further, statistical significance of the estimated trends is assessed using absolute values of the ratio of trend to corresponding uncertainty ω σ ω . σ ω is calculated as where σ N is standard deviation and ϕ is autocorrelation coefficient of noise. If the value of ω σ ω is greater than 2, corresponding trend (ω) is considered to be statistically significant at 95% confidence level 53 .
MERRA2 reanalysis. In order to investigate the role of atmospheric dynamics, horizontal and vertical components of winds are used from Modern-Era Retrospective analysis for Research and Applications Version 2 (MERRA2) reanalysis, by NASA's Global Modeling and Assimilation Office (GMAO). MERRA2 is advanced, compared to MERRA, in terms of assimilation of more satellite observations in the reanalysis 54 . In order to assimilate newer satellite observations, MERRA2 uses an upgraded version of Goddard Earth Observing System Model (GEOS-5), compared to that used in MERRA reanalysis 55 . MERRA2 provides data of atmospheric parameters at a spatial resolution of 0.5° × 0.625° at 72 pressure levels from surface to an altitude corresponding to 0.01hPa 54 . MERRA2 data set used in the present study is obtained through https:// disc. gsfc. nasa. gov.

Oceanic Niño Index. Oceanic Niño Index (ONI) is one of the parameters used to identify El Niño and
La Niña conditions and quantify the strength of ENSO. ONI is estimated as the three month running average of sea surface temperature anomaly over the Niño-3.4 region (5° S-5° N, 120° W-170° W) 33 . Thus, ONI value corresponding to the month of August is the running mean of the anomaly for the three months from July to September. Conditions with ONI ≥ 0.5 °C is considered as El Niño event and that with ONI ≤ − 0.5 °C is considered as La Niña 56  TOA fluxes and CRE from CERES. Changes in TOA fluxes and CRE are examined using the Energy Balanced and Filled (EBAF Ed4.1) level 3b data from the Clouds and the Earth's Radiant Energy System (CERES) 58 . This data set is available through https:// ceres. larc. nasa. gov. The data set includes TOA outgoing shortwave and long wave radiations under clear sky and all sky conditions, which are used to compute CRE SW , CRE LW and CRE NET . CRE SW is computed as, CRE SW = F SW Clear Sky − F SW All Sky , where F SW Clear Sky and F SW All Sky are the TOA shortwave fluxes under clear sky and all sky conditions respectively. Similarly, CRE LW is computed as CRE LW = F LW Clear Sky − F LW All Sky , where F LW Clear Sky is the TOA longwave flux under clear sky conditions and F LW All Sky is the same under all sky conditions. Net cloud radiative effect is obtained as the resultant of shortwave and longwave cloud radiative effects (CRE NET = CRE SW + CRE LW ). Trends of the TOA outgoing fluxes and CREs are estimated, by excluding the years of El Niño and La Niña conditions. Estimated annual trends are multiplied with the total number of years to obtain the changes in fluxes and CREs during the study period.