Impacts of Cloud Microphysics Modifications on Diurnal Convection and the ISO over the Maritime Continent: A Case Study of YMC-Sumatra 2017

Relationship between diurnal convection and the intraseasonal oscillation (ISO) over the western Maritime Continent (MC) was investigated by a case study of an ISO event that occurred during the Years of the Maritime Continent (YMC)-Sumatra 2017 campaign. Two sets of global cloud-permitting simulations using cloud microphysics settings for ISO prediction (CTL) and for climate simulation (MOD) were performed to clarify their impacts. CTL had biases of weaker diurnal variation and smaller precipitation amounts over land than in observations; these were reduced in MOD by higher probabilities of local intense convection in the middle troposphere and higher precipitation efficiency. The enhanced convection over land coincided with suppressed convection over the surrounding ocean, especially at the diurnal peak time of land convection. Exception is the onset period of the ISO convection, when upward moisture advection and precipita- tion increased also over ocean in MOD than in CTL at the diurnal peak time of oceanic convection. These results suggest that the enhancement of local convection over the MC by the cloud micro-physical processes basically hinders the ISO convection by the activation of land convection, but it also favors the ISO convection development over ocean during the onset period. ( modifications on diurnal convection and the ISO over the Mari - time Continent: A case study of YMC-Sumatra 2017.


Abstract
Relationship between diurnal convection and the intraseasonal oscillation (ISO) over the western Maritime Continent (MC) was investigated by a case study of an ISO event that occurred during the Years of the Maritime Continent (YMC)-Sumatra 2017 campaign. Two sets of global cloud-permitting simulations using cloud microphysics settings for ISO prediction (CTL) and for climate simulation (MOD) were performed to clarify their impacts. CTL had biases of weaker diurnal variation and smaller precipitation amounts over land than in observations; these were reduced in MOD by higher probabilities of local intense convection in the middle troposphere and higher precipitation efficiency. The enhanced convection over land coincided with suppressed convection over the surrounding ocean, especially at the diurnal peak time of land convection. Exception is the onset period of the ISO convection, when upward moisture advection and precipitation increased also over ocean in MOD than in CTL at the diurnal peak time of oceanic convection. These results suggest that the enhancement of local convection over the MC by the cloud microphysical processes basically hinders the ISO convection by the activation of land convection, but it also favors the ISO convection development over ocean during the onset period.

Introduction
The tropical intraseasonal oscillation (ISO) is a major target of subseasonal to seasonal prediction due to its marked impacts on global weather (Zhang 2013;Vitart et al. 2017); however its accurate prediction over the Maritime Continent (MC) has been challenging (Vitart and Molteni 2010;Kim et al. 2016;Wang et al. 2019;Ahn et al. 2020). The MC is characterized by diurnally forced active convection that is driven by local processes associated with complex land-ocean distribution and steep orography (Yang and Slingo 2001;Mori et al. 2004). Active local convection over the MC affects the ISO intensity and propagation (Innes and Slingo 2006;Kim et al. 2014;Zhang and Ling 2017). Hagos et al. (2016), by sensitivity experiments with and without the diurnal insolation cycle, highlighted the blocking effects of diurnal convection on the ISO propagation. To improve the prediction of the ISO over the MC, adequate representation of local and large-scale convective processes is essential.
The development of general circulation models (GCMs) has led to remarkable progress in reproducing the ISO propagation over the MC (Wang et al. 2019;Ahn et al. 2020). Ahn et al. (2020) demonstrated that the models participated in Coupled Model Intercomparison Project Phase 6 (CMIP6) outperform those in Phase 5 (CMIP5) by reducing mean moisture field bias due to updates in convection schemes. On the other hand, the latest GCMs still exhibit biases in the simulation of diurnally forced MC convection (Branowski et al. 2019;Argüeso et al. 2020). Regional cloud-permitting models are useful for studying the MC convection, including its dependence on the ISO (Hagos et al. 2016;Lane 2017, 2018;Argüeso et al. 2020), while they tend to overpredict diurnal convection over land. Appropriate modeling of cloud microphysics and shallow convection is a key issue (Argüeso et al. 2020). In addition, scale interactions in regional models are restricted by their domains. With this respect, global cloud-permitting models have an advantage (Miura et al. 2007(Miura et al. , 2015Miyakawa et al. 2014;Fujita et al. 2011;Nasuno et al. 2017;Nasuno 2019).
The international project Years of the Maritime Continent (YMC) was conducted to improve our understanding and prediction of the global MC impacts; one YMC campaign collected field data in Sumatra in 2017 (Yoneyama and Zhang 2020), for which near real-time forecasts were executed using a global 7-km mesh model. Nasuno (2019) used this simulation dataset to quantify the moisture transport in the ISO, and found that the high-frequency variabilities contributed to the ISO-scale moistening in the preconditioning phase of the ISO over ocean, whereas such relationship was not clear over land. However, the role of diurnal variation could not be thoroughly examined due to the single simulation setup and model biases.
In the present study, two series of simulations, each with different cloud microphysics settings, were performed, and the resulting systematic differences in diurnal convection over the MC were examined to clarify how the model physics altered the behavior of local convection and its effects on the simulated ISO.

Simulation setup
In this study Nonhydrostatic Icosahedral Atmospheric Model (NICAM, Satoh et al. 2014) was used with a horizontal mesh size of 7 km globally. Moist convection was explicitly calculated using a six-category single-moment bulk cloud microphysics scheme (NSW6, Tomita 2008) without operating implicit convection schemes. The details of the model configurations for the near real-time forecasts (CTL) are provided in Nasuno (2019). The simulations were initialized daily at 0000 UTC throughout the YMC-Sumatra 2017 campaign period using the National Centers for Environmental Predication (NCEP) final operational global analysis (NCEP FNL). In this study, 5-day outputs for the period between 9 November and 9 December 2017, covering the full life cycle of an ISO event, were analyzed. For the sensitivity simulations (MOD), the model configurations were based on the climate simulations (Kodama et al. 2021) and initialized using the same data as used in CTL. The most significant difference between the two setups was in the cloud microphysics settings; other minute changes had little impacts on the week-long simulation results (Supplement 1).
In the CTL settings, the terminal velocities of the precipitating condensates (i.e., snow, graupel and rainwater) were reduced to avoid excessively intense, sporadic precipitation and to facilitate large-scale organization of convection, which were valid for the month-long ISO simulations (Miyakawa et al. 2014;Miura et al. 2015). In the MOD settings, the modification of NSW6 proposed by Roh and Satoh (2014), which was based on evaluation of cloud itation patterns and the diurnal differences over land and ocean, but with less contrast than observed with CMORPH ( Fig. 2a−2f). In comparison with CTL, MOD produced smaller deficits in land precipitation (Fig. 2, left panels) and diurnal differences (Fig. 2, right panels; Fig. 3). The time series of precipitation exhibited typical ISO evolution, with early activation of land precipitation overtaken by oceanic precipitation during the ISO onset (Fujita et al. 2011;Peatman et al. 2014;Zhang and Ling 2017). Diurnal variation was greater over land than over ocean, especially during the suppressed period prior to the ISO onset. This contrast was clearer in MOD than in CTL (Fig. 3).

Moisture advection
The essential role of the moisture variability in the ISO dynamics has been well established (Fuchs and Raymond 2005;Raymond and Fuchs 2009;Sobel et al. 2001;Sobel and Maloney 2013;Nasuno et al. 2015Nasuno et al. , 2017. In this study, sum of the horizontal and vertical components of the moisture advection terms (Eq. S3-1) is presented (definition and analysis of the moisture tendency terms are provided in Supplement 3). Time series of the column integrated moisture advection were positive throughout the period (Fig. 3d) and nearly balanced with the sum of precipitation and surface flux. Figure 4 shows time-height section of moisture advection in the land and ocean area of the target domain (90°E−120°E, 12°S− 8°N). During the ISO onset period (21−25 November), positive moisture advection over ocean was greater in MOD than in CTL, with clearer diurnal deepening (Figs. 4a, 4c, and 4e), concurrent with the ISO-scale anomalous moistening (Fig. 4g). These suggest a supporting role of oceanic diurnal convection to the ISO convection onset through the intensification of upward motion which is tied to the enhancement of local latent heat release. Over land, anomalous moistening was greater in MOD than in CTL during the ISO suppressed period (e.g., before 18 November; Figs. 4b, 4d, and 4f). At the ISO time scale, variation over land was generally the opposite of that over ocean (i.e., a seesaw relationship between land and oceanic regions; Figs. 4g and 4h). microphysics processes by a satellite simulator, were employed, with parameter tuning for global cloud climatology (Kodama et al. 2015). The MOD settings decelerated conversion of cloud ice to snow and growth of graupel, which led to increased (reduced) amount of ice (graupel) with lower peak level of snow in MOD than in CTL (Fig. 1d). These changes in the ice condensates were accompanied with little changes in latent heat release, but some increase (decrease) in latent heating (water vapor content) occurred accordingly (Figs. 1b and 1c).

Analysis method
Following Nasuno et al. (2017) and Nasuno (2019), the 5-day outputs were combined to create time series covering the entire period (i.e., averaging outputs initialized on different dates for each valid time), and horizontal and vertical components of moisture advection terms were calculated (Supplement 3) using these time series; the ISO-scale components were extracted using a 7-day running average. For comparisons, precipitation data from the Climate Prediction Center morphing method (CMORPH) products (Joyce et al. 2004) and atmospheric data from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA-Interim and ERA5; Dee et al. 2011;Hersbach et al. 2020) were used.

Precipitation
The period-mean distribution and time series of simulated precipitation over the MC are presented in Fig. 2 and Fig. 3, respectively. Large precipitation amounts over the western MC ( Fig.  2a) were associated with the ISO event that occurred in this period (Fig. 3a, Fig. 1 of Nasuno 2019). In the high-precipitation region (e.g., along and off coasts), diurnal differences were pronounced, with opposite signs over land and ocean (Fig. 2b), indicating that diurnal convection accounted for a large proportion of the mean precipitation there. Both simulations reproduced the major precip-    To more clearly see the responses of the land-ocean contrast to the microphysics modifications, time-averaged moisture advection and its diurnal differences were calculated (Fig. 5). Differences between MOD and CTL mainly appeared at height (z) of 4−10 km, with enhanced (reduced) moistening over land (ocean) (Figs. 5a, 5b, and 5c). These trends were more evident at the diurnal peak time of land convection, especially in the suppressed period prior to the ISO convection onset (Fig. 5d). The moistening tendency (MOD -CTL) over land continued to increase during the ISO convection onset period (21−25 November; Figs. 5f and 5g), and then decreased during the latter active period (26 November to 2 December; Figs. 5h and 5i).
Over ocean, the drying tendency (MOD -CTL) was more pronounced, reflecting intensification of compensating subsidence via enhanced upward motion over land. The exception was the ISO convection onset period, when the moistening tendency also appeared over ocean at the diurnal peak time of oceanic convection (Fig. 5g). In both simulations, the continuous increase in the moistening was robust over ocean, in contrast to the reduction in the latter active period over land (Fig. 5h). Such land-ocean contrast is supporting of the known precipitation variation associated with the ISO (e.g., Peatman et al. 2014).

Decomposition of vertical motion
Given the compensating relationship between convection over land and ocean in the western MC (Fig. 5), a question arises which change (e.g., intensity, area) of the upward/downward motion by the cloud microphysics modifications was the major factor. Vertical motion W over the target domain can be decomposed as follows: W = W u + W d = a u w u + a d w d = a l w l + a s w s = a ul w ul + a dl w dl + a us w us + a ds w ds , where a u + a d = a s + a l = 1.0; a and w stand for the areal fraction and the mean vertical velocity of each component; and the subscripts l, s, u, and d denote land, ocean, upward, and downward,  Fig. 6, together with w l and w s in the ECMWF reanalyses. The difference between w l and w s was greater in ERA5 than in ERA-Interim, and those in the simulations were within the realistic range (Figs. 6a and 6b). The following paragraphs discuss the differences between MOD and CTL. On average, w l (w s ) [red (blue) lines] increased (decreased) below z = 10 km by the cloud microphysics modifications (Fig.  6b), elucidating the difference in moisture advection (Fig. 5). In all components, w l [red] and w s [blue] were greater (smaller) in MOD [broken lines] than in CTL [solid lines] below (above) z = 10 km (Fig. 6d). The increase in w l /w s [red/blue lines] was especially pronounced in w ul and w us in the middle troposphere (z = 4−10 km) [positive side in Fig. 6d], which accounted for the enhanced late evening land convection and the morning oceanic convection during the ISO onset period (Figs. 5f and 5g). These were accompanied with decrease in a u (increase in a d ) (Fig. 6c), indicating compensation between the upward and downward motion within the target domain. The reduction in a us and increase in w ds accounted for the mean decrease in w s [blue lines in Fig.  6b] and moisture advection over ocean [blue lines in Fig 5a]. The intensification of w u was mainly attributable to the reduction of  graupel (and snow) amounts (Fig. 1d) with the lessened drag force in MOD. In the lower troposphere, intensification of w d (Fig. 6d) and reduction in a d (Fig. 6c) cancelled out with little difference in w u . These were responsible for the smaller impacts of microphysics modifications below z = 4 km than those above (Figs. 4 and 5). Changes in a d and w d (i.e., more occurrences of local intense downdrafts than slow broad subsidence) were related to a reduction in rainwater amounts (Fig. 1d) and less evaporative cooling due to the greater terminal velocity. In the upper troposphere (z > 10 km), w u and w d were reduced with increase in a u both over land and ocean (Figs. 6c and 6d). These were owing to the increase in cloud ice with decrease in precipitating condensates (snow and graupel) in the presence of cloud-radiative interaction (Figs. 1c and 1d) [i.e., broader coverage of thinner clouds, which was the major improvement by Roh and Satoh (2014) against satellite observations]. The impacts of these changes in the upper troposphere were secondary to those in the lower levels. Table S4 summarizes the contribution of each component to the total difference between MOD and CTL.
Finally, Fig. 7 demonstrates the total (area-weighted) difference between MOD and CTL in W and a u in the ISO. Anomalous positive a u in the middle troposphere (Figs. 7a and 7b) and W below (Figs. 7c−7f) during the onset period contributed to the deepening of the ISO convection more in MOD than in CTL. Subsequently, positive a u anomalies extended to the upper and lower troposphere with peak W anomalies at z = 8−10 km, representing the development of deep, organized convection. In the latter active period, decreases in a u and W were more significant in MOD than in CTL, indicating the weaker sustainability of the organized convection.

Conclusions
In this study, two series of global cloud-system-resolving simulations with different cloud microphysics settings, for ISO prediction (CTL) and for climate simulation (MOD), were examined to clarify the effects of the cloud microphysics modifications on local convection and their linkage with the ISO.
Intense upward motion with latent heat release in the middle troposphere occurred more frequently in MOD than in CTL, owing to the reduced drag force of precipitating condensates (Figs. 6 and S1-2). These led to enhanced local diurnally forced convection and a resultant increase in the mean precipitation over land (Figs. 2 and 3), accompanied by systematic differences in moisture advection (Figs. 4 and 5). The enhancement of upward motion selectively occurred in the convectively active regions (e.g., over the MC, tropical continent, and inter tropical convergence zone), without significant changes in the global circulation (Fig. S5). Thus, compensating subsidence was rather constrained around the neighboring domains. As a result, the upward motion and moisture transport were suppressed over the surrounding ocean within the western MC, through the intensification of mean downward motion and reduction in the fractional area of upward motion. In terms of bulk moisture balance, the enhanced precipitation in MOD with little change in moisture gain led to higher precipitation efficiency (Supplement 3). Changes in the lower troposphere (reduced fractional area of downward motion and intensification of mean downward motion) and in the upper troposphere (increased ice clouds with slower upward motion and broader coverage) also appeared, although their impacts on changes in moisture advection and precipitation were secondary.
In view of the impacts on the ISO, the temporal variation of moisture advection showed generally opposite tendencies over land and over ocean, consistent with the above analysis. The exception was the ISO onset period, when anomalous upward moisture advection with enhanced diurnal variation occurred over ocean. This can be attributed to stronger radiative forcing over the ocean with less cloud amount in MOD (Fig. 1c), as well as gradual development of large-scale circulation associated with the ISO. This result implies that the MOD settings enhance interactions between ISO and diurnal variation over ocean, with anomalous moistening by the latter (e.g., Nasuno et al. 2015;Tseng et al. 2015). Responses of the radiative forcing associated with the ISO convection (e.g., stronger/weaker surface insolation in the convectively inactive/active period) to the microphysics modifications,  The enhancement of local deep clouds in MOD, which selectively appear over land, was consistent with the dominance of convective precipitation over land in observations (Sakaeda et al. 2017(Sakaeda et al. , 2020 and in regional cloud-permitting simulations (Vincent and Lane 2018), whereas the development of stratiform clouds, which accounts for the major body of the ISO convection especially over ocean, was less evident in MOD than in CTL, despite the general improvement of the representation of the upper-level clouds (Kodama et al. 2021;Roh and Satoh 2014). The results of this study suggest that more realistic settings lie between the two settings, which is pursued in forthcoming investigations, as well as better understanding of scale interactions, including the long-term responses of the large-scale conditions through extended-range simulations.