Figure 18 – uploaded by Yi Ting Thean

See full PDF downloadDownload figure

Table 1. Comparison of continuous incidents that were processed by the real-time flood alert system. The real-time flood alert system can provide effective notification from the comparison results when taking continuous events. It was found that the rain data obtained from the system can inform approximately 20 minutes early warning. This system can alert and provide location and depth information of flood levels, and it can be advance and effective and can be used to make informed decisions of the authorities in the area drainage management. REFERENCES This research tested t data from the RTF system, rain data from the automatic rain gauge, comparing t and rain da he test resu hree types of rainfall data: rain ta from the rain radar. When ts from different rainfall data sources, it was found that rainfall data from the RTF system could be notified automatic ra faster than the rain data from an in detector in 30 minutes, and the rain data from the rain radar in 20 minutes as shown in Table 1 below. For t he accurate evaluation of the RTF system, it was performed by selecting 116 historical rainfall events in ten days duration. The results showed that the RTF system provided 71 out of 116 pre-warming events (61.2%) with rainfall forecasting accuracy above 69.1%. — Table 1 Comparison of continuous incidents that were processed by the real-time flood alert system. The real-time flood alert system can provide effective notification from the comparison results when taking continuous events. It was found that the rain data obtained from the system can inform approximately 20 minutes early warning. This system can alert and provide location and depth information of flood levels, and it can be advance and effective and can be used to make informed decisions of the authorities in the area drainage management. REFERENCES This research tested t data from the RTF system, rain data from the automatic rain gauge, comparing t and rain da he test resu hree types of rainfall data: rain ta from the rain radar. When ts from different rainfall data sources, it was found that rainfall data from the RTF system could be notified automatic ra faster than the rain data from an in detector in 30 minutes, and the rain data from the rain radar in 20 minutes as shown in Table 1 below. For t he accurate evaluation of the RTF system, it was performed by selecting 116 historical rainfall events in ten days duration. The results showed that the RTF system provided 71 out of 116 pre-warming events (61.2%) with rainfall forecasting accuracy above 69.1%.

Related Figures (268)

Table 1. Indicators for calculating DHI, DEI, and DVI.

Fig. 1. Study area showing four administrative regions. Figure 1 shows the study area, which includes 27 provinces in Vietnam belonging to four administrative regions: the South Central Coast (Khanh Hoa, Ninh Thuan, and Binh Thuan); the Central Highland (Kon Tum, Gia Lai, Dak Lak, Dak Nong, Lam Dong); the South East (Ho Chi Minh, Ba Ria — Vung Tau, Binh Duong, Binh Phuoc, Dong Nai, Tay Ninh); and the Mekong Delta region (Long An, Tien Giang, Ben Tre, Tra Vinh, Vinh Long, Kien Giang, Hau Giang, Soc Trang, Bac Lieu, Ca Mau, Can Tho, An Giang, and Dong Thap). DATA AND METHODS Selection of indicators For calculating DRI, SPI and VHI data obtained from SWCEM products was extracted for the study area from global CDF data by using “Make net CDF raster layer” technique in ArcGIS (ESRI [date unknown]-b). This spatial data was classified and mapped to demonstrate monthly drought progression and then transformed into statistical data at the administrative level using Zonal statistics tool of ArcGIS (ESRI [date unknown]-c). The average statistical value of SPI and VHI for each province was used to identify the month with lowest average SPI and VHI value.

The mapping results for drought hazard, exposure, vulnerability, and the combined drought risk index are presented in Fig. 2. Fig. 2. Drought Hazard, Exposure, Vulnerability, and Risk Maps.

Table 1. Detail of all ten experiments performed in this study.

Minimum MSLP time series from all the 3DVar experiments along with CNTL simulation and IMD best- track data are provided in Fig. 3(a). Experiments L03 V02, L10V02, LO3V08 and CNTL have shown an under- estimation from the IMD, whereas L06V10, L10V06, L06V08, L10V10 and LO6V06 have depicted an over- estimation with a lag of 6-hours. L10V08 and L06V04 have equalled the IMD in terms of minimum MSLP though a lag of 6-hours exists with these experiments. However, MSLP error is found lowest with the LIOV08 experiment, shown in Fig. 3(b). The model has exhibited remarkable results as the MSLP error is less than 5 hPa up to 66 hours forecast. The average MSLP error for the entire simulation period is 3.4 hPa. Histogram plots for other experiments are not provided here.

Fig. 3. (a) MSLP time series from all the 3DVar experiments along with CNTL simulation and IMD best- track; (b) MSLP error (hPa) histogram from L10V08 experiment calculated against the IMD best-track. Fig. 3. (a) MSLP time series from all the 3DVar Fig. 2. Track error (km) histogram from L10V08 experiment calculated against the IMD best-track.

Similar to MSLP, the MSW time series is shown in Fig. 4(a). Some experiments have a time lead, whereas some have a time lag to attain their peak with underestimation or overestimation from the IMD. The MSW error (Fig. 4(b)) up to 66 hours forecast time is found lowest in the L10V08 experiment with an average of 2.8 m/s for the entire simulation hours.

Fig. 4. (a) MSW time series from all the 3DVar error is found 45.7 km, whereas the average minimum sea level pressure (maximum sustained wind) error is found ave’ A kDa £9 7 tasla\ Anta eee tA: hE ALORA.

We use GEMS level 1C data at 0545 UTC 14 March 2021 after thinning and cloud screening. To prevent potential correlated errors of high-reso observation, simple thinning is applied, pixel every 100 km. For cloud screening, pixels that cloud fraction is more than 0.2 are removed. ution satellite selecting one After cloud screening, the number of observations d 1575 to 436 (Fig. 1). For this study, we channels (310.0 nm—310.8 nm with 0.2 nm interval and 311.0 nm—326.0 nm with 1.0 nm interval) and will extend ecreased from use just a few to the entire channels of GEMS in further study. Fig. 1. GEMS level 1C radiances at 310 nm on 0545 UTC 14 March 2021 with thinning and before (left) and after (right) cloud screening.

Figure 2 demonstrates the maximum DFS of each GEMS channel. The maximum DFS value is 0.9746 at 326.0 nm, and the minimum is 0.0003 at 312.0 nm. The value of DFS decreases from 310.0 nm, reaching a minimum at 312.0 nm, and then gradually increases. It turns out that the channels of 310.0—313.0 nm have much less information for control variables than those of 314.0— 326.0 nm. It may be due to lower absorption of troposphere trace gases (e.g. Chemical control variables: O03, NO2, SO2, and HCHO) over 310.0-313.0 nm. Fig. 2. The maximum DFS obtained by GEMS level 1C radiance assimilation. 326.0 nm. It may be due to lower absorption of

Fig. 3. A horizontal distribution of DFS obtained by GEMS level 1C radiance assimilation for (a) 312.0 nm and (b) 326.0 nm.

Table 1. Optimized tuning parameters — amplitude (0), decorrelation length (L) and time scale (t) for different experiments.

Fig. 1. Scatter plots of the ensemble mean spread and RMSE for (a) soil temperature (ST), (b) temperature at 1000 hPa (T1000), (c) soil moisture (SM), and (d) water vapor mixing ratio at 1000 hPa (Q1000). We investigate the ensemble spread and error using the scatter plot (Fig. 1). It shows that the experiment using the optimized tuning parameters generally increase ensemble spreads in all variables. As for the soil variables, the diurnally-varying tuning parameters (e.g., STP2 and SMP2) help to improve the desired linear relationship between the ensemble spread and error. On the other hand, when they propagate to the atmospheric model, the single (i.e., daytime only) tuning parameters (e.g., STP1 and SMP1) are more effective to increase the ensemble spread. The daytime amplitude and time scale, which show larger values in STP1 and SMP1, respectively, can generate larger random forcing.

Fig. 2. The analysis increment (colored contours; positive in red, negative in blue, and zero in gray) and the background error against GFS analysis (shaded) for temperature (in K) in (a) CTRL and (b) STP1, and for water vapor mixing ratio (in g kg") in (c) CTRL and (d) SMPI1. Results are averaged from 850 hPa to 1000 hPa. variable perturbations to the atmosphere, through the sensible and latent heat fluxes, etc., is less effective than the single (i.e., daytime only) tuning parameter. variable perturbations to the atmosphere, through the

Fig. 1. Track of 120 typhoons approached through Japan during 2006-2019.

thresholds of 5, 10, 20 mm over the different regions of Figure | represents the tracks of the 126 typhoons that landfalled/passed near Japan during 2006-2019. It shows the typhoons approaching Japan mostly form in the region of 5N—-15N and move initially northwestward and, after recurving, northeastward to the southern coast of Japan. The tracks indicated that the typhoons mostly landfalled in the Pacific Ocean side of Japan during 2006-2019 and are more in numbers compared to those landfalled in the Sea of Japan side. Figure 2 shows the probabilities of the spell durations of precipitation exceeding various thresholds of 5, 10, 20 mm over the different regions of ea of Japan side. Figure 2 shows the probabilities of the

Fig. 1. Lightning density in 2009-2020 (stroke*km?* year!, log scale). The numbers are associated with proposed area division. The white circle marks Yakutsk city. The average lightning density decreased to the north: the linear fit was —0.0033 (with 95% confidence bounds: from —0.0035 to —0.0031) and R* = 0.93. Also, the strong density relation to orography can be noted (Fig. 1). The average lightning density was high (> 2 times) on the western slopes of the Verkhoyansk ridge (along Lena and Aldan rivers) and between Momskiy and Cherskiy ridges (northeast of Yakutia). Aldan rivers) and between Momskiy and Cherskiy ridges

Fig. 2. Annual total lightning density in the parts according to the division indicated on Fig. 1 (stroke*km?*year"', log scale). different origin of storms that formed under the influence of air mass transfer from Okhotsk Sea.

The coordinate of the WRF-3DVar is 38.1°N and 98.9°E. For assimilation data author used AMSU-A satellite radiance data (from ATOVS d ataset) and for upper air we have used data from NCAR archives. Model forecast runs occur at 0000, 0600, 1200, daily. The first experiment, Control (CN1 Global Forecast System (GFS) as initia and 1800 UTC TL), used NCEP and boundary conditions. And as observation data author used BMD Table 1. WRF model configure. where J(x) is the current state at (x) to the background state (x) which is summed and squared and also to the observations (y’) in which the inverse of the error covariance matrices is weighted to misfits. The forward observation operator H maps model state to observation space. An observation error (R) and the covariance matrices of background error (B) are Gaussian in nature. The WRF-Var analysis and WRF model configuration, along with the parameterization schemes used in the present study, are summarized in Table 1. The first experiment, Control (CNTL), used NCEP Global Forecast System (GFS) as initial and boundary conditions. Two hailstorm events were simulated and then assimilated. The physical schemes of the model were the same in all three experiments as outlined in Table 1. physical schemes of the model were the same in all three

Table 2. Root mean square errors (RMSE) for rainfall, surface temperature and relative humidity for 27" February 2019 at Dhaka station. As it is found that the assimilation gives better agreement with the observational data, author used the assimilation technique to simulate the hailstorm event over Bangladesh which occurs on 27" February 2019 to investigate the associative atmospheric conditions responsible for the occurrence of hailstorms as sea level pressure, rainfall, accumulated hail (mm), reflectivity (dBZ), vertical profile of wind, vertical profile of CAPE, hail mixing ratio, graupel mixing ratio etc. and from the analyses it found that over the location where hailstorm occurred specially over Dhaka, Faridpur, Natore and Sirajganj these parameters were intense at the time of hail. As it is found that the assimilation gives better

Fig. 1. Comparison of accumulated hail amount between WREDA and CNTL run from the model for 27" February 2019. As case study, high-resolution numerical simulation o the 27 February 2019, severe hailstorm event has been conducted using mesoscale model WRF ARW and 3DVAR data assimilation system over Bangladesh. The model was having been run with 02 nested domain of 09 km and 03 km horizontal resolutions using the appro- priate physics schemes. For improvement of extreme hai event,the radiance data assimilation and without assimi- ation (CNTL run) simulation of WRF was performed. By comparing the RMSE of the assimilation and simulation without assimilation, and the simulated data with the BMD observed data, it is found that the hail simulation with assimilation give better results than the CNTL simulation. The CNTL run cannot capture the hail ocation over Dhaka, but WRFDA can capture the exact ocation of the 27" February 2019 hailstorm event. The model simulated weather elements and microphysical parameters have been found in close agreement with the BMD observation during the hail event. In terms of other weather variables WRFDA performs better than without assimilation (CNTL run). As case study, high-resolution numerical simulation of

Table 1. CMIP5 models (underlined if no RCP4.5). FIGURES AND TABLES

Table 2. Minimum central pressure (unit in hPa). REFERENCES Table 3. Additional sensitivity experiment (unit in hPa)

Fig. 1. Statistical Post-processing (SPP) procedure of probabilistic and single deterministic precipitation forecasts. In this study, we aim to produce well-calibrated 7-day accumulated precipitation forecasts through statistical Post-Processing (SPP), including probabilistic and single value (often called “deterministic”) forecasts. How to produce these two calibrated forecast products? Figure 1 shows the SPP procedure for precipitation forecasts. First, analog post-processing (AP; Hamill and Whitaker, 2006; Hamill et al., 2015) is used to get historical forecast analogs that most resemble the current ensemble forecast. That is, we search for the 20 past dates where the forecast is most similar to today’s forecast. Then we use the 20 observations corresponding to the forecast analogs as a calibrated forecast ensemble (called AP ensemble or APE hereafter). Probability-Matched mean (PM; Ebert, 2001) is then applied on APE to derive a single deterministic forecast (called APPM) on one hand, and frequency counting is used to obtain probabilistic forecasts on the other.

Fig. 2. Frequency distribution of rain rate using all the samples from the 22 years data set in this study. The blue, red and black curves are results from the AP ensemble mean, APPM, and corresponding observations. Figure 2 shows that the under-forecasting for heavy rainfall and over-forecasting for light rainfall, which results from the averaging process in the ensemble mean, is mitigated in APPM: its frequency distribution of rain rate is close to the truth. In other words, the rainfall pattern of the APPM displays more detailed fine scale variability that is present in nature, as compared with that of the AP ensemble mean.

Fig. 3. Tukey boxplots of correlation for summer, winter, and Mei-yu season. The blue and red boxplots are the results from the raw and calibrated forecasts respectively. For the two adjacent boxplots, the left is week 1 and the right is week 2 forecast. These results suggest that the precipitation forecasts downscaled and calibrated using the AP and PM methods substantially improve forecast quality and value.

A total of 79 cm, 172 cm, and 406 cm rainfall recorded at CCPO, MACPO & HACPO locations respectively during the Indian Summer Monsoon (ISM) 2019. During this season 12, 58 and 48 BB events are observed. It is noted that before-BB a sufficient amount of precipitation of 2.2 cm (CCPO), 24.9 cm (MACPO) and 41.7 cm (HACPO) observed shown in Fig. 1. Maximum number of BBs occur at MACPO but the longest duration of events seen at HACPO location. The total rainfall contributed by BB in before-during period is highest (512 mm) at HACPO. At the time of BB events, 6 mm and 125 mm were noted at CCPO and MACPO. Fig. 1. Rainfall (mm) associated with before-BB and during-BB events at CCPO, MACPO and HACPO locations. CONCLUSION

Further, to analyze the scenario of the DSD before-BB and during-BB we classify the rain rate into four different categories i.e, R1 (0.1 < R <1 mmh!), R2(1<R< 5 mmlr!), R3 (5<R<10mmb) and R4(R>10 mmh’!). The DSD spectra for all rain rates at each location are presented in Fig. 3. It is observed that at CCPO majority of rainfall occur in R1 rain rate category for during-BB, whereas before-BB the R1 and R2 rain rate contributes to total rainfall for these periods, however at MACPO and HACPO all rain rate categories contribute to surface rainfall. At each location number concentration of small drops (0.1—0.6mm) observed higher before BB, whereas large drops (greater than 3mm) are much more observable during BB. Fig. 3. Mean raindrop concentration (N(D), m? mm) versus drop diameter (D, mm) for before-BB (a—c) and During-BB (d—f) at CCPO, MACPO and HACPO near ground level for rain rate R1 to R4.

Fig. 2. Contoured frequency altitude diagrams (CFAD) of fall velocity (ms!) BBB (a-c) and DBB (d-f) events at CCPO, MACPO and HACPO.

Fig. 1. SLP at (a) 0000 UTC on 9 August (b) 1800 UTC on 10 August 2011.

Figure 2 shows strong circulation at 950 hPa over West-Bengal and adjoining Bangladesh with a trough extended to east-southeastwards. There was_ strong southwesterly flow of winds from the Bay of Bengal through Bangladesh coast. The circulation was found to extend vertically to 500 hPa (not shown in the figure) at 0000 UTC on 9 August 2011 and continued to 10 August. The maximum wind speed at 950 was 18.75 ms? (67.5 km/hr) at Rajshahi at 0000 UTC on 9 August. Fig. 2. Simulated wind vectors and circulations at 950 hPa on 9-10 August 2011.

Fig. 4. Simulated 6 hourly reflectivity over Bangladesh on 9-10 August 2011. Fig. 5. Simulated 6 hourly CAPE over Bangladesh on 9— 10 August 2011.

Fig. 3. Simulated 3 hourly rainfall in Bangladesh on 9 August 2011.

the centre of the depression. Figure 6(b) shows the cloud water mixing ratio at stations. At Khepupara, maximum cloud water mixing ratio of more than 600 mgm* was at 0000 UTC on 10 August 2011. Maximum cloud water mixing ratio of 600 mgm* was found at 0900 UTC on 9 August at Rajshahi. Cloud water mixing took place between 950 and 300 hPa and was maximum with greater depth near the centre of the depression.

Those previous studies indicated that there is nteraction between CENS and the DCR. However, the Fig. 1. Mean of ERAS surface wind (vector) and meridional wind (shaded) in boreal winter from 2000- 2020. Red box indicates CENS indices and the purple box indicates Java Island as the study area. and it is further defined as the cross equatorial northerly surge (CENS) by Hattori et al. (2011). Previous studies (Wu et al., 2007; Hattori et al., 2011; Yulihastin et al., 2020) agreed that CENS is associated to heavy rainfall events over Java Island. It is also associated to the early morning peak over the northwestern coast of Java Island that causes severe flood (Yulihastin et al., 2020).

Fig. 2. Subset area for western, central, and eastern Java which are indicated by dashed box r.1, 1.2, and r.3 respectively.

Fig. 3. Hovmoller diagram of DCR over Java Island in the three different regions; r.1 ((a) & (d)), r.2 ((b) & (e)), r.3 ((c) & (f)); during nCENS days (top) and CENS days (bottom). JS is for Java Sea, IO is for the Indian Ocean, and area within two dashed black line is the island.

Fig. 1. Variation between MCRA and Tbs on 19 Dec 2017.

Fig. 2. Bias between MCRA, GSMaP, IMERG with actual rainfall on 19 Dec 2017. Fig. 3. Rainfall map by MCRA on 04 Nov 2017 (1200 UTC).

observed smaller between 1 mm and 3 mm. The hourly bias between the estimates by MCRA, GSMaP and IMERG with the actual rainfall is found to be smaller between —2 mm and 2 mm but the smallest is observed on 19 December 2017 (Fig. 2). Meanwhile, RMSE is observed to be consistent and closer to each other at some points where the values are closer to 0 mm and the highest is observed at 8 mm on 20" Jan. The hourly bias between the estimates by MCRA with the satellite’s derived products show smaller values in between —1 mm and 0 mm. The RMSE is generally observed smaller between 1 mm and 3 mm. between —1 mm and 0 mm. The RMSE is generally

are observed similar with the actual Himawari’s enhance image of IR on channel 13 (Fig. 4). Fig. 4. Himawari’s actual image of IR on channel 13 on 04 Nov 2017 (1200 UTC). REFERENCES CONCLUSION The algorithm at the threshold rainfall value of 0.8 mm

Since Lorenz (1963), chaotic solutions have been a focal point for several decades, yielding the statement of “weather is chaotic”. In fact, depending on the relative strength of heating, the L63 model also produces non- chaotic solutions such as steady-state and limit cycle solutions (e.g., Figure 1 of Shen et al., 2021b). Given a model configuration, only one-type of solution appears, referred to as monostability, as shown in Fig. 2 (left). Fig. 2. Monostability illustrated by the chaotic solution of the L63 model (left), and multistability by coexisting chaotic and steady-state solutions of the GLM (right).

tiny change in an IC will eventually lead to a very different time evolution for a solution. Fig. 1. An illustration of SDIC. Control and parallel runs were performed using the same model and the same model parameters. The only difference is the inclusion of an initial tiny perturbation within the parallel run. Two runs initially produced almost the same result for t € [0, 25], as shown with the red curve. This feature is called CDIC. During longer time integrations, the appearance of two curves (in red and blue) indicates significant differences (i.e., the “rapid divergence”) of solutions for the two runs. Such a feature is then called SDIC.

To illustrate SDIC, Lorenz (1993) applied the activity of skiing (left in Fig. 3) and developed an idealized skiing model for revealing the sensitive dependence of time- varying paths on starting points (middle in Fig. 3). The eft panel for skiing may indicate monostability when slopes are steep everywhere. By comparison, the right panel for kayaking is used to illustrate multistability. In the photo, both strong currents and a stagnant area outlined with a white box) can be identified, suggesting both instability and local stability. As a result, when two kayaks move along strong currents, their paths display SDIC. On the other hand, when two kayaks move into the stagnant area, they become trapped. The features of kayaking reveal the nature of multistability. Fig. 3. Skiing as used to reveal SDIC and monostability (left and middle, Lorenz, 1993), and kayaking as used to indicate multistability (right, Copyright: ©Carol- stock.adobe.com).

Fig. 1. SWCRF and LWCREF as a function of size resolved AOD at TOA. Blue and red dot represents the small and large particles respectively.

Fig. 2. Same as Fig. | but at the surface. IGP region is known for the elevated aerosol loading during both monsoon and pre-monsoon period, with increased concentration of coarse mode absorbing aerosols (Jose et a et al., 2012). The TOA and surface ., 2021; Mangla et al., 2020; Srivastava decreased SWCRF and LWCREF at the for large particle could be probably due to the aerosol induced heating inside the clouds resulting in the evaporation of cloud droplets. IGP region is known for the elevated aerosol loading

Table 1. Frequency distribution of model and sounding derived relative humidity (surface — 500 hPa). 1umidity do not match on a correct interval with large Table 1 shows frequency distribution of model and sounding derived relative humidity for points of Eastern Siberia and Russian Far East (number of cases is 13194).

Table 2. Frequency distribution of model and sounding derived index Jice (number of cases is 13194). Values of Jice frequency are decreased as it moves away from correct icing category. Slight overestimation

Table 3. Contingency table. Table 3 shows that Peirce Skill Score is 0.83; Heidke

Fig. 2. Speciation result of bulk soil samples. Fig. 1. Total metal concentrations of soil samples.

Fig. 3. Metal concentrations in different size fractions.

Fig. 4. Geo-Accumulation Index of soil samples.

Fig. 5. Contamination Factor of soil samples. the metals; hence the risk cannot be overestimated as in the case of considering total metal concentration for risk estimation.

heterotrophic respiration. With each experimental plot, soil solutions at 10 cm depth in triplicate were regularly sampled each month using self-made zero-tension lysimeters in growing season (May to October) each year (Xu et al., 2021). Two composed mineral soils (0-10 cm depth) within each experimental plot were randomly sampled in both non-growing and growing seasons using a soil auger (3.1 cm in diameter) each year, for measuring soil properties such as water content, pH, extractable carbon and nitrogen pools, and microbial biomass carbon and nitrogen. Fig. 1. Apparatus for measuring soil respiration and its components in the field.

Fig. 1. The geographical locations of TCCON stations in North America. The color legend represents 5D values for each station, also showing in Fig. 5.

RESULTS Fig. 2. The monthly mean 6D of TCCON stations from 2011 to 2021 in North America, color-coded according to their latitudes.

Comparison of correlation between humidity and 6D Fig. 4. The comparison of correlation between water columns and 6D for all TCCON stations in North America, each line is from the average value of the original observations as shown in Fig. 3, color-coded according to their latitudes. Figure 3 shows the correlations between humidity and 6D for all six TCCON stations in North America. All six graphs show positive correlations between the water columns and dD values. This correlation can be explained by the Rayleigh process, which relates to the isotopologues changes during the evaporation and condensation processes, or the air mixing process, which relates to the mixing of air masses with different 6D.

Fig. 3. The correlations between water columns and 6D for TCCON stations in North America. TCCON stations, which is shown as light blue line in the graph. The lowest 5D value is from Eureka station, which is shown as orange line in the graph. The 6D values from all stations fluctuates as the season changes. In summer, the 5D values reached the highest point in each year while they dropped and reached the lowest points when the season turns to winter. Compared to the 5D values in all stations, the 6D value tends to be higher when the latitudes of the stations are higher.

positive. When the humidity keeps constant, t 6D are lower as the latitude —500%o in all six stations, w subduction of air from higher hich may be re depleted in 6D. According to Fig. 5 shown be values are decreasing as the America. The slope of the linear correlation is calculated atitude increas becomes higher. When the humidity tends to zero, the values of 6D tend t atmosphere that is dry and he values o o be around ated to the ow, the dD es in North as —1.46, ignoring the station with the lowest latitude which is Pasadena. Therefore, when the latitude increases by | degree, the 5D value will decrease by about 1.46%o. Fig. 5. The correlation between 6D and latitude for the six TCCON stations in North America. The mean values of OD are adapted from Fig. 4 between 1000 and 2000 ppm for XH20. America. The slope of the linear correlation is calculated CONCLUSIONS

Table 1. ML models performance evaluation for SSL inflow prediction. Fig. 1. Taylor diagram for ML models performance evaluation.

Fig. 2. Sediment flushing curves are drawn for (a) SW and (b) NREB sites.

Fig. 1. Flow chart of SVR method, Wavelets and EnKF approach. This study used daily RWT and related minimum, maximum, and mean AT data from January 1, 1989, to January 1, 2004. Based on the sensitivity analysis, the maximum AT was found to be a highly sensitive variable; hence we have taken the maximum AT as an input ence we have taken the maximum AT as an input maximum AT was found to be a highly sensitive variable; parameter for prediction. Results demonstrated improved forecasting accuracy of the hybrid WT-SVR model (NSE: 0.87 and RMSE: 0.92) compared to the standalone SVR model (NSE: 0.84 and RMSE: 0.99). It was discovered that RWT prediction ML models with a monthly time scale performed better than those with a daily time scale. To develop a continuous intelligence system, the EnKF data assimilation approach is applied to improvise the performance of the WT-SVR model in each simulation step. The WT-SVR-EnKF model’s findings with merged data at different simulation steps demonstrate better outcomes from simulation-1 (January 1, 2001, to January 1, 2002) (NSE: 0.890 and RMSE: 0.740) to simulation-2 (January 1, 2002, to January 1, 2003) (NSE: 0.91 and RMSE: 0.73). These outcomes reveal that the WT-SVR- EnKF model performs well with blended data in RWT prediction. Further improvement of RWT prediction can be investigated by using other types of WT coupled with the SVR algorithm. Because of a lack of a complete dataset, flow discharge was not included in this work. After training the models on local data, creating a global model using the federated ML approach is the future scope. maximum, and mean AT data from January 1, 1989, to This study used daily RWT and related minimum,

Fig. 1. The methodological framework of the proposed reservoir inflow forecasting model.

Table 1. The specification of Soil Moisture and Ocean Salinity. Fig. 2. Soil Moisture and Ocean Salinity product over the catchment area. Where 7 is the soil porosity (dimensionless) or ratio of non-solid volume to the total volume of soil, L is the soil depth (m) specified in the parameter setting, (1-qi) is Gamma in RRI algorithm. Once F is found, the infiltration rate f (m/h) can be obtained by: solution based on approximate physical theory to include soil moisture content and soil characteristics, is one advantage of the RRI Model. The cumulative depth of infiltrated water F (m) is:

Fig. 1. Study area (left) and landuse (right) of upper Brantas River basin in East Java, Indonesia. solution based on approximate physical theory to include

(see Fig. 2) tends to give higher value. In order to obtain the value of soil water content, the time variation of infiltration is extracted using the Green-Ampt equation. This amount is needed for soil moisture value validation. Fig. 3. RRI Model result showing observed discharge (black) and simulated runoff (red).

Fig. 4. RRI Model result showing the spatial simulated runoff (left) and inundation (right). Fig. 5. Green-Ampt cumulative depth of infiltrated water.

Fig. 6. Downscaling result and the original SMOS product.

Fig. 1. Schematic diagram of the water balance model developed for the INS-IBWT. Blue boxes highlight groundwater related components of the RF scheme.

obtained for The vertica strategy is s Fig. 2. Pareto-approximate water transfer strategies the RF (blue) and NRF (green) formulations. axis represent a variable of interest at donor D), recipient (R), and system (S) level, arranged to have the preferred value at the top. Each Pareto-approximate hown by a line crossing the vertical axis at the ocations denoting its consequence for the variable of interest. M EF — minimum environmental flows, RF: recharge formulation, NRF: no-recharge formulation. strategies also performs better at system as well as individual basin level in maintaining minimum environmental flows. Notably, the mean annual deficits are reduced upto 50% in the strategies allowing artificial groundwater recharge. However, we did not find any systematic reduction in the required transfer volumes. Although the median of the annual transfer volume is lower for RF, both RF and NRF entail similar range of annual transfers volumes. Thus, the additional water storage provided by artificial recharge in the recipient basin is successful in reducing the system-level water stress significantly but not reducing reliance on the water transfers. basin is successful in reducing the system-level water

Fig. 3. Seoul National University Gwanak campus (Seo et al., 2012; Tak et al., 2017). University Gwanak campus. Seoul National University,

In this study, the balanced urban water cycle was defined in terms of sustainability, safety, equity, and efficiency according to the theoretical studies. Following Fig. 2 illustrates the four states of the balanced urban water cycle. Sustainability, equity, safety, and efficiency were defined as the securing of water circulation management system adapted to future climate change, the absence of inequity in the movement of water spatially, less damage to property and life caused by water, and the securing of the right amount of clear water to anyone, respectively. Fig. 2. Definition of balanced urban water cycle. absence of inequity in the movement of water spatially, The process of quantifying the four indicators and

Fig. 1. Box plots of 8!8O and SD at different surface water sampling areas Content. Samples of 5D and 8'8O from the Upper Indus River Basin are around the LMWL and the GMWL (Fig. 2) indicating that the surface water is of meteoric origin. However, the data points are on and below the LMWL, indicating that the surface water was affected by evaporation. Saline lake samples are below the GMWL suggested that these samples have been more intensely evaporated. suggested that these samples have been more intensely indicating that the surface water was affected by The d-excess values of mainstream ranged from 6.1% to 15.6%o with a mean value of 11.0%o. The values of d-excess for tributaries and glaciers ranged from 12.3% to 19.4%o with a mean of 15.3%, 14.7%o to 20.6%o with a mean value of 17.2%o respectively. Our finding is also consistent with prior records (Pande et al., 2000; Rai et al., 2016; Karim and Veizer, 2002). All of the d-excess values of the glacier/glacial meltwater and river water recharged by glacial/snow meltwater are large, which indicates the characteristics of the d values of the glacier and glacial/snow meltwater. Large and small-scale glaciers are developed at the relatively high altitude of the Upper Indus River and are the source of the tributaries and The d-excess values of mainstream ranged from 6.1%

streams. Glacial/snow meltwater recharges the river, so the d value of the sampling points in these places is the largest. This finding indicates that the surface water of the Upper Indus River Basin originated from glacial/ snow but also atmospheric precipitation. Fig. 2. Plots of 65D and 5!8O of Mainstream, Tributaries, Glaciers, and Groundwater. The groundwater samples of 5!8O and 5D of the Upper Indus River Basin were around and below the GMWL and LMWL, indicating that the groundwater is derived from glacial/snow meltwater and meteoric water. The value of d-excess of groundwater ranged from 10.3%o to 2.0%0 with a mean value of 11.2%o were positive, indicating that p g the groundwater may have originated from glacia Snow meltwater. the distributions of the surface water and groundwater samples are relatively concentrated, indicating that they may have a common source.

Fig. 3. Shows the variation in 5D values of river water along the flow path of the Indus River. This figure were plotted from the Indus river samples of (Karim & Veizer, 2002, black circle) and samples of this study (red circles).

Table 1. Calibrated model parameters. CQOF stands for the catchment overland flow runoff coefficient. Zing, and rir: ES SS RO RAD TE) pee eee EE oe re EE ae” rete CE ee ee CE | until the simulated runoff matched the observed values.

The observed hydro-meteorological data required for the RR model were o btained from the Sri Lanka Departments of Meteorology and Irrigation. The catchment boundaries and river network were generated using Shuttle Radar Topography Mission (SRTM) data, with a resolution of 1 arcsecond (Reuter et al., 2007). Land use and land cover d Department of Sri Lanka. ata were obtained from Survey Fig. 1. Location of rainfall and river gauges in the basin. Methodology The RR model was calibrated for the period 2000- 2003 and validated for the period 2004—2007 using the input data described above. The catchment area was divided into two sub-catchments for model para- meterization based on the locations of the river gauge stations. The validated model was used to simulate two major 2017. 7 flood peak discharge and flood volume, we conducted [o evaluate the impacts of and-use change on t flood events that occurred in May 2003 and May ne flood simulations under two conditions, with and without land-use changes within the catchment area during t 1999-2017 period (Jayapadma, 2019). The RR mod performance was evaluated according to the Nas Sutcliffe efficiency (NSE) and he el _ volume ratio (Vr), calculated using Equations (1) and (2) below. NSE value varies from 0 to 1 where NSE of 1 agreement of observed values wit would suggest a perfect h the simulated values.

The NSE values (> 0.6) suggested that the model accurately simulated runoff in the Gin River catchment area and was thus suitable for flood simulation in this region. Figs. 2(a) and 3(a) compare the observed and simulated discharges at an upstream gauge station (Thawalama) during model calibration and validation, respectively, and Figs. 2(b) and 3(b) compare those at a downstream gauge (Baddegama) station (Jayapadma, 2019). The model performed well in simulating both low and peak flows at the Baddegama gauge station, whereas peak flows at the Thawalama gauge station were underestimated, perhaps due to the effects of average catchment parameter approximation, observed data uncertainties, and coarser input data resolution (DHI, 2014a; Horritt and Bates, 2002; Moseley and McKerchar, 1993). Fig. 2. Comparison of river discharge during model calibration.

Fig. 3. Comparison of river discharge during model validation.

Fig. 1. Marshyangdi Basin. Finally, trend analysis was performed by Mann Kendall’s and Sen’s slope (Mann, 1945; Sens’s Slope, 1968) method.

Table 1. Ecological Implications of selected hydrologic indices (m?/s/year). Note: HA: Hydrologic Alteration; S: Significant; S: Significant, NS: Insignificant; M: Moderate

degree of hydrologic alteration. CONCLUSION The hydrologic alteration was low with an overall mean of 30% in the Marshyangdi basin. However, in the future, alteration in a flow regime due to climatic and anthropogenic activities could impact river health. The mean annual streamflow in the basin is 222 m?/s which is insignificantly increasing with a trend of 0.6 m*/s/y. Hydrologic alterations (HA) in the basin vary from low to moderate within five groups. An increase in the median flow values during the period of monsoon and _ post- monsoon values (Fig. 2) and consequent statistically significant increasing trend in the 30-day and 90-day maximum values indicate the possibility of flood in the basin in the future (Table 1). Poff and Zimmerman (2010) had also mentioned that almost all published research found negative ecological changes to alteration in the flow variability. NE re ee SSE eee pee ei Satee ng ee Sai asec sa ste ete oz.

Table 1. Classification of drought intensity based on SRI value and for DNBC.

Fig. 1. Location of the Han River basin with sub-basin ID number, elevation, and stream network. METHODOLOGY Standardized Runoff Index (SRI) was proposed by Shukla and Wood (2008) to characterize and monitor hydrological drought using streamflow data. The calculations of SRI are quite similar to Standardized Precipitation Index (SPI), in which log-normal distribution was fitted to streamflow data and cumulative probabilities of marginal distribution were transformed into standard normal variate with zero mean and unity standard deviation. In this study, daily data of streamflow was aggregated into monthly timescale and SRI was calculated at 3-month time scale.

Fig. 2. Different classes of hydrological drought assessed through DNBC during 2001—2013 and compared with the observed SRI in watershed a) #1003, (b) #1010, (c) #1012, (d) #1018. where P = {J,+ 1 =i/Y;} is the probability of hydrological drought classes in next month given hydrological drought classes in current month J;. Pi is the transition probability matrix for hydrological classes in next month condition on the hydrological drought class in current month.

Fig. 3. DNBC in the classification of hydrological drought classes. (a) accuracy and (b) precision. The accuracy estimated for class 1 varied from 50% to 63%; for Class 2 from 46% to 70%; for Class 3 from 59% to 67%; for Class 4 from 45% to 67%; and for Class 5 from 33% to 50%, while the precision estimated for class 1 varied from 45% to 58%, for class 2 it varied from 55% to 85%, for class 3 it varied from 46 % to 78%, for class 4 it varied from 24% to 50% and for class 5 it varied from 33% to 71%.

Fig. 3. Simulating 169 climate change cases (13 precipitation * 13 temperature). Each climate change case is used to generate drought time series.

Fig. 2. Spatial maps of 215 catchments presented in the figure. Catchments are classified based on Aridity Index.

Fig. 4. Classification and Regression Trees (CART) for a catchment: Critical threshold of precipitation and temperature are obtained from classification and regression tree for SPEI and PDSI.

Table 1. Statistical characteristics of SGSs. Results The best-fit probability distribution function of each D-day low flow sequence was identified for each SGS in the Kucuk Menderes River basins (Table 2). No probability distribution function was identified for D = 1, 7, 14 and 30-day low flows in D06A001 and E06A001, and for D = 1, 7, 14, 30 and 90-day low flows in D06A012. Why a probability distribution function does not exist is because of the total number of non-zero D-day low flows. We applied frequency analysis on D-day flow

Fig. 1. Location of Kucuk Menderes River Basin in Turkey and layout of SGSs.

Table 2. Best-fit probability distribution functions of D- day low flows. Parameters of the selected probability distribution

Fig. 2. LFDF curves for SGS D06A001. Parameters of the selected probability distribution functions were determined and used in calculating the D- day low flows in every individual SGS for return periods of 2, 5, 10, 25, 50 and 100 years. SGS DO6A001 was selected to demonstrate the LFDF curves (Fig. 2). LFDF curves decrease quite fast towards very low discharges and even to zero values other than the 273-day LFDF curve. The D-day low flow time series have mostly zero values. It means that the river basin is prone to get dry. It is also important to notice that no curves were obtained for D = 1 and 7 days. The curves extend up to T = 5 years for D = 14 days, and to T = 10 years for D = 30 and 90 days. The duration-frequency curve is complete up to T = 100 years only for D = 273 days. This is because of the number of zero flows in each D-day low flow sequence. 100 years only for D = 273 days. This is because of the CONCLUSION Frequency analysis shows that Log-Pearson Type III probability distribution function fits the D-day low flows

"RMSE: root mean square error. >RMSPE: root mean squared percentage error. “NSE: Nash-Sutcliffe efficiency. 4N: number of samples Table 1. The statistical model validation.

Fig. 2. Conceptual diagram of ecological model. Fig. 1. The computational range of the model. The dotted line illustrates the outer boundaries. The red point is the model validation point. Model validation The model was validated by comparing the model outputs with the field measurements data (Saito et al., 2020). Three statistical variables were used to validate this model, namely root mean square error (RMSE), root mean squared percentage error (RMSPE), and Nash- Sutcliffe efficiency (NSE). The validation target position is near the estuary of the Kuma River, shown as the red point in Fig. 1.

And we validated the fitting results of the hydrodynamic module of this model, using salinity, temperature, and ot as objects, and the fitting results of the ecological module, using DIC, TA, and pCO as objects. The validation results are shown in Table 1, and all these values are in an acceptable range, so the model fit of this model was regarded as good. Fig. 3. Field measurement results (a), and simulation results (b) of ot distribution (top) and pCOz distribution (bottom), during the strong stratification period (August 2, 2019). objects. The validation results are shown in Table 1, and

We inspected the vertical pCO2 dynamics during the flood season. The inspection cross section is shown as the red line in Fig. 4, starting at the estuary of the Kuma River. Fig. 4. Inspection cross section of vertical pCO2 dynamics.

Fig. 5. Vertical distribution at 12:00 on July 7, 2018. CONCLUSIONS

Fig. 6. Vertical distribution at 12:00 on July 27, 2018. From the distribution of salinity at 12:00 on July 7, it can be seen that the freshwater spreads on the seawater surface. In addition, the distribution of pCO2 shows that freshwater with high pCO2 and seawater with low pCO: are stratified. Therefore, it was confirmed that the freshwater concentrates on the seawater surface immediately after flooding, which promotes CO2 emissions into the atmosphere.

Fig. 1. Calculation domain. METHODS This study uses the hydrodynamic model and lower- trophic ecosystem model by Delft3D, which are the same as Tadokoro and Yano (2019). Calculation domain is a combination of the Ariake Sea and the Yatsushiro Sea (Fig. 1). The horizontal grid is a rectangular grid with a Cartesian frame of reference and the vertical grid uses o coordinate system, where the number of o-layers is defined as 10 layers with 5% x 3, 10% x 4, 15% x 3 from surface to bottom. Open boundaries are located outside the inlet to the Ariake Sea and on the line connecting ABSTRACT: The Ariake Sea is considered as an estuary or inner bay-type region of freshwater influence (ROFD) located in Kyushu Island, western Japan, which is one of the most important shallow sea for fisheries and seaweed production. In recent years, extreme flooding events occurred frequently in Japan. Heavy rainfall may cause a large amount of freshwater inflow, which will have serious impact on the marine environment and ecosystem in the bay. For proposing the adaptation, this study conducted different scenarios of effluent patterns to evaluate the development of hypoxia by coupling the three-dimensional hydrodynamic model (Delft3D-FLOW) and the ecosystem model (Delft3D-WAQ) in the bay. The results indicate that the duration and peak discharge of effluent can significantly affect the development of hypoxia in the bottom layer.

Table 1. Calculation results of hypoxia for six effluent patterns. Fig. 3. Observation and calculation results of DO at Sta.B3 in 2020.

Fig. 4. Relationship of normalized total effluent volume and duration of hypoxia at the bottom layer as well as the peak discharge in each case.

Fig. 2. Conceptual diagram of lower-trophic ecosystem model.

Fig. 5. Isopleth of DO at Sta.B3 in six effluent patterns. The effluent period is extracted when the most severe hypoxia occurred. Figure 5 shows the isopleth of DO at Sta.B3 in six effluent patterns (Fig. 1). The duration of continuous hypoxia lasted for more than two months in case 1. Among the three months in summer (from Jun. to Aug.), the hypoxic water mass at the beginning of August was larger than the others and it distributed vertically in the entire water column. REFERENCES Figure 5 shows the isopleth of DO at Sta.B3 in six

Fig. 1. the Ramkhamhaeng polder, Bangkok, Thailand. Bangkok, the capital city of Thailand, is located in the Shao Phraya River basin. Chao Phraya River split down he aadddAleae at thse. cei Tha eraind larval ta Whatiyaan +O nO

Fig. 2. Real-time flood forecasting system framework. In the central part of Bangkok, the Ramkhamheng area is representative of urbanization. Bangkok Metropolitan Administration (BMA) has a department of drainage and sewerage (DDS) working on flood problems in Bangkok. Several studies from DDS have shown that the drainage capacity of the existing system is still not perfect enough. The Ramkhamhaeng polder has an area of 11.444 sq. km. The north boundary is the Saen Saeb canal. South is the East Railway. The east is Khlong Tooyor and Khlong Hua Mak, and the west is the Saen Saeb Canal and Khmer Canal. The surface ground level is between 0.00 to +0.50 m.MSL, some areas are below sea level, such as Ramkhamhaeng University. Real-time flood forecasting system divided into three main parts. The first part is the rainfall input data process. This research measured rainfall data with 10 minutes of rainfall from a radar station near this study area. The forecasted rainfall data with 30 minutes of rainfall duration collected from the forecasting system. This research assigned the TITAN software to track and forecast the storm and analyze movement by determining the linear movement of the storm groups, calculating the overlapping area between the storm boundary and the study area to calculate the amount of precipitation entering the study area in advance. The optimal success of CSI (Critical Success Index) of 66% was obtained at a forecasted time of 30 min as shown in Fig. 3. Real-time flood forecasting system divided into three

The schematic of the real-time flood forecasting system was designed based on the optimum rainfall duration, which contained 40 minutes including 10 minutes measured and 30 minutes forecasted rainfall durations. Moreover, it should be included 10 minutes for the computer processing time. Fig. 4. The schematic of the real-time flood forecasting system. duration, which contained 40 minutes including 10 The second part of the real-time flood forecasting system is the main process for simulating flood maps. The PCSWMM software assigned to simulate as a hydraulic model which consist of flows in pipe and overland flows. A database centers the system for storing input and output data. The database system is responsible for keeping and importing the preparation data via a script command from the python and automatically running through the window task scheduler. system was designed based on the optimum rainfall

Fig. 3. Critical success index, CSI (%), of rainfall forecasting results.

Table 1. Status of formulation of BCP and other emergency response plans. Meanwhile, 29 companies have defined priority recovery operations and their target recovery time in their plans, which must be defined in the BCP, and 19% of the companies have plans that can be considered BCPs (Table 2).

Table 3. Degree of external dependence on management resources. Furthermore, when asked what resources related to

Table 4. Difficulty in collecting information and making decisions due to external dependence on management resources. Meanwhile, when asked what kind of cooperation with stakeholders is envisioned in BCPs and other disaster response plans, and what kind of cooperation with stakeholders was used in past disasters and other emergencies, there were no responses that envisioned or cooperated with stakeholders related to social infra- structure. Meanwhile, when asked what kind of cooperation with stakeholders is envisioned in BCPs and other disaster

Fig. 1. Empirical and estimated distribution functions of extreme rainfalls X; (mm) for different durations for London CS station.

Fig. 2. Probability plots between the observed and estimated short-duration ERs of station #7-Ottawa CDA for the calibration period 1961-1990. Yellow circle markers represent the empirical probability plots while red cross markers and dash lines show the at-site frequency analysis plots. The gray lines and boxplots show the derived distributions of short-duration ERs based on the proposed approach and data from 21 GCMs. Fig. 1. Boxplots comparing different indices (CC, MADr, and RMSEr) computed based on the difference between observed and adjusted data or observed and regional data for (a) calibration and (b) validation periods.

84 stations located in Ontario, Canada. The jackknife technique was used to represent the ungaged site condition at 15 chosen sites among them. In general, based on different indices, the results produced by the proposed approach highly agree with the observed data. More specifically, the low values of RMSEr and MADr and high values of CC indicate the accuracy of the new method in the estimation of short-duration extreme design rainfalls for an ungaged location in the context of a changing climate. Fig. 1. Boxplots comparing different indices (CC, MADr,

LUCC spatial analysis, change and future scenario models The study is carried out in two water bodies environmentally differentiated: the reservoir of As Conchas in the province of Orense, Northwest Spain and the Albufera of Valencia, a Mediterranean coastal lagoon, in the province of Valencia, East Spain (Fig. 1). Fig. 1. Study areas location: 1 As Conchas Reservoir, 2 Albufera de Valencia Lagoon. Information on land use/cover changes (LUCC) has Spatial and temporal analysis of LUCC dynamics should imply the application of quantitative models to establish the projection of historical land covers to future scenarios (Belda-Carrasco et al., 2019) and to foresee how LUCC affect ecosystem services performance, including those services related to water regulation in the interaction of LUCC with soil functions. Land Use/Cover Changes related to water quality and quantity processes

Fig. 2. Percentage of land cover for the years 1990, 2006. 2018 and the 2030 and 2050 scenarios in As Conchas. urban covers (U10) and the agricultural mosaic rainfed (A21) the ones that grow the most, with clear reduction in the rest of the agricultural ones.

Fig. 3. Percentage of land cover for the years 1990, 2006, 2018 and the 2030 and 2050 scenarios in Albufera of Valencia.

Fig. 4. LUCC-Soils water retention capacity in As Conchas Reserovir. Fig. 5. LUCC-Soils water retention capacity in Albufera of Valencia. CONCLUSIONS LUCC analysis linked to the geospatial model for the generation of future scenarios allows relating water parameters with historic trends and future evolution of land cover. The LUCC scenarios obtained with the model

To overcome this problem, K-water (2016) proposed a unique disaggregation method to disaggregate the annual drought frequency inflow into a monthly scale. Figure 1 below illustrates the KDM process. The solid curve indicates the annual frequency inflow estimated by the frequency analysis. The dotted curve is the sum of monthly frequency inflows estimated by the monthly frequency analysis. In order to satisfy the additivity condition, annual drought inflow is distributed according to the same sum of the combination of monthly frequency inflows. For example, the return period of 100-year in the solid curve can be disaggregated according to the ratio of the return period 16.5-year monthly inflows. By using KDM, the annual sum of the drought inflow

is shown in Fig. 2. With the negative inflow issue related to the August 2011 flood event, Method 2 was ruled out. Fig. 2. Comparison of return period 20-year drought inflows in Boryeong Dam (m?3/s). The results all followed the same pattern: high during

Fig. 1. Trend in was measured hydro-meteorological parameters. The variation in the streamflow was found sensitive to both climate and anthropogenic activities in the watershed. The contribution of the climate change in the variation of streamflow remains primary factor accounting 70% to 76% in the total change. However, the anthropogenic actives contribute 24% to 30% in the tota change in the watershed streamflow (Table 1). The fractional contribution of anthropogenic activities observed gradually increase in every year in the pos change period. This increase in the impact o anthropogenic activities may result in the hydrologica disaster in the watershed. The variation in the streamflow was found sensitive to

Learning from the Central Sulawesi disastrous event in 2018 (Cipta et al., 2020), we investigate the interior of Jakarta Basin to better understand the subsurface structure that may have hidden under the thick sediment. This technique is able to prove the existence of Cisadane, Ciliwung and Kali Bekasi faults that firstly have been identified by Moechtar & Poedjoprajitno (2003). The Citarik fault which was identified by Sidarto (2008), could be a southwest continuation of the Kali Bekasi fault, has also been captured by the HVSR method. This method was able to detect the Baribis fault as well (Fig. 1). Fig. 1. (a) West-east and (b) south-north cross section showing a quite deep of Jakarta basin, reaching > 500 m (Vs of basement = 800 m/s) and suspected faults crossing in and around the basin. (c) Location of the cross sections.

Fig. 2. (a) and (b) are S-N cross sections showing Baribis- Kendeng Fault and other suspected fault south of the Baribis- Kendeng. (c) soft sediment (Vs = 150 m/s) thickness inferred from HVSR inversion is compared with (d) subsidence rate in Semarang city (Kuehn et al., 2009). Location of cross-sections and data points (black dots) are showing in (c).

Table 1. ABM Parameter Values. We ran the model 10 times for each of the scenarios, with 150 steps each run. The averages of the results were then computed, and the results for the no vaccination scenario were validated with the COVID-19 data from March 21 to May 21, 2021 (60 days).

We implemented the agent-based model using Mesa (Project Mesa Team, 2021) and MesaGeo (Corvince, 2018), two Python application libraries that are both publicly available (see Fig. 1). We then used this implementation to simulate the four vaccination scenarios of interest in this study. Fig. 1. GIS visualization of agent-based model.

Fig. 2. Validation of the death (left) and recovery (right) age distributions for the no vaccination scenario. Furthermore, our model also demonstrates the bias of COVID-19 against the elderly which was already pointed out by Bongolan et al. (2021) as early as 2020. In our simulations, we observed that older individuals (60+) generally die when hit by the disease and the younger ones recover. Ground-truthed with March 21 to May 21, 2021 (60 days) COVID-19 data, this captured age dynamics validates with actual observation (see Fig. 2).

Table 1. Variation of soil properties with magnitude of seismic shear strain according to Ishihara (1996). 'Department of Science and Technology — Philippine Institute of Volcanology and Seismology (DOST — PHIVOLCS), Quezon Cit 1101, Philippines. "Department of Education (DEPED), Pasig City 1605, Philippines. 3Department of Science and Technology (DOST), Taguig City 1631, Philippines. Microtremor survey is conducted within the confines of public schools situated along the coast of Manila Bay in Greater Metro Manila. Quaternary alluvium classified with saturated and unconsolidated material of varying

Table 2. Liquefaction Potential Index and descriptions.

Table 3. Strain range with SPT and SDS LPIs.

Table 4. Strain range with SPT and SDS LPIs.

thickness is distributed along the near ground surface along the coastal lowlands. Figure 1 shows liquefaction susceptibility increasing towards Manila Bay where soil thickness and groundwater saturation are believed to increase as well. Fig. 1. Liquefaction hazard map of Greater Metro Manila showing all experimental sites and relative features.

Meycauayan City West Central Elementary School (MCWCS), Bulacan Fig. 3. Distribution of predominant period, relative ampli- tude, and seismic shear strain in MCWCS, Bulacan. Also shown are the locations of SPT and SDS boreholes.

Fig. 4. Distribution of predominant period, relative ampli- tude, and seismic shear strain in TNCHS. Also shown are the locations of SDS boreholes.

Bagbag Elementary School II & Bagbag National High School (BES), Cavite Fig. 5. Distribution of predominant period, relative ampli- tude, and seismic shear strain in BES. Also shown are the locations of SDS boreholes.

Fig. 6. Non-linear curve fitting of y (x 10°) vs. LPIs from SDS and SPT.

Fig. 1. Disease transmission through respiratory droplets.

Figure 3 clearly shows the movement droplets near the mouth region of a human. At initial time step (t = 0.5 sec), the droplet particle emerged out from a human travel over a longer distance in the supplement of velocity. The larger (100 um) and medium-sized (30 um) particles lose their inertia and settle rapidly under the influence of gravity. But in the case of smaller droplets, requires larger time to lose its inertia and remains in the air, travels over a longer distance. The distribution of droplets near the hip area is also reported in this study. Figure 4 indicates that the medium sized particle concentration (30 um) accumulates more within 2-4 meters. In case of smaller particles, it needs more time to gain its weight and settle down. The 1 um particle concentration is most often located at a distance of 5 metres. Similarly, the concentration of 0.1 um droplet particles is more noticeable at a distance of 6 meters. Fig. 3. Spatial spread of respiratory droplet near mouth region.

Fig. 4. Spatial spread of respiratory droplet near hip region.

Sedimented droplets are most often occupied near the foot area. The spatial spread of droplets near foot region is shown in Fig. 5. The large droplet (100 um) settles down quickly at a distance of 2 meters and the time taken is also less than 5 seconds. The distance covered by medium-sized particles (30 um) is 2-3 m. Further, the smaller droplets take longer time to settle down. The concentration level of these droplets is low in this region.

Fig. 6. Spatial spread of respiratory droplet in XY direction at Initial time step. Fig. 7. Spatial spread of respiratory droplet in XY direction at Final time step.

In the LBM method, respiratory droplets are regarded as a particle that undergoes collision and propagation in a discrete domain space. The above study considered XY plane for simulating the spread of droplets. At an initial stage, the droplet particle is concentrated in the centre region. Concurrently, it undergoes collision and stream- ing process. The above process creates an unstable condition and make the particles move to the neighbour- ing cells. At the final stage, the particle loses its velocity and starts to settle down. Figures 6 and 7 indicate the spatial spread of respiratory droplet in XY direction at Initial time step and final step.

Average Wet Season (May-Oct) Low Sea-Level Rise Scenario

Fig. 1. Average wet season (May—Oct) SLR scenario and its impact on roadway vulnerability. Groundwater map was adopted from Obeysekera et al. (2019). (A) Low SLR scenario (B) High SLR scenario (C) Detailed view of High SLR scenario. Blue line indicates concrete pavement. Table 1. Vulnerability classification based on ground- water condition.

be utilized for establishing the cumulative event rainfall— duration (ED) thresholds. As for the rainfall data, the rainfall event variables, such as cumulative event rainfall, rainfall duration, and three-day antecedent rainfall, are split by over 12 hours non-rainfall based on hourly rainfall data using Python programming. station, the Automated Agriculture Only one gauging Observing System (AAOS), is available for soil moisture observation data around this study area. Noah LSM requires the atmospheric forcing data, including air temperature, specific humidity, surface pressure, direction, rainfall, downward so downward longwave radiation. Among them, ERA-Land hourly data is used for downward downward longwave radiation data. 1 Automated Synoptic Observing System (ASOS) has al required meteorological data, so we wind speed, wind ar radiation, and solar radiation and The only Chuncheon can derive one soi moisture data with four layers from running Noah LSM. Fig. 1. Geomorphic setting of study area.

We designed the Landslide Early Warning System (LEWS) as in Fig. 2. We defined a normal level as when it starts in the summer season and under the Cumulative event rainfall-duration thresholds. We proposed watch levels using the ED thresholds of the local area, which were presented in our previous study (Lee et al., 2021). Then, we made warning levels and severe warning levels using the Bayesian results of landslide probability based on rainfall and soil moisture data. Fig. 2. Landslide Early Warning System (LEWS).

Fig. 3. Comparison of soil moisture variation in the Noah LSM model and observation data in the landslide triggering year 2013.

landslide scarp, in cyan, is facing south to SE along its extent. The profile also shows that the boundary between the zones of depletion and accumulation is marked by a local uplift less than 5 m in height. The landslide toe farther downslope, which signifies the lower boundary of the zone of accumulation, is represented by blue-cyan color which reflects the distinct change in aspect and SRR. Fig. 1. The PC image composition with section line, corresponding elevation profile and color slice, and interpretation of the surface morphology of the landslide in Parasanon. White hachured lines represent field- mapped landslide scarps.

morphological characteristics through color contrasts with the surrounding areas. Such analysis can be performed prior to field surveys for targeted landslide mapping. Fig. 2. The PC image composition with section line, corresponding elevation profile, and an oblique view of the landslide in Manghulyawon. morphological characteristics through color contrasts

Figure 18 – uploaded by Yi Ting Thean

Related Figures (268)

Connect with 287M+ leading minds in your field