A pathway for multi-stage cyclone-induced hazard tracking—case study for Yaas

A Very Severe Cyclonic Storm ‘Yaas’ developed over the Bay of Bengal (BoB) on 23 May 2021 and crossed over the Odisha coast on 26 May with maximum sustained wind speed of 75 kts. Herein, a pathway has been developed and exemplified for ‘Yaas’ through three-stage cyclone-induced hazard tracking. Days before the cyclone formation, cyclone genesis potential parameter, sea surface temperature (SST) (> 30 °C) and tropical cyclone heat potential (anomaly of 40–80 kJ/cm2) indicated a strong possibility of cyclogenesis in the BoB. A Lagrangian advection model used for its track prediction with 24-h lead-time provided an accuracy of ~ 19 km and ~ 6 h in its landfall location and time. Further, intensity prediction was done using numerical weather prediction model. Geostationary satellites, INSAT-3D/3DR, were used to visualize cyclone structure. Passing of cyclone had its reverbarations in oceans, which are observed in SST drop of ~ 3 °C, salinity and density increase by ~ 1 psu and ~ 2 kg/m3, respectively. During the period, 23–26 May 2021, the Ekman suction velocity and chlorophyll concentration were found significantly high at ~ 5 m/day and > 0.5 mg/m3, respectively. Forecast of storm surge was found to be between 3.5 and 4 m at coastal locations. Significant wave height was found to be 5.5–9.2 m. The coastal inundation forecast for 24 May 2021 provided its quantitative maximum inland extent. Finally, loss of the crop, fishery and forest areas by strong winds and inundation/ingress of saline water associated with storm surge were examined using SAR and optical data.


Introduction
Over the last few decades, the space-based observations have revolutionized the study of severe weather systems such as tropical cyclones (TC) with the wealth of observations from pre-cyclogenesis stage to the aftermath of their formation. A detailed study of TCs with conventional data is insurmountable due to sparsely distributed measurements of fewer parameters.
The North Indian Ocean (NIO) surrounding the Indian peninsula is a potentially energetic region for the development of TCs. There is an estimated loss of about 2% of India's GDP, and about 12% of the central government revenue occurs due to cyclones hitting the country's coast every year (National Cyclone Risk Mitigation Project, https:// ncrmp. gov. in/ cyclo nes-their-impact-in-india/). The ruinous impact of tropical cyclones (TC) is much more pronounced in the east coast of India, as compared to the west coast. However, the recent studies show an increase in the frequency of TCs in the west coast of India (Murakami et al. 2017;Deshpande et al. 2021). The cyclones crossing over eastern coastline of India originate over the Bay of Bengal (BoB) or migrate from the west pacific basin crossing the Myanmar coast. The distribution of tropical storms/ cyclones in BoB is bimodal with its peak during the post-monsoon months (September-December) and pre-monsoon months (April-June). The seasonal and intra-seasonal variability of temperature, moisture and low-level potential vorticity in both ocean and atmosphere plays a key role in genesis, intensification and movement of tropical cyclones (Fan et al. 2020).
TCs are one of the most violent manifestation of the weather that cause devastation in the sea and coastal regions from storm surge, flooding, extreme winds, torrential rains, thunderstorms, lightening, etc. As discussed above, each year tremendous economic losses as well as loss of lives are caused by the TCs in all parts of the world. To reduce these losses, the accurate prediction of TC formation, evaluation, intensity and precise track is necessary. TC formation involves interaction among various processes of different scales ranging from meso to synoptic scales. Cyclones form over oceans, and therefore, the atmospheric and oceanic conditions prior to their formation and intensification are crucial for their role in determining cyclone structure and energetics (Emanuel 2004), and also for their effects on various oceanic processes and ecosystem. The strong forcing of TC produces two kinds of dynamic responses in the upper ocean, namely turbulent mixing and upwelling, both of which can induce a decrease of sea surface temperature (Pan et al. 2018). The analysis of oceanic parameters such as sea surface temperature (SST) and tropical cyclone heat potential (TCHP) do provide an insight into cyclone formation and intensification (Gray 1975;Cione et al. 2003;Lin et al. 2013). Upwelling in the oceans driven by the wind stress curl is the most important process for density changes in the thermocline through divergence of upper layer transport (Dickey et al. 1998), and that causes nutrients from below the mixed layer uplifted to the surface, and thereby setting conducive environment for enhancing the biological productivity. This can be monitored through changes in the chlorophyll concentration from satellite optical measurements. The intense winds associated with cyclone perturb the ocean surface and pile up the water, and thus generate storm surges that on arriving to the coast can cause coastal inundation. After landfall of a cyclone, there is a widespread damage to property and, agriculture and forest areas due to ingress of the saline water, uprooting of the trees and lodging of the standing crops. There is a strong need for assessing the losses to provide relief and rehabilitation of the affected people.
The present study focuses on TC Yaas formed in the North Indian Ocean. It was developed as a low-pressure system over the east-central BoB in the morning (0830 IST/0300 UTC) of 22 May 2021. Under favourable environmental conditions, it intensified into a depression over the east-central BoB in the noon (1130 IST/0600 UTC) of 23 May 2021. It moved northwestwards and intensified into a deep depression (DD) over the east-central BoB in the midnight (2330 IST/1800 UTC) of 23 May 2021. It turned into the cyclonic storm in the early morning (0530 IST/0000 UTC) of 24 May 2021 and named by Indian Meteorology Department (IMD) as TC 'Yaas'. Thereafter, it moved north-northwestwards and attained a peak intensity of 75 kts while laying centred over northwest BoB ~ 30 km east of Dhamra Port, Odisha, during early morning ( A large number of studies documented cyclones' formation and intensification, role of prior and post-formation atmospheric and oceanic conditions, their effect on marine ecosystem, coastal inundation and damages of vegetation including agriculture and properties on their landfall, but none of them addressed all these inter-woven processes in a consolidated study. This is required because of possible feedback and cascading effect of various processes, which, however, fall in very diverse domains of meteorology, hydrology, agriculture science, physical and biological oceanography, etc.
A timely forecast of cyclogenesis and subsequent intensity and track prediction may help the relief agencies to take timely action to evacuate affected areas to save lives. The change in biological productivity of the oceans prior to post the event may help fisherman to take appropriate action to exploit the conditions for better fish catch in suitable locations. The distribution of precipitation, storm surge and associated coastal inundation helps providing relief to the people stuck by the disaster. The loss of agriculture area and forest cover help making assessment for providing compensation to the victims, especially to settle crop insurance claims and taking actions for restoring forests. This paper addresses all these issues in a well-organized study associated with the TC 'Yaas'.
The paper is divided in the following sections: Section 2 provides details of the variety of data products used, Sect. 3 provides methodology and Sect. 4 contains results and discussion. Because the present study provides an insight into various diverse aspects of the cyclone Yaas, this section is divided into many sub-sections for more organized structuring. Finally, Sect. 5 provides the conclusion.

Data used
The study utilizes variety of data products including a large number of satellite radiance and derived geophysical parameters, analysis and forecast from numerical weather prediction models and other static data as well as in situ data. The geospatial data from satellites, analysis, forecast and multi-sensor merged product, static and ground data are summarized in Table 1.

Methodology
This section provides details of three-stage tracking of cyclone-induced hazard. Three stages represent prediction of pre-onset features in stage-1, cyclone track-intensity prediction after onset, monitoring the cyclone propagation and forecasting pre-landfall ocean state in stage-2 and post-landfall impact assessment in stage-3, respectively. An overview of this multi-stage tracking is given in Fig. 1.

Stage-1
Scatterometer-derived wind pattern matching-based technique has been developed by Kishtawal (2011, 2013) that effectively predicts tropical cyclogenesis (TCG) 1-4 days prior to cyclone formation in North Indian Oceans. However, due to non-availability of coverage over Yaas by any other wide-swath scatterometer, this technique was not applicable to Yaas. Hence, the numerical weather prediction (NWP) model-derived forecast fields were examined to investigate the signatures of cyclogenesis. Kotal et al. (2009) proposed a cyclone genesis potential parameter (GPP) for Indian seas that give an early indication of the intensification of an existing system. GPP utilizes vorticity at 850 hPa, mid-tropospheric relative humidity and instability and the inverse of vertical wind shear averaged over an area of 2.5° radius from the centre of an existing system. These meteorological parameters are taken from GFS reanalysis data.

Stage-2
After the formation of tropical cyclone over the Indian Ocean region, two different models were used for its track and intensity prediction. These models are: (a) Lagrangian advection cyclone track prediction model (LAGAM) (Singh et al. 2012a, b) Hurricane Weather Research and Forecasting (HWRF), which is a coupled land-ocean-atmosphere model customized for Indian Ocean region. Starting from 23 May 2021, the LAGAM was used for the track prediction, and HWRF model was used for intensity prediction. HWRF is computationally expensive model, and it is not feasible to run it at multiple times in a day. Therefore, the real-time intensity prediction is generated once in a day at 00 UTC using HWRF model. However, SAC track prediction model (is limited to track prediction only) is very fast and use less computational resources so it can be run for all synoptic hours (00, 06, 12 and 18 UTC) for real-time track prediction. LAGAM requires the high-resolution 0.5° × 0.5° forecasted atmospheric wind and temperature from GFS model and the initial position of the cyclone from Joint Typhoon Warning Centre (JTWC) advisories. Using this model, cyclone track predictions were generated up to 96 h with 6-hourly intervals. In the first step, the steering flow was computed for every 6-h forecast intervals up to 96 h, using the analysis as well as forecast wind fields data at 21 pressure levels (100-1000 mb) by the weighted average scheme. The weight for each level was assigned by estimating the potential vorticity (PV) as discussed by Hoover and Morgan (2006). Then, cyclonic vortex was removed using a synthetic cyclone to erase existing cyclonic wind fields present in the steering flow to achieve the residual steering current. The synthetic cyclone was constructed by using the vorticity equation as discussed by Chan and Williams (1987). The steering flow obtained after removing the cyclonic vortex was used in Lagrangian advection model to forecast cyclone track. The computation for the trajectory of cyclone (or cyclone track) was initiated by interpolating the steering wind from model grid points to the initial location of cyclone (Brand et al. 2018). For cyclone intensity prediction, numerical weather prediction model, HWRF was utilized. The HWRF model is a primitive-equation, non-hydrostatic and coupled atmosphere-ocean model with an atmospheric component that employs the non-hydrostatic mesoscale model (NMM) dynamic core of the WRF model (WRF-NMM), with a parent and two-nest domains (Biswas et al. 2018). The HWRF version 4.0a (HWRF4.0a) model, used in the present study, is available from the NCEP website (www. dtcen ter. org). The model was evaluated for its default configuration for NIO tropical cyclones. In the present study, the results of intensity prediction of TC Yaas for three different initial conditions at 00 UTC on 23, 24 and 25 May 2021 are presented in Sect. 4.1. The GFS analysis and forecasts (0.25° × 0.25°) were used to provide the initial and lateral boundary conditions. All the three experiments were performed in the uncoupled ocean mode and without the data assimilation module.
Each of India's two geostationary meteorological satellites, viz. INSAT-3D and INSAT-3DR, provides coverage every 30-min over India and surrounding regions. During the active cyclones in NIO, INSAT-3DR satellite was used in rapid scan mode, and it provided observations over TC in every 4-min intervals. Both INSAT-3D/3DR provided a unique opportunity to observe the development and intensification of tropical cyclones. One of the essential ingredients to the intensification of TCs is vigorous convection with associated latent-heat release through condensation processes. Identifying and quantifying active convection in geostationary satellite images from INSAT Imager can be potentially useful for the prediction of cyclone intensification. Cyclone centric products from INSAT-3D and INSAT 3DR imager channels were generated to visualize cyclone structure, viz. its centre, brightness temperature of the eye and its surrounding environment. Radius of maximum winds (R max ) is one of the most critical parameters that determine the tropical cyclone wind structure often required to assess the area of damage after the landfall. In infrared satellite images, the central region of cyclone is mostly obscured by high-level cirrus clouds that cause difficulty in precise identification of R max . High-resolution observations from visible (central frequency: 0.65 μm) and SWIR (central frequency: 3.82 μm) channels provide unique opportunity to identify the cloud structure near the centre of cyclone. A procedure has been developed to produce cyclone centric products from each half-hourly image of INSAT-3D satellite (Jaiswal and Kishtawal 2016). These images are very useful to study the structural changes in the core of a tropical cyclone.
The tropical cyclone heat potential (TCHP) was calculated using the temperature and salinity profile simulated from data assimilative Modified version of Modular Ocean Model-3 (MVMOM3, Mallick et al. 2019Mallick et al. , 2020. Extreme waves and surge-induced coastal inundation associated with Yaas was critical as it landed in low-lying areas of Bengal. State-of-the-art coupled ADCIRC + SWAN model was used for the estimation of storm surge, details of which are available from Luettich and Westrink (2004) and Manadal et al. (2020). The storm surge simulations were carried out with a spatial resolution of 1-km near land and 60 km in the open ocean. ADCIRC model was primarily forced by tides, winds and wind waves. During a cyclone, wind stresses are the dominant forcing. ETOPO-2 global topography was used as bathymetry. ADCIRC model is coupled with a coastal wind wave model (SWAN) for capturing wave-induced setup (Dietrich et al. 2011). In the present study, NCMRWF analysed 6-hourly winds at 0.25° spatial resolution were used as primary forcing, and ADCIRC + SWAN model was integrated for 10 days to accomplish model spin up.
Further, satellite data assimilative WAVEWATCH III (WW3) model (Tolman 2009) was used for wave forecasting. It uses National Centre for Medium-Range Weather Forecast (NCMRWF) analysed winds as the forcing field. The wave model runs daily at the 0000 UTC. SARAL/AltiKa, Jason3 and Sentinel measured significant wave heights are being assimilated into this model, and 5 days' forecast (Seemanth and Suchandra 2018) was generated. Coastal inundation forecast was generated in coupled ADCIRC + SWAN model using 100-m fine resolution (near coast) mesh. The total water elevation as a combined effect of storm surge, wind waves and tides was simulated using NCMRWF 0.25° resolution forecast winds, which are interpolated on to the model grid using flux conservation scheme. This methodology has already been verified for several past cyclones (Mandal et. al. 2019(Mandal et. al. , 2020. Apart from storm surge, the low-lying inland regions were also flooded by intense precipitation. A hydrological model, SACHYDRO, developed by us (pl. see, https:// vedas. sac. gov. in/ vedas/ downl oads/ ertd/ Hydro logy) was utilized here to derive flood inundation probability. This model was a deterministic, distributed and unsteady flow model that simulates the hydrological fluxes at regional scale (5 × 5 km grid). To solve different hydrological processes in these modules, different datasets such as meteorological parameters from WRF model forecast along with remote sensing data products for land surface variables (e.g. rainfall and snow cover) were utilized.

Stage-3
Flooded area over land is detectable using remote sensing measurements from different satellite borne instruments such as synthetic aperture radar (SAR), radar scatterometer, microwave radiometer and visible band imagers. In the present study, we have utilized SAR and microwave radiometer observations to demarcate flooded areas. Sentinel-1, a C-band SAR, was used to map inundation regions at 10-m resolution.
Microwave radiometer-derived brightness temperature polarization ratio (PR), also known as Microwave Polarization Difference Index (MPDI), is commonly used to study soil moisture, surface inundation and vegetation characteristics (Gupta et al. 2019;Njoku et al. 2003;Zheng et al. 2016). In this study, MPDI from AMSR-2 is used to retrieve flood index maps. AMSR-2-derived MPDI at 36.5 GHz helps to observe large-scale flooding with very high temporal resolution (1-2 days) and wide swath. SAR data, with repeat pass of ~ 12 days and limited swath, can miss the onset and progression of flood; whereas, passive microwave radiometer provides global coverage every 2 days making it extremely useful in analysing regional-scale floods with very high repetivity. Thus, MPDI at 36.5 GHz from AMSR-2 at 0.1° spatial resolution was utilized for estimating the inundated areas over a larger area (swath of ~ 1450 km) with 1-2 days repetivity.
Using the complementary information from SAR and optical remote sensing data, a joint approach has been adopted to map the affected areas of agricultural patches. The available Sentinel-1A (S1A) SAR, Sentinel-2-derived NDVI and its maximum value composites during pre-and post-cyclone periods were used in the present study. The NDVI data were used to demarcate active agricultural area using pre-defined threshold and forest mask. Several metrics, such as probability density functions (PDFs) of S1A-derived VH and VV backscatter before and after cyclone periods, differences in VH and VV backscatter as well as NDVI between pre-and post-cyclone, were computed. All these helped in mapping probable affected area and three damage classes of low, moderate and high affected areas for agriculture using equi-quantile segregation of the 95 percentile values. In addition, pre-and post-cyclone SAR polarimetric parameters such as entropy, anisotropy and alpha using S1A SLC and Eigen value/Eigen vector decomposition (Pottier and Cloude 1996) were also used for areas of probable damage with three classes for agriculture.
For analysing changes in forest areas, phenological pattern and NDVI difference (NDVI May -NDVI June ) were used. At the first stage, phenological pattern of representative areas was analysed for year 2019 using NDVI time series derived from Sentinel-2 datasets. It was noticed that there is increasing trend in forest phenology (NDVI) from May to June. This information was used to calculate NDVI difference, such that the areas where there is change/ loss in the forest vegetation, it will be indicated by positive value in NDVI difference image (NDVI Precylcone -NDVI Postcyclone ). All other areas of natural change in vegetation phenology will indicate negative values in NDVI difference image. The difference (NDVI Precylcone -NDVI Postcyclone ) therefore attributes to vegetation changes due to cyclone only and not due to natural changes in vegetation phenology. The damage due to cyclone was assessed over coastal forest especially mangroves.

Results and discussion
The present study focuses on prediction of cyclogenesis, track, intensity and landfall of a tropical cyclone 'Yaas'. It also examines environmental conditions prior to the formation of cyclone, during its presence and aftermath, and its effects on marine ecosystem and coastal processes. Finally, it analyses flooding of the coastal areas and its other destructive effects on agriculture and forest areas. The study makes strides in many inter-woven processes from diverse areas of natural sciences, viz. meteorology, hydrology, physical and biological oceanography, agriculture and forestry.

Prediction of pre-onset cyclonic feature
The most important pre-onset feature, such as cyclogenesis potential, has been predicted in stage-1. The tropical cyclogenesis (TCG) is one of the important features that have been quantified using GPP. The threshold value of GPP to detect cyclogenesis was determined by maximizing the probability of detection (POD) and minimizing the false alarm ratio (FAR) . A GPP threshold value of '2' was used to identify the TCG in GFS forecasted fields (Singh et al. 2013). The maximum GPP values for all analysed forecasts for Yaas for 00 and 12 UTC on 20, 21 and 22 May 2021 were found to be 2.18, 3.80, 3.98, 5.86, 8.09 and 9.32, respectively. An increasing GPP value indicates strengthening of the conditions for cyclogenesis with time. IMD is an operational agency for tropical cyclone forecast in the NIO. Prediction of cyclogenesis of TC Yass was provided on 00 UTC of 18 May 2021 by the IMD using the GPP approach. IMD utilizes the IMD GFS data in generating the GPP values and generate the graphical output showing that the potential cyclogenesis zone is also provided and updated for all synoptic hours (IMD report, 2021).

Post-onset prediction of cyclone track and intensity
The real-time predicted tracks of cyclone Yaas by LAGAM for different initial times are shown with the observed track of IMD in Fig. 2. In Table 2, the track forecast error for all the forecasts was computed w.r.t. IMD observed track positions. IMD reported the landfall point of cyclone Yaas near Balasore, Odisha (landfall point: 21.35 N, 86.95 E) between 0500 and 0600 UTC (IMD, Report). The landfall position error for all the forecasts with different initial time is computed and reported in Table 3. For 24-h lead prediction, the landfall position and time errors from LAGAM for Yaas are found to be ~ 19 km and ~ 6 h, respectively.
The intensity of simulated cyclone from different experiments was evaluated in terms of the maximum sustained wind speed (MSW). The real-time intensity prediction of cyclone Yaas for all initial conditions (00 UTC 23-25 May 2021) along with IMD intensity estimates is provided in Fig. 3. The results suggested that the model overestimated the intensity for all forecast hours w.r.t. initial condition of 00 UTC 23 May 2021. This might be due to the reason that cyclone was in the depression stage on 23 May 2021 which might have resulted into less accurate wind structure simulations. However, the forecasts initiated on 00 UTC 24 and 25 May 2021 were relatively close to the observed intensity by IMD. Further, forecasting of ship avoidance region was carried out by demarcating > 34-knot wind field from HWRF model-derived wind fields (figure not included).

Cyclone eye and wind structure
Detection and propagation of cyclone eye, structural changes and rainfall have been monitored using INSAT 3D and 3DR imageries. A sample product generated from INSAT on 0830 UTC 25 May 2021 is presented in Fig. 4a, b, c, d, e and f. TC Yaas was continuously observed by the half-hourly acquisition of INSAT-3D satellite. Using the half-hourly TIR imageries, the centre location of cyclone was estimated by centre determination algorithm given by Jaiswal and Kishtawal (2016). On 25 May 2021, Yaas further moved towards north-eastern coast of India and gained its intensity to very severe cyclonic category at 1200 UTC. The observation over TC Yaas using microwave sensor like scatterometer was also analysed. The ASCAT scatterometer on-board METOP-A satellite provided observations over the cyclone Yaas during 24 and 25 May. The scatterometer observed winds over the cyclone Yaas are shown in Fig. 4g and h. The winds show asymmetric cyclonic circulation with maximum wind speed as 24.04 and 26.16 m/s during 24 and 25 May, respectively.

Pre-landfall ocean state forecast
Pre-landfall ocean state forecast of storm surge, coastal inundation and its inundation probability have been made. At the time when deep depression over the east-central Bay of Bengal intensified into cyclonic storm 'Yaas' at 00 UTC of 24 May 2021, maximum simulated wave height found was ~ 5.5 m. As Yaas evolved to a severe and Very Severe Cyclonic Storm at 0000, 1200 and 1800 UTC on 25 May 2021, SWH increased to ~ 8.5 m, ~ 9 m and ~ 9.2 m, respectively (figure not shown) in the northern Bay of Bengal (BoB). At the time of coast crossing, about ~ 20-km south to Balasore, the maximum wave height was found to be ~ 5.5 m in the north BoB and ~ 2-2.5 m along the Orissa coast. Figure 5 shows the time-series evolution of model-simulated SWH at different coastal locations along the east coast of India during 'Yaas'. The model simulations were validated using in situ wave rider buoys and were found to be in agreement with observations. At landfall, the peak surge height forecast was ~ 3.5-4-m near Digha and Sagar Island areas (Fig. 6). The surge residual validation was carried out at Dhamra port (86.95°E, 20.78°N) using tide gauge 1 3 observation provided by INCOIS that showed good correlation coefficient of 0.9 between simulation and observation.
The coastal inundation forecast for 24 May 2021 (Fig. 7) showed maximum storm tide of ~ 4.5-m near Contai and Sagar Islands, ~ 3.5-m near Haldia and ~ 3.8-m near Sundarban areas, which resulted into maximum inland extent of coastal inundation of around 3.8-km near Sutahata (Haldia), 4.5-km near Kumbhigari, 1-km near Bakkhali and around 2.5-km near several places in Sundarban area. The summary of maximum storm tide, significant wave height and horizontal extent of coastal inundation are shown in Table 4. Inundation probability based on amount of surface runoff and topography slopes is scaled from 0 to 1. Figure 8 showed low-lying regions of Odisha and West Bengal, that were prone to cyclonic floods, were affected by cyclone Yaas during the period of 23-28 May 2021.

Changes in ocean physical and biological parameters
The spatial plots of the sea surface temperature (SST) from INSAT-3D (Gangwar and Thapliyal 2020) before, during and after Yaas are shown in Fig. 9. It can be seen that  prior to the cyclone formation on 20 May 2021, the ocean was very warm with temperature > 30 °C in the central BoB. Correspondingly, tropical cyclone heat potential (TCHP) anomaly (figure not shown) was highly positive (40-80 kJ/cm 2 ), making the oceanic conditions conducive for cyclogenesis, which resulted into formation of a deep depression at 16°N; 90.5°E on 23 May 2021. The TCHP anomalies are computed with respect to longterm mean (35 years) of TCHP simulated from same model. Spatial distribution of SST for 26 and 29 May (Fig. 9) revealed a ~ 2.5-3.0 °C reduction in temperature after cyclone had passed. Figure 10a and b showed the snapshots of 8-day averaged SSS (sea surface salinity) from SMAP for pre-(20 May) and post-cyclone (28 May) conditions. Large-scale, low salinity features (< 32 psu) in the central Bay of Bengal are noticeable in the SSS image for 20 May 2021. This low salinity resulted into strong surface stratification, which, in turn, led to surface heating due to high downwelling solar radiation. This is consistent with the SST image in the central Bay of Bengal that showed large-scale warming (~ 32 °C) in low salinity region. This further indicates conducive ocean surface conditions for the formation and sustainability of a cyclone in this region. Sea surface density (figure not shown) was also low prior to the cyclone formation as compared to the surrounding regions. After the passage of cyclone, surface became cool and saline (Chaudhuri et al. 2019) due to Ekman pumping associated with high cyclonic winds. There was a drastic change in the physical properties (i.e. SST, SSS and density) after the passage of cyclone. While there was a temperature drop of ~ 3 °C during 20-28 May 2021, the salinity increased by ~ 1 psu, due to which the density increased by ~ 2 kg/m 3 . Density changes were associated with both salinity and temperature changes. However, the causal physical parameter (SST or SSS) that contributed more to density changes was assessed by computing density changes in two ways, one by keeping SST fixed and another by SSS fixed (both corresponding to 20 May 2021). It was found that density changes (> 50%) resulted mainly due to cooling of SST, while the salinity changes contributed marginally (~ 15%) in the total density changes. Daily averaged TCHP anomalies over the cyclone track, starting 23 May 2021, were analysed. These anomalies were computed using long-term ocean model simulations (Mallick et al. 2020) and were extracted over the cyclone track and compared with the averaged cyclone intensity (maximum wind speed) for that particular day. Just before the cyclone arrives at the location, the TCHP anomaly for that location is large positive (62.7 kJ/cm 2 ), and it becomes maximum (101.8 kJ/cm 2 ) just before the maximum intensity of the cyclone (63.13 Knots) is observed. Once the cyclone passed, the TCHP anomalies became negative, indicating cooling down of the oceans due to wind mixing.
In order to quantify upwelling, wind stress curl and Ekman suction velocity were computed on a daily basis from 17 May 2021 to 9 June 2021 with given wind speed, wind-dependent drag coefficient and water and air density (Halpern 2002; Sun et al. 2010). For the same period, 8-day composite images of chlorophyll were obtained from GlobColour multi-sensor merged products. Further, 8-day composites of sea level anomaly (SLA) from AVIOS product and geostrophic current velocity components were generated and overlaid on SLA data to infer cyclonic and anti-cyclonic mesoscale eddies. Following Faghmous et al. (2015), eddies were detected using outermost closed contours of sea level anomalies. Using every 6-h position of cyclone track as provided in Fig. 2, cyclone translation speed was calculated as the average speed during the 6-h period from the 6-h track position and thermocline displacement due to the cyclonic wind stress following Price et al. (1994) and Walker et al. (2005). Analysis of SST chlorophyll and SLA were carried out for three 8-day periods, i.e. pre-cyclone period (17-24 May 2021), during cyclone period (25 May-1 June 2021) and post-cyclone period (2-9 June). However, the Ekman suction velocity was analysed during all these three periods using only 4-day averaged products. In this case, 4-day averages were considered appropriate because the cyclonic storm was active only for 4 days (23-26 May 2021), and 8-day averaged data might not be likely to pick up cyclone-induced wind stress curl. The pre-Yaas, during Yaas and post-Yaas Ekman suction velocities are shown in Fig. 11a and b. Ekman suction velocities in pre-Yaas and post-Yaas period were found significantly lower (< 1 m/day), while the period from 23 to 26 May 2021 is characterized by high Ekman suction velocities of about 5 m/day (Fig. 11a). The rotating storm system caused Ekman mass divergence and upwelling due to the wind stress curl. Thermocline displacement ranging from 7 to 24 m was observed along the track ( Table 5). The impact of Yaas cyclone in the 8-day averaged chlorophyll concentration (CC) images (Fig. 12) showed a region of high chlorophyll plume on the left side of the track in the coastal and offshore waters with CC > 0.5 mg/m 3 between 12° and 17°N (Fig. 12). To investigate the change in chlorophyll after the cyclonic event, data of precyclone and post-cyclone period were analysed by observing the difference (Fig. 12d, e and f). The chlorophyll maps for 17-24 May (Fig. 12d) and 2-9 June, clearly showed increase in chlorophyll concentration in the post-cyclone period. The average SST and chlorophyll in a rectangular box (shown in Fig. 12b) around the cyclone path for the  We also investigated OCM-2 single day passes over the Bay of Bengal in order to quantify the changes in chlorophyll concentration due to higher spatial resolution of OCM-2 (360 m). Cloud-free OCM-2 scenes of pre-cyclone (17 May) and post-cyclone (29 May) Yaas were selected, and chlorophyll was retrieved using OC-2 algorithm in Fig. 12g and h, respectively (O'Reilly et al. 1998(O'Reilly et al. , 2019. A patch of increased chlorophyll, occurred in coastal and offshore waters near Chilka lake, left of the cyclone track was observed in OCM-2 imagery on 29 May, 3 days after cyclone Yaas landfall. The average chlorophyll concentration in the rectangular box was 0.16 mg/m 3 on 17 May and 0.6 mg/m 3

Inland flooding
Cyclone induces flash floods in many low-lying areas in the coastal regions. While inland off from the coast, the primary cause of flooding is intense rainfall, but in areas close to coastline, flooding may be caused by both rainfall and storm surge. Cyclone Yaas brought a high amount of rainfall that caused inundation in many parts of Odisha, West Bengal and Bihar. The intense rain on its landfall and after it moves over the land is shown in Fig. 4g and h. SAR-based inland inundation after cyclone Yaas in the eastern coast of India is shown in Fig. 13. The regional threshold was determined based on the histogram backscatter values to delineate the still water surface and quasi-inundated areas. Heavy flooding was observed in the coastal districts of Orissa. For cyclone Yaas, using SAR, we have done the delineation exercise for the districts of Bhadrak, Kendrapara, Baleshwar and Purba Medinipur, and that were found to have 1280, 933, 413 and 530 km 2 inundated areas, respectively, which were about 52%, 38%, 11% and 13% of the total inundated area.
Water bodies and inundated regions show high value of MPDI. Flood index for pixels with MPDI in the range of 0.01 and 0.1 is mapped on a linear scale of 0-1. MPDI-based flood index maps are generated for cyclone Yaas during 27-28 May 2021. The BT difference and surface flooding maps ( Fig. 14a and b) clearly show the affected regions of Orissa, Bihar, Jharkhand and West Bengal.

Damage assessment of agricultural crops and forest in coastal regions
The damage to the crop and forest vegetation especially coastal fishing area in Odisha and West Bengal were mainly due to storm winds, and inundation (Fig. 14) by both intense rain, wind and ingress of saline water associated with storm surge (Fig. 6) from sea on 26 May 2021. Areas that are affected include the coastal region of Odisha and West Bengal including Bhadrak, Balasore, East Midnapore and South 24 Parganas districts. A detailed study of affected regions was carried out using both SAR and optical sensor data. The results over agricultural patches showed that both VH and VV backscatter (in dB) increased from pre-to post-cyclone periods in parts of Balasore, Mayurbhanj, Bhadrak, East Midnapore and South 24 Parganas districts. However, the increase in dB was more profound in VH channel rather than VV (Fig. 15a). Similar increase in VH backscatter post a cyclonic event has been reported by other researchers also (Bell et al. 2019;Ajadi et al. 2020). The increase in SAR backscatter in a pixel can be attributed to either increased volume scattering due to structural changes associated with crop lodging or more double bounce effects due to soil exposure or the combination of both conditions (Ajadi et al. 2020). It was found that both the differences decreased in affected areas. This might be due to inundation of crop fields. On the other hand, in other parts, particularly near Simlipal forest and parts of West Midnapore district, the backscatter difference between pre-and post-cyclone was found to be negative. This could be due to increase in backscatter either due to increase in surface soil moisture (of exposed soil) or increase in surface roughness due to crop canopy disturbances. It was found that both skewness and kurtosis of frequency  distribution of histograms ( Fig. 15a) varied between pre-and post-cyclone for both VH and VV polarizations. However, the effective changes in both the parameters were higher in case of VH backscatter than VV backscatter. Therefore, VH backscatter difference over the study sites was used for further mapping of affected areas. In mapping of the affected areas, NDVI difference from pre-to post-cyclone has also been employed. Those pixels with positive VH difference vis-a-vis positive NDVI anomaly were retained as probable affected areas. Backscatter difference of − 1 to − 4.5 dB (coming within 95 percentile) was classified through equi-quantile segregation into three classes of affected areas. Low affected areas are having dB difference of − 1 to − 1.6 dB, moderately affected areas are having dB difference of − 1.61 to − 3.2 dB and highly affected regions are having dB difference of − 3.21 to − 4.5 dB. The affected area of agricultural patches was estimated to be 1,002,277 ha (Table 6) with higher proportion in West Bengal state (52%) than in Odisha state (48%). Estimation was also made under different levels of impact (high, moderate and low). They constitute 10.1%, 31% and 58.9%, respectively. The high level of impact was observed in Odisha (11.9%) as compared to West Bengal (7.8%). Moderate level of impact was at par in both the states (29.9-30.4%) while the low level of impact was higher in West Bengal (60.4%) than in Odisha (53.9%). A limited number of GT locations were collected over 203 agricultural patches dominated by summer groundnut in the eastern and western Midnapore districts of West Bengal by a survey team from West Bengal Science and Technology-Biotechnology. Our validation showed that the area affected map generated through the current methodology using Opti-SAR combination matches well to the order of 93% cases. Pre-and post-cyclone data NDVI composite during 1-20 May 2021 and 26 May-1 June 2021 from Sentinel-2 was analysed over the forest regions of Balasore, Bhadrak and Kendrapara, Mayurbhanj (Simlipal Wildlife Sanctuary) districts, and mangrove vegetation of coastal Odisha, particularly Bhitarkanika National Park to generate maximum NDVI composite. Vegetation phenology from representative areas in Simlipal forest was first analysed using time series of NDVI for a normal year (2019) which showed increasing trend with June NDVI higher than May NDVI making the difference (May-June) negative due to change in phenophase. However, positive NDVI change was found between pre-and post-cyclone periods that must not have been due to phenological shift but due to impact of cyclone. The affected patches Balasore, Bhadrak and Kendrapara districts, and Simlipal forest and mangroves in and around Bhitarkanika surroundings are shown in Fig. 15b, and areas are provided in Table 7. In summary, it can be noted that the forest patches of Bhitarkanika (41%) and Simlipal forest (46%) were largely affected followed by forest patches of Bhadrak, Kendrapara and Balasore districts in the range of 10-31% which have also been supported by several ground-based reports. In this region, strong winds were the main reason for uprooting/felling-off of forest trees. The present study is carried out using available SAR and optical observations over the study area. However, for more precise and timely assessment of the damaged area, it may be necessary to get access to the frequent (with 1-2 days repetivity) high spatial resolution SAR and optical data (better than Sentinel data suite).

Conclusions
TCs are the strongest manifestation of the violent weather that can cause huge losses of the lives of human and animals, damage to the property, vegetation, marine, coastal and inland environment and displacement of scores of the people. In the present study, a threestage cyclone hazard tracking is proposed, which includes its formation, structure, tracking, Fig. 13 Inundation after cyclone Yaas in the Kendrapara district, Odisha, and Bhadrak district, Odisha, using Sentinel-1 C band 10-m spatial resolution dataset intensification, landfall, storm surge, coastal inundation and effect on marine and terrestrial ecosystem. In the present study, an attempt is made to address all these aspects of cyclone including its effect on atmospheric, marine, coastal and inland areas with emphasis on maximum use of the satellite measurements that may help to manage the disaster of such a huge scale in most effective way. This study becomes difficult because it ventures into the domains of diverse sciences including meteorology, hydrology, agriculture science and physical and biological oceanography. This paper is based on a case study of a Very Severe Cyclonic Storm (VSCS), named Yaas, developed over the Bay of Bengal (BoB) that hit the Odisha coast to the south of   Balasore on 26 May 2021. Prior to the formation of TC, SST and ocean heat potential were found favourable for the cyclogenesis, which was subsequently successfully observed as well as forecasted. After the cyclone was formed, its intensity and track predictions were generated, which were found following closely to the observations. The frequent observations from INSAT-3D/3DR were potentially used in locating the centre of the cyclone, and distribution of rain and clouds. Further, as cyclone passes, change in SSS and SST, thermocline displacement and Ekman suction were computed, and associated changes in ocean productivity were examined. Numerical model-based simulations are presented for storm surge, ocean waves and swells at the coast, which were found to be matching with available buoy observations. After the landfall, the inundation of the coastal and inland regions by storm surge alone with torrential rains was examined. The inundated area was also identified in SAR, visible band and passive microwave measurements. Finally, an assessment of damage of crops and forest area in the cyclone-affected regions was analysed and presented. The study is completed in all aspects, and results are presented with limited validation from the available ground observations. This study may become a template for following the similar approach to manage the cyclone disaster, assessment of the losses and providing the relief to the affected people.