Prediction of Rain Occurrence and Accumulation Using ... - IEEE Xplore

9 downloads 0 Views 2MB Size Report
Abstract—A nowcasting technique has been proposed to esti- mate the impending rain accumulation using ground-based radio- metric measurements at ...
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 5, MAY 2018

2789

Prediction of Rain Occurrence and Accumulation Using Multifrequency Radiometric Observations Animesh Maitra , Senior Member, IEEE, and Rohit Chakraborty

Abstract— A nowcasting technique has been proposed to estimate the impending rain accumulation using ground-based radiometric measurements at Kolkata (22.65°N, 88.45°E), a tropical location. It has been observed that the normalized variation of brightness temperature (BT) at 31 GHz along with the standard deviation of BT at 22 GHz and instability indices, namely, lifting index, have shown definite changes before rain events. A combination of these three parameters can be effective in predicting rain events both qualitatively and quantitatively. Accordingly, a prediction model is developed and tested on several intense rain events during the period 2014–2015. The model is found to perform reasonably well in predicting intense rain about 70–75 min in advance with an efficiency of 80%. Index Terms— Atmospheric instability, brightness temperature (BT), convective rain, microwave radiometry, prediction algorithm.

I. I NTRODUCTION NTENSE convective activities are known climatic phenomena in the east and the north-eastern parts of the Indian subcontinent during the premonsoon and monsoon seasons [1], [2]. These types of activities induce several adverse effects on various fields of life and, thus, prediction of such events is useful. In the past, prediction of convective activities was done from space-borne and ground-based radars [3]–[7]. However, these techniques did not prove successful as the obtained lead time is low. A combination of various kinds of measurements involving radar, satellite, atmospheric profiler, and lightning detector is very effective in nowcasting intense convective activities. Atmospheric profiles and instability indices obtained from radiosondes can also be useful tool in nowcasting rain occurrences [8]–[10]. On the other hand, microwave radiometers generate atmospheric profiles of humidity and temperature with a high temporal resolution, which make it more suitable for characterization and prediction of intense rain events compared with

I

Manuscript received April 1, 2017; revised August 16, 2017; accepted October 24, 2017. Date of publication January 4, 2018; date of current version April 20, 2018. This work was supported by the Indian Space Research Organization through the projects “Ku/Ka Band Channel Modelling for SATCOM Links Over Indian Region” under Grant ISRO/RES/4/614/2014-15 and “Space Science Promotion Scheme” under Grant E 33013/3/2009-V. The work of R. Chakraborty was supported by the Science and Engineering Research Board, Government of India, through the award of a National Post-Doctoral Fellowship. (Corresponding author: Animesh Maitra.) A. Maitra is with the Institute of Radio Physics and Electronics, University of Calcutta, Kolkata 700009, India (e-mail: [email protected]). R. Chakraborty is now with the National Atmospheric Research Laboratory, Department of Space, Government of India, Gadanki 517112, India (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TGRS.2017.2783848

other techniques [11]. Short-term variations in the brightness temperature (BT) around the water vapor line obtained from the microwave radiometer can be very useful to predict rain [12], [13]. It has been shown that an increase in BT is observed in water vapor channels (22–30 GHz) about 2 h before rain [13]. Values of K-index have found to give a good correlation with the atmospheric instability obtained from the radiometer [14]. In the recent past, many attempts were reported to nowcast intense rain events using instability indices [15]–[17]. The Cumulus Tracking and Monitoring algorithm was used by [18] to detect storms with a lead time of 30–60 min and a prediction efficiency of 58%. The Warning Decision Support System-Integrated Information algorithm was implemented by [19] to nowcast thunderstorms. The technique proved successful in predicting rain about 30 min in advance with a minimal false alarm rate (FAR) of 0.03 but with a very low probability of detection of 0.46. Later, a technique was proposed to identify hailstorms using Meteosat Second Generation data from Spain [20]. This technique provided an accuracy of 76.9% with an FAR of 16.7%. More recently, BT at 10 GHz has been utilized to predict rain with a hit ratio of 74% and an FAR of 7% [21]. Precipitation intensities have a direct effect on flood monitoring, which is a major weather hazard in various parts of the world. Hence, studies on precipitation intensity and accumulation have been a subject of utmost interest in many fields of life. Recently, many space-borne techniques have evolved to predict such activities. Additionally, attenuation obtained from both rainfall and cloud particles have been used by L’ecuyer and Stephens [22] to estimate rain parameters using space-borne radars of CLOUDSAT. However required accuracies could not be achieved for surface rain rates >10 mm h−1 . Some other techniques have obtained certain thresholds for BTs from geostationary thermal infrared channel at 10.8 μm to predict rain intensities [23], [24]. A new technique, the precipitation evolving technique, was proposed in [25] which projects the rainfall maps from advanced microwave sounding unit (AMSU) and Microwave Humidity Sounder observations using infrared brightness temperature maps of water vapor (6.2 μm) and thermal-IR (10.8 μm) channels from the Spinning Enhanced Visible and Infrared Imager. The mentioned technique provides a good lead time of 2–3 h with a prediction efficiency of 70%, but the FAR is very high (35%). Recently, a new technique has been suggested on canonical correlation and calculation of suitable thresholds provided to the linear combination of the BTs using Special Sensor Microwave Imager Sounder, TRMM-PR, and AMSU satellite

0196-2892 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

2790

Fig. 1.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 5, MAY 2018

Variation in BT at various frequencies for rain events. (a) 22 GHz. (b) 26 GHz. (c) 31 GHz. (d) 51 GHz.

data [26], the prediction ratio being 0.55–0.7. At the same time, a probabilistic precipitation technique was also developed utilizing 3B42RT generated from TRMM to obtain daily maps of precipitation accumulation in Nepal and surrounding areas with a minimal estimation error [27]. Apart from these, there have also been many other attempts to predict rain using radiometers, radars, and rain gauge networks [28]–[37]. Most of the above techniques have provided either a prediction efficiency less than 80% or an FAR greater than 20%. Recently, it has been reported that BT at both water vapor and oxygen absorption line can successfully predict 90% of rain events [38]; however, the obtained lead time was low. Later on, this work was improved [11] using instability indices like lifting index (LI) that predicted many rain events with a lead time of 70 min, but it could not predict the rainfall amount in advance. In this paper, the radio environment over Kolkata (22.65°N, 88.45°E) has been studied using BT observations at 22 and 31 GHz and instability indices from multifrequency profiler radiometer. Since BT variation at 31 GHz can effectively represent the amount of liquid water associated with the rain events, this parameter has been additionally used with other two inputs from the previous nowcasting model to predict rain, both qualitatively and quantitatively with a good lead time.

University of Calcutta, has been utilized for the study. It consists of two receiving units, rain sensor, GPS clock, and meteorological sensors [39]. This instrument measures BTs with a temporal resolution of 3 s and provides an accuracy of 0.5 K in two separate bands. The first frequency band (22–31.4 GHz) is sensitive to water vapor and hence is utilized for humidity profiling. The second frequency band (51.2–58 GHz) is sensitive to oxygen absorption and hence is utilized to obtain the temperature profiles. A detailed assessment on the quality of data provided by this instrument has already been done in previous studies [40], [41]. A Waldgovel-type impact disdrometer (RD-80) has also been used for rain rate measurements. A microrain radar (MRR) operating at 24.1 GHz has been used for generating profiles of rain rate in this paper [42]–[44]. It may be noted that all these instruments are stationed at the same location. For the present investigation, BTs’ observations at 31 GHz from radiometer for 80 convective rain events and 120 clear days during 2011–2012 have been used to develop this model. For testing the technique, a set of 86 events of 2013–2014 has been utilized. Finally, for validation of the prediction technique, BT data at 22 and 31 GHz and instability index LI for 50 events during 2015 have been utilized. Rain events having duration between 10 and 90 min and total accumulation greater than 5 mm have been considered for the study.

II. I NSTRUMENT D ESCRIPTION AND DATA A multifrequency radiometer (the Radiometer Physics GmbH - Humidity and Temperature Profiler) located on the roof top of Institute of Radio Physics and Electronics,

III. M ETHODOLOGY Convective processes are generally triggered by moisture ingress from the surroundings or from a sudden heating of the

MAITRA AND CHAKRABORTY: PREDICTION OF RAIN OCCURRENCE AND ACCUMULATION

2791

TABLE I C ORRELATION C OEFFICIENT B ETWEEN R AIN A CCUMULATION AND N AT D IFFERENT T IME G APS B EFORE R AIN

Fig. 2.

Variation in BT at 31 GHz for a heavy rain event.

atmosphere. Generally, at the growth stage of such activities, temperatures at the lower atmosphere increase resulting in updrafts of moisture laden to create clouds [45]. Simultaneously, lower atmospheric heating disturbs the thermodynamic balance leading to instability. The strength of this atmospheric instability generated in 850–500 mbar pressure levels of the atmosphere plays an important role in sustaining the convective processes. Because of the convective growth, the temperature balance in all the layers of lower atmosphere gets disturbed leading to condensation of cloud to liquid with an emission of latent heat. This increase in liquid water content in the maturity stages leads to intense rain. However, it is possible that even in the growth stage before heavy rain, minuscule instability can develop, which stimulates formation of small liquid water content producing marginal rain that does not reach the ground. Consequently, the generated liquid water produces enough latent heating to boost the instability resulting in stronger updraft and downdrafts. At a later stage, the instability growth triggers convective maturity producing rain. It is also expected that stronger rain events have higher rain accumulation that is expected to be triggered by larger instabilities manifested by greater variations in BTs close to liquid water absorption. Thus, BT changes about 1 h before rain can be potential tools to forecast and estimate the impending rain. For the present radiometric observations, BT data are available in the 22–31- and 51–58-GHz frequency bands sensitive to water vapor and oxygen absorption, respectively. It is also seen from the absorption spectra that the absorption due to liquid water increases with frequency. Thus, to have an idea of which frequency would be most sensitive to liquid water growth, superepoch analysis of BT variation is obtained at 22, 26, 31, and 51 GHz and shown in Fig. 1. In this analysis, average BT values are taken at various instants before rain event shown by point “0” in the x-axis for premonsoon, monsoon, and yearly average. The analysis indicates that for the 31-GHz BT case, especially during premonsoon, the average BT has increased from 48 to 102 K showing a maximum increase of 54 K 1 h before rain. This prompts the use of 31-GHz BT data for further analysis. To further investigate this phenomenon, BT variations have been studied for a set of 80 heavy rain events. The variation

of BT at 31 GHz prior to rain is shown for April 4, 2011 in Fig. 2. From Fig. 2, it can be seen that rain has started at 19:00 Indian Standard Time (IST) and experienced 10.5 mm rain accumulation from disdrometer observations. A sharp increase in BT is seen just before rain. However, a number of small peaks of BT are also seen at various time instants before the rain event. The variation of BT at 31 GHz is checked for other rain events revealing similar changes in BT. However, it is persistently seen that during rain, the BT values saturate. The actual change in BT much before rain is comparatively small. Also, the BTs show seasonal and diurnal variation. Hence, a normalized variation of BT needs to be calculated to show whether any significant variation occurs much before rain starts. The normalized variation of BT at 31 GHz has been termed as N for convenience. It has already been mentioned in [11] that a microwave radiometer provides definite signatures of impending rain about 75 min in advance. Hence, the time frame for calculation of N has been taken to be around 90 min. This normalized variation (N) has been calculated from BT measurements using the formulation N=

(BTmax − BTmin ) BTinit

(1)

where BTmax , BTmin , and BTinit refer to the maximum, minimum, and the initial magnitudes of BT for a time frame of 90 min. It can be inferred from Fig. 2 that BT shows a number of peaks before rain. However, it is still not clear enough that which peak of BT before rain is to be considered the most significant estimator of the impending rain. For this reason, the values of rain accumulation are recorded for all rain events of 2011–2012. The correlation coefficient between rain accumulation and the BT variation (N) at different time gaps before rain to the start of rain is recorded and shown in Table I. Table I shows that maximum correlation is observed for a time gap of 1.25–1.5 h. Since the previous study [11] showed that the present location experiences a significant enhancement in convection about 75 min before the events, the analysis is carried out taking BT variation 75 min before the events. This means that the normalized variation of BT at 31 GHz at about 75 min before rain has the best relation with the impending rain amount. Next, the normalized BT variation at

2792

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 5, MAY 2018

Fig. 3.

Variation of rain accumulation with variation of BT at 31 GHz. TABLE II

S TATISTICAL S IGNIFICANCE T ESTING OF R AIN A CCUMULATION P REDICTION AT VARIOUS F REQUENCIES

this time instant is plotted with respect to rain accumulation to determine the nature and order of relationship between them. Fig. 3 shows that a linear positive relationship exists between rain accumulation and N. Thus, N calculated 75 min before rain can give an estimate of the impending rain amount. In the previous section, it has been shown that the normalized variation of BT at 31.4 GHz can be suitable in rain magnitude prediction. However, it also has to be checked whether other frequencies of the microwave radiometer can provide similar or a better performance. Hence, the normalized variation is calculated for all the test events at 75 min before the occurrence of rain events at six different frequencies, namely, 22.23, 23.84, 26.04, 27.84, 31.40, and 51.26 GHz. Linear regression fits are done with respect to rain accumulation and the statistical significance of the fit is calculated in terms of the correlation coefficient, standard deviation, and the rootmean-squared error (RMSE). For a particular frequency to be reliable for the prediction purpose: 1) linear slope of the curve should be least; 2) correlation with the measured accumulation should be maximum; and 3) standard deviation and mean error of fit should be minimum. Now it can be seen from Table II that two frequencies, namely, 26.24 and 31.40 GHz, have higher correlation and lower standard deviation than the rest. However, 31.4 GHz has much higher correlation value with lower slope and RMSE, which would enable a more stable fitting. Hence, N calculated for this frequency at 75 min prior

Fig. 4. Superepoch variation. (a) Diurnal variation of N in clear days. (b) Average N variation before rain. (c) Probability distribution of N for clear days. (d) Probability distribution of N for rainy days.

to rain can be most suitable for rain accumulation estimation at the present location. To investigate whether N changes due to definite processes or some random phenomenon, a set of 120 clear days of 2011–2012 is selected and the hourly diurnal variation

MAITRA AND CHAKRABORTY: PREDICTION OF RAIN OCCURRENCE AND ACCUMULATION

of N is studied in Fig. 4. The average variation of N is also seen for 80 rain events of 2011–2012 with an interval of 15 min from 2 h before the event till 1 h after the event to find out any difference exists in the presence of rain. From Fig. 4(a) and (b), it can be seen that N is generally in the range of 0–0.2 in clear days, but 1 h before the event, it increases to 0.3. Next, probability distributions of N for clear and rainy days are investigated separately. From Fig. 4(c) and (d), it can be seen that there is a difference between the ranges of N during both the cases. During clear days, N lies in the range of 0–0.4, whereas in convective days, it is greater than 0.5, indicating convective growth before heavy rain. Thus, N can be a definite estimator for nowcasting heavy rain events. The next concern is to investigate the physical processes underlying such variation of BT before rain. About 1 h before rain, instability sets up in the atmosphere resulting in updrafts carrying water vapor that gets saturated to form liquid water. However, due to a dearth of sufficient instability, it cannot produce enough lifting force, and hence the resultant rain is only marginal. To show this phenomenon, the average variation of atmospheric instability is recorded for all test events. To study the growth of instability, total totals index (TTI) obtained from radiometer is utilized. TTI is formed by summation of two indices, namely, vertical totals index, which is the difference of temperature between 500- and 850-hPa levels and related to instability, and cross totals index, determined by moisture ingress at 850-hPa level. Thus, TTI gives a complete picture of the convective process. TTI values have been averaged for 15 min starting from 2 h before the rain events and shown in Fig. 5(a). Fig. 5(a) shows that TTI increases before the rain. However, a small peak of TTI is observed at 75 min before rain, which supports the possibility of a weak instability. Weak instabilities can favor marginal transport of water vapor and its condensation to liquid water. Hence, it can be expected that, at the same time, there should also be marginal rain in the atmosphere that rarely reaches the ground. To show it, the average rain profiles from MRR observations have been recorded and plotted. The colorbar in the contour plot shows the rain rates in logarithmic scale. Fig. 5(b) shows, as a typical case, that marginal rain has occurred 75 min before rain. It is known that this conversion of vapor to liquid results in latent heat release, which heats up the atmosphere leading to more instability. A dense cloud can release a significant latent heat to trigger a large convective cell. Thus, the resultant rain accumulation is generally influenced by latent heat release about 1 h before the event that is related to liquid growth or BT variation at 31 GHz. To investigate this phenomenon, the latent heat has been calculated from temperature profiles got from radiometer measurements at about 1 h before every rain event using the following relation [2]:   L c qs 1 + (RT dT )  = d  (2) s = − ε L 2c qs dz 1+ (c p RT 2 )

where ε = 0.622, d is the dry adiabatic lapse rate, and L c is the latent heat of condensation. The real root of L c has been taken into account. Values of latent heat and rain accumulation

2793

Fig. 5. Superepoch variation. (a) Average variation of instability index TTI. (b) Average rain profiles from MRR.

are plotted against each other in Fig. 6, which shows that the BT variation has a good agreement with the latent heat released before rain. It should also be mentioned that BT variation is calculated from the 22–31-GHz band while the latent heat is calculated from 51 to 58 GHz. Hence, their mutual independence further validates the physical basis behind the estimation of rain accumulation using BT at 31 GHz. Also, N has a good linear relation with the impending rain amount with a lead time of 75 min. The entire process is summarized in form of a flow diagram given in Fig. 7. The illustration of this flow diagram through the variation of different atmospheric parameters is shown in Fig. 8 for a typical rain event on August 27, 2013. Fig. 8 shows that rain started at 08:00 h IST. However, a rise in BT has been noted at around 07:00 IST. The normalized change N was calculated from BT measurements and a value of 1.5 was obtained. The estimated rain accumulation was around 30 mm. However, the actual rain accumulation of 28.5 mm has been recorded from disdrometer observations. Rain profile observations from MRR also indicated a patch of light rain at about 07:00 IST that occurred from low instability as shown by TTI values in Fig. 8(d). However, this rain led to heating of the atmosphere in various layers, shown

2794

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 5, MAY 2018

Fig. 6. Inter-relationship between various parameters. (a) Heat generated and normalized BT variation. (b) Rain accumulation and heat generated.

Fig. 7.

Flow diagram of a convective rain process.

in Fig. 8(c), which further increased the instability growth to reach convective maturity. The above estimation technique has been tested on 86 rain events of 2013–2014. The BT observations were recorded from 150 to 60 min before the occurrence of each event, and accordingly, the normalized variation has been calculated. Next, the rain accumulation has been estimated and compared with actual recorded rain accumulation measurements from disdrometer observations as shown in Fig. 9(a). The results show that there is a good match between the estimated and the recorded accumulation. However, it can be inferred that the estimation error increases for heavier rain. To test

Fig. 8. Illustration of the convective process. (a) BT variation at 31 GHz. (b) Rain profiles from MRR. (c) Temperature profiles from radiometer. (d) Variation of instability index TTI.

the statistical significance of the errors, the total test set is divided into three classes based on rain accumulation, namely: 1) 5–10 mm; 2) 10–20 mm; and 3) greater than 20 mm. Next, the error distributions have been plotted for these three classes separately in Fig. 9(b). It can be seen that the error spread

MAITRA AND CHAKRABORTY: PREDICTION OF RAIN OCCURRENCE AND ACCUMULATION

2795

Fig. 9. Performance constraints of the rainfall estimation technique. (a) Comparison between the predicted and recorded rain accumulations. (b) Distributions of estimation error for different rain accumulation classes.

increases with increasing rain accumulation. To quantify the expected errors, the standard deviations of errors are calculated and found to be 1.10, 2.77, and 6.89 mm, respectively, for the three classes in the order of increasing rainfall accumulation. This shows that this technique provides a reasonable estimate of the forthcoming rainfall amount with an error of 20%–30%. IV. D EVELOPMENT OF A C OMPOSITE M ODEL From the previous sections, it can be noted that BT variations at 31 GHz can be useful to estimate rainfall amount. However, this method does not have the full capability of providing a reliable prediction of rain occurrence. It has been seen from previous studies that information of water vapor or liquid content alone cannot create the necessary conditions for rainfall occurrences [46], [47]. Instability and lifting force are also important for sustenance and maturity of such events. Standard deviation of BT at 22 GHz along with an instability index like LI can be used to qualitatively

Fig. 10. Performance of the complete prediction technique. (a) Flowchart of the prediction algorithm. (b) A test case showing the performance of the technique for a rain event on August 26, 2013.

predict rain occurrence with high prediction efficiency and a lead time of 75 min [11]. The present technique can also efficiently predict the rain quantity with almost similar lead time. Hence, a combined system is proposed, which can predict rain occurrence and rainfall accumulation. The present prediction technique would check the standard deviation of BT at 22 GHz

2796

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 56, NO. 5, MAY 2018

and LI after every 90 min. If both the conditions regarding the moisture content and instability cross certain thresholds, as provided in [11], an alarm for rain occurrence is generated. Once the rain alarm is activated, N is calculated from 31-GHz BT to estimate the amount of rain to occur after 75 min. This composite prediction system would enable nowcasting of both rain occurrence and rainfall amount with a good lead time. The schematic of the prediction algorithm is shown in Fig. 10(a). The present technique is validated for a set of 50 rain events. It is seen that the technique has a prediction efficiency of about 80% with an FAR of 18%. A demonstration of the final prediction technique is shown in Fig. 10(b) for a typical heavy rain event on August 26, 2013 starting at 15:00 IST. However, prior to the rain event, at 13:30 IST, the standard deviation of BT at 22 GHz and LI crossed their thresholds causing a rain alarm to generate. Accordingly, N was estimated from BT variation at 31 GHz to be 1.45, and the system predicted a rain accumulation of around 29 mm that is close to the actual amount of 26 mm measured from disdrometer. Hence, this system serves to be an effective tool for rainfall prediction. V. C ONCLUSION This paper utilizes radiometric observations of BT at 22 and 31 GHz and the instability index LI to predict heavy rain events over Kolkata. A nowcasting model has been developed using these parameters to predict rain occurrence 75 min prior to the events along with the rain accumulation. It is seen that that a combination of LI and BT variations at 22 and 31 GHz can be effective to nowcast convective activities with prediction efficiencies of about 80% and FARs of 18%. This method can therefore be suitable to predict intense convective rain events. R EFERENCES [1] A. Maitra, S. Jana, R. Chakraborty, and S. Majumder, “Multi-technique observations of convective rain events at a tropical location,” in Proc. 31st URSI Gen. Assembly Sci. Symp. (URSI GASS), Beijing, China, Aug. 2014, pp. 1–4. [2] R. Chakraborty, S. Talukdar, U. Saha, S. Jana, and A. Maitra, “Anomalies in relative humidity profile in the boundary layer during convective rain,” Atmos. Res., vol. 191, pp. 74–83, Jul. 2017. [3] S. Mecklenburg, J. Joss, and W. Schmid, “Improving the nowcasting of precipitation in an Alpine region with an enhanced radar echo tracking algorithm,” J. Hydrol., vol. 239, nos. 1–4, pp. 46–68, 2001. [4] P. Wang, A. Smeaton, S. Lao, E. O’Connor, Y. Ling, and N. E. O’Connor, “Short-term rainfall nowcasting: Using rainfall radar imaging,” in Proc. 9th Irish Workshop Comput. Graph. Eurogr. Ireland, Dublin, Ireland, Dec. 2009, pp. 1–9. [5] D. Dutta et al., “Nowcasting of Yes/No rain situations at a station using soft computing technique to the radar imagery,” Indian J. Radio Space Phys., vol. 39, no. 2, pp. 92–102, 2010. [6] Z. Sokol, “Nowcasting of 1-h precipitation using radar and NWP data,” J. Hydrol., vol. 328, nos. 1–2, pp. 200–211, 2006. [7] A. Zahraei, K.-L. Hsu, S. Sorooshian, J. J. Gourley, Y. Hong, and A. Behrangi, “Short-term quantitative precipitation forecasting using an object-based approach,” J. Hydrol., vol. 483, pp. 1–15, Mar. 2013. [8] D. W. McCann, “WINDEX—A new index for forecasting microburst potential,” Weather Forecasting, vol. 9, no. 4, pp. 532–541, 1994. [9] B. Geerts, “Estimating downburst-related maximum surface wind speeds by means of proximity soundings in New South Wales, Australia,” Weather Forecasting, vol. 16, pp. 261–269, Apr. 2001. [10] A. Manzato, “A climatology of instability indices derived from Friuli Venezia Giulia soundings, using three different methods,” Atmos. Res., vols. 67–68, pp. 417–454, Jul./Sep. 2003.

[11] R. Chakraborty, S. Das, and A. Maitra, “Prediction of convective events using multi-frequency radiometric observations at Kolkata,” Atmos. Res., vol. 169, pp. 24–31, Mar. 2016. [12] P. W. Chan and Y. F. Lee, “Application of a ground-based, multichannel microwave radiometer to the alerting of low-level windshear at an airport,” Meteor. Zeiteschrift, vol. 20, no. 4, pp. 423–429, 2011. [13] H. Y. Won, Y.-H. Kim, and H.-S. Lee, “An application of brightness temperature received from a ground-based microwave radiometer to estimation of precipitation occurrences and rainfall intensity,” Asia–Pacific J. Atmos. Sci., vol. 45, no. 1, pp. 55–69, 2009. [14] P. W. Chan, “Performance and application of a multi-wavelength, ground-based microwave radiometer in intense convective weather,” Meteor. Zeiteschrift, vol. 18, no. 3, pp. 253–265, 2009. [15] A. Madhulatha, M. Rajeevan, M. V. Ratnam, J. Bhate, and C. V. Naidu, “Nowcasting severe convective activity over southeast India using ground-based microwave radiometer observations,” J. Geophys. Res., Atmos., vol. 118, no. 1, pp. 1–13, 2013. [16] J. G. Galway, “The lifted index as a predictor of latent instability,” Bull. Amer. Meteor. Soc, vol. 37, no. 1, pp. 528–529, 1956. [17] U. Saha, A. Maitra, S. K. Midya, and G. K. Das, “Association of thunderstorm frequency with rainfall occurrences over an Indian urban metropolis,” Atmos. Res., vol. 138, pp. 240–252, Mar. 2014. [18] N. Dotzek and C. Forster, “Quantitative comparison of METEOSAT thunderstorm detection and nowcasting with in situ reports in the European Severe Weather Database (ESWD),” Atmos. Res., vol. 100, no. 4, pp. 511–522, 2011. [19] M. Kohn, E. Galanti, C. Price, K. Lagouvardos, and V. Kotroni, “Nowcasting thunderstorms in the Mediterranean region using lightning data,” Atmos. Res., vol. 100, no. 4, pp. 489–502, 2011. [20] A. Merino, L. López, J. L. Sánchez, E. García-Ortega, E. Cattani, and V. Levizzani, “Daytime identification of summer hailstorm cells from MSG data,” Natural Hazards Earth Syst. Sci, vol. 14, no. 4, pp. 1017–1033, 2014. [21] P. Dvorak, M. Mazanek, and S. Zvanovec, “Short-term prediction and detection of dynamic atmospheric phenomena by microwave radiometer,” Radioengineering, vol. 21, no. 4, pp. 1060–1066, 2012. [22] T. S. L’Ecuyer and G. L. Stephens, “An estimation-based precipitation retrieval algorithm for attenuating radars,” J. Appl. Meteorol., vol. 41, pp. 272–285, Sep. 2001. [23] P. A. Arkin, “The relationship between fractional coverage of high cloud and rainfall accumulations during GATE over the B-scale array,” Monthly Weather Rev., vol. 107, pp. 1382–1387, Oct. 1979. [24] P. A. Arkin and B. N. Meisner, “The relationship between large-scale convective rainfall and cold cloud over the western hemisphere during 1982-84,” Monthly Weather Rev., vol. 115, pp. 51–74, Jan. 1987. [25] F. Di Paola et al., “Combined MW-IR precipitation evolving technique (PET) of convective rain fields,” Natural Hazards Earth Syst. Sci., vol. 12, pp. 3557–3570, Nov. 2012. [26] D. Casella et al., “A novel algorithm for detection of precipitation in tropical regions using PMW radiometers,” Atmos. Meas. Tech., vol. 8, no. 3, pp. 1217–1232, 2015. [27] N. Y. Krakauer, S. M. Pradhanang, J. Panthi, T. Lakhankar, and A. K. Jha, “Probabilistic precipitation estimation with a satellite product,” J. Clim., vol. 3, no. 2, pp. 329–348, 2015. [28] S. J. Wood, D. A. Jones, and R. J. Moore, “Accuracy of rainfall measurement for scales of hydrological interest,” Hydro. Earth Syst. Sci., vol. 4, pp. 531–543, Dec. 2000. [29] F. S. Marzano, D. Cimini, P. Ciotti, and R. Ware, “Modeling and measurement of rainfall by ground-based multispectral microwave radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 5, pp. 1000–1011, May 2005. [30] E. Goudenhoofdt and L. Delobbe, “Evaluation of radar-gauge merging methods for quantitative precipitation estimates,” Hydrol. Earth Syst. Sci., vol. 13, pp. 195–203, Feb. 2009. [31] J. C. Bergès, I. Jobard, F. Chopin, and R. Roca, “EPSAT-SG: A satellite method for precipitation estimation; its concepts and implementation for the AMMA experiment,” Ann. Geophys., vol. 28, pp. 289–308, Jan. 2010. [32] S. K. Midya, H. Sarkar, and U. Saha, “Sharp depletion of atmospheric refractive index associated with Nor’wester over Gangetic West Bengal: A possible method of forecasting Nor’wester,” Meteorol. Atmos. Phys, vol. 111, nos. 3–4, pp. 149–152, 2011. [33] U. Saha, S. K. Midya, H. Sarkar, and G. K. Das, “Sharp depletion of absolute humidity associated with squall over Kolkata (22°34 N, 88°26 E): A possible method of forecasting squall,” Pacific J. Sci. Technol., vol. 13, no. 1, pp. 683–688, 2012.

MAITRA AND CHAKRABORTY: PREDICTION OF RAIN OCCURRENCE AND ACCUMULATION

[34] R. Chakraborty, U. Saha, A. K. Singh, and A. Maitra, “Association of atmospheric pollution and instability indices: A detailed investigation over an Indian urban metropolis,” Atmos. Res., vol. 196, pp. 83–96, Nov. 2017. [35] S. Das, R. Chakraborty, and A. Maitra, “A random forest algorithm for nowcasting of intense precipitation events,” Adv. Space Res., vol. 60, no. 6, pp. 1271–1282, 2017. [36] G. Rakshit, A. Adhikari, and A. Maitra, “Modelling of rain decay parameter for attenuation estimation at a tropical location,” Adv. Space Res., vol. 59, no. 12, pp. 2901–2908, 2017. [37] R. Renju, C. S. Raju, N. Mathew, N. V. P. Kirankumar, and K. K. Moorthy, “Tropical convective cloud characterization using ground-based microwave radiometric observations,” IEEE Trans. Geosci. Remote Sens., vol. 54, no. 7, pp. 3774–3779, Jul. 2016. [38] R. Chakraborty, S. Das, S. Jana, and A. Maitra, “Nowcasting of rain events using multi-frequency radiometric observations,” J. Hydrol., vol. 513, pp. 467–474, May 2014. [39] T. Rose and H. Czekala, RPG-HATPRO Radiometer Operating Manual, Version 7.99, Radiometer Physics GmbH, Meckenheim, Germany, 2009. [40] C. Matzler and J. Morland, “Refined physical retrieval of integrated water vapor and cloud liquid for microwave radiometer data,” IEEE Trans. Geosci. Remote Sens., vol. 47, no. 6, pp. 1585–1594, Jun. 2009. [41] D. Cimini et al., “Thermodynamic atmospheric profiling during the 2010 Winter Olympics using ground-based microwave radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 12, pp. 4959–4969, Dec. 2011. [42] R. Chakraborty and A. Maitra, “Retrieval of atmospheric properties with radiometric measurements using neural network,” Atmos. Res., vol. 181, pp. 124–132, Nov. 2016. [43] G. Rakshit, R. Chakraborty, and A. Maitra, “Effect of vertical wind on rain drop size distributions in the boundary layer,” in Proc. Int. Symp. Antennas Propag. (ISAP), Okinawa, Japan, Oct. 2016, pp. 846–847. [44] B. K. Guha, R. Chakraborty, B. Nandy, and A. Maitra, “Rain attenuation over earth space path from radar and propagation measurements at tropical location,” in Proc. 6th Int. Conf. Comput. Devices Commun. (CODEC), Kolkata, India, Dec. 2015, pp. 1–4. [45] U. Saha and A. Maitra, “Effects on stratospheric moistening by rates of change of aerosol optical depth and ozone due to solar activity in extra-tropics,” Ind. J. Phys., vol. 88, no. 11, pp. 1117–1126, 2014. [46] S. C. Sherwood, R. Roca, T. M. Weckwerth, and N. G. Andronova, “Tropospheric water vapor, convection, and climate,” Rev. Geophys., vol. 48, no. 2, p. RG2001, 2010. [47] S. Das, S. Talukdar, R. Chakraborty, and A. Maitra, “Nowcasting of tropical rain using dual frequency atmospheric brightness temperatures at Kolkata,” in Proc. 5th Int. Conf. Comput. Devices Commun. (CODEC), Kolkata, India, Dec. 2012, pp. 1–4.

2797

Animesh Maitra (M’03–SM’06) received the M.Sc. degree in physics from Cotton College, Gauhati University, Gauhati, India, in 1975, and the Ph.D. degree in radio physics and electronics from the University of Calcutta, Kolkata, India, in 1986. From 1976 to 1981, he was a Research Fellow, and from 1981 to 1983, he was a Research Assistant with the Institute of Radio Physics and Electronics, University of Calcutta, where he joined as a Lecturer in 1983 and later became a Professor and the Head of the Institute. He was a Commonwealth Academic Staff Fellow at the Rutherford Appleton Laboratory, Didcot, U.K., from 1988 to 1989. He was also a British Council Visitor and a Royal SocietyINSA Exchange Visitor at a number of universities in the U.K. from 1986 to 2002. He also served as the Director of the S. K. Mitra Centre, University of Calcutta, from 2007 to 2017. He is currently a UGC Basic Science Research Faculty with the Institute of Radio Physics and Electronics, University of Calcutta, where he is leading a research group that has set up an extensive experimental facility for microwave sensing of the atmosphere, earth-space propagation studies, and aerosol measurements. He has contributed to a large number of publications addressing the impact of tropical radio environment on communications, multitechnique sensing of precipitation processes, nowcasting techniques, and role of urban pollutants in convective atmosphere. His research interests include radio propagation, remote sensing, radar, and atmospheric and space sciences. Prof. Maitra was a recipient of the Young Scientist Fellowship by the International Union of Radio Science (URSI) in 1987. He was the Founding Chairman of the IEEE GRSS Kolkata Chapter from 2011 to 2014, and is the Indian Representative in URSI Commission F. Rohit Chakraborty received the B.Tech. degree in electronics and communications from the West Bengal University of Technology, West Bengal, India, in 2010, and the M.Tech. and Ph.D. degrees in radio physics and electronics from the University of Calcutta, Kolkata, India, in 2012 and 2017, respectively. He is currently a National Post-Doctoral Fellow at the National Atmospheric Research Laboratory, Gadanki, India. He has authored and co-authored around 40 publications in international journals and conferences. His research interests include radio remote sensing, convective instability and precipitation, boundary layer dynamics, and earth-space propagation study. Dr. Chakraborty was a recipient of the International Union of Radio Science (URSI) Young Scientist Award in 2017 at the URSI General Assembly held in Montreal, Canada.