In IMD, upper-air observations are made at 43 RS stations, 34 RS stations are being used and 62 pilot balloon observatories to provide pressure, temperature, humidity and wind at various levels in the atmosphere up to an altitude of 30–35 km. Figure 2 shows the location (marked in red) of 34 RS stations. Observations from these stations are utilized for the comparison with INSAT-3D TPW. The types of ground equipment used in RS observatories are (1) Radiosonde Ground equipment (ECIL/DIGITAL) along with X band Win, (2) d finding Radars (EEC/MULTIMET) at 401 MHz and (3) IMS-1500 Radiotheodolite at 1680 MHz and SAMEER Radiotheodolite at 401 MHz. The performance of IMD's GPS radiosonde stations has been very thoroughly examined using ECMWF global data (Kumar et al., 2011).

IMD is augmenting Integrated Network of Global Navigation Satellite System (GNSS) receivers from 5 to 30 for integrated precipitable water vapour (IPWV) measurements. The network is capable of using other GNSS network data of research institutes in real-time basis for enhancing data spatial density and processing. The equipment has advanced meteorological sensors to measure the temperature, pressure, and humidity of the station and is capable of working independently in all-weather conditions with a high temporal resolution. Though satellites do not often fail, if one does, GNSS receivers can pick up signals from other satellites of the system. The data can be found from http://gnss.imd.gov.in/TrimblePivotWeb/ (last access: 18 October 2018).

The NOAA (National Oceanic and Atmospheric Administration) Satellite and Information Service provides timely access to global environmental data from satellites and other sources to monitor and understand the atmospheric variation over the Earth in efficient manner. In this study, we used blending TPW from two satellite sources, one from the Advanced Microwave Sounding Unit (AMSU) instruments on NOAA satellites (Ferraro et al., 2005) and the other from the Special Sensor Microwave Imager (SSM/I) instruments on Defence Meteorological Satellite Program (DMSP) satellites. In the blended TPW product, individual biases of the data sources have been mitigated to produce a more meteorologically significant product. A blending retrieval procedure was detailed (Kidder et al., 2007) and the methodology provides seamless global coverage without gaps to allow for the analysis of atmospheric moisture over land and ocean (Schmit et al., 2002; Smith et al., 2007). The products are on a Mercator projection with 16 km resolution at the Equator. The products are hourly in a HDF-EOS file format. These operational products were produced by the NOAA/NESDIS (National Environmental Satellite, Data, and Information Service) Office of Satellite and Product Operations (OSPO).

There is an on-board black body, which is responsible for generation of calibration information for all the IR channels in the geostationary satellite sounder. In-orbit readings of black-body temperatures revealed a gradient among the sensor, which led to inaccuracy in getting the correct black-body temperature. It was also observed that, during satellite midnight, Sun rays from behind the Earth directly enter the sensor and lead to an increase in black-body temperatures. This phenomenon leads to generation of incorrect calibration information. In order to provide climate quality products and to improve the calibration coefficients, GSICS (Global Space based Intercalibration System)-based intercalibration is used for INSAT-3D. The GSICS aims to intercalibrate a diverse range of satellite instruments to produce corrections ensuring consistency in satellite data sets. Allowing usage of calibration data, it produces globally homogeneous products for environmental monitoring. In addition, GSICS develops common methodologies to check the quality of sensors operated by various satellite agencies over the worldwide. The post-launch calibration strategy involves spectral response function of sensors, sensor performances and intercalibration of satellite sensor. Finally, recalibration of archived data or products of sensors is carried out, if necessary. The GSICS corrections are routinely applied in the IMDPS system of IMD by SAC (ISRO). The channelwise GSICS coefficients are found and applied during the radiometric correction process.

INSAT-3D-derived TPW is available at hourly interval over the Indian region. For validation purposes of TPW and its usefulness in weather monitoring and forecast, it is desirable to compare INSAT-3D TPW at different timescales with different sets of data. Thus, on a daily scale, we compared the INSAT-3D TPW with all the collocated measurements of RS TPW. On a monthly scale, monthly averaged data on collocated points were used. For the subdivision scale, five different regions were categorized according to meteorological subdivisions: northern India (NI), eastern India (EI), central India (CI), western India (WI) and peninsular India (PS) (Fig. 2).

Figure 3 shows the comparison of INSAT-3D and RS TPW on a daily scale during May–September 2016. On a day-to-day basis, INSAT-3D TPW agrees well with RS TPW. INSAT-3D TPW is able to measure the synoptic features of weather phenomena on a monthly scale over the Indian region very well. However, the magnitude differs, it can be termed as source of error due to registration and navigation issues at night-time. The more consistent and better correlation was seen above 40 mm, whereas for TPW values less than 40 mm, INSAT-3D underestimates slightly. This might be attributed to seasonal variation, topography of the region, and different climatic zones across India. The largest differences are observed mainly over mountainous areas and/or near the sea, which reveal differences in representativeness. Good confidence in INSAT-3D TPW estimates is gained during periods of moderate to heavy rain. The overall correlation on a daily scale was found to be 0.86. In a previous study, Mitra et al. (2015) reported 0.73 correlations using 10 IMD stations.

Figure 4 shows the comparison of INSAT-3D and RS TPW on a monthly scale during May–September 2016. The correlation coefficients are in the range of 0.78–0.87. It can be noted that, during the monsoon period, especially in the months of June, July, and August, when heavy rainfall (above 64.5 mm) occurs, INSAT-3D TPW shows good agreement with RS TPW. Mostly INSAT-3D TPW is higher when rainfall occurrence is higher above 40 mm. The statistics corresponding to this comparison is shown in Table 2. INSAT-3D coefficients of variation are high compared with RS, which indicates the higher variability in total precipitable water. The mean difference between RS and INSAT-3D TPW is much higher in the month of July, at 5.57 mm. It is due to the substantial rainfall during the monsoon season and in the subsequent months of August and September is 5.24 and 5.3 mm. It was also reported by Venkat Ratnam et al. (2016) that mean differences in the water vapour are as high as 20 %–30 %. The dry bias of 10 %–25 % in the INSAT-3D channel was compared to a similar satellite and the reanalysis data set was also noted. The coefficient of variation is lower for the months from July to September 2016. The coefficient of skewness is found to be negative between the INSAT-3D and RS measurements, which indicates that the mean is less than the mode of the data. The correlation coefficient shows good agreement with RMSE from June to September 2016, except in the month of July. The Student's t test is calculated for the significance of the computed parameter. The Student's t test shows the statistical significance of a linear relationship among the data, i.e. INSAT-3D and RS TPW.

Figure 5 shows the comparison of INSAT-3D with RS and NOAA TPW on a subdivisional scale from May to September 2016. It can be clearly seen from the figures that INSAT-3D TPW underestimates RS TPW, whereas it overestimates the NOAA TPW for the entire region during the monsoon period. A good correlation is observed for the regions CI and PS compared to EI and NI. However, opposite trends were found while comparing INSAT-3D with NOAA TPW. INSAT-3D TPW is always higher over NOAA TPW data. One of the possible reasons is that INSAT-3D sounder-derived TPW was calculated from the radiances sampled every hour, while NOAA TPW were based on only two satellite passes with Equator crossing times of 02:30 and 14:30 LT (in Indian Standard Time). Therefore, the sampling frequency of the radiometer is much higher in a geostationary satellite than a polar satellite. In general, subdivisional comparison reveals that the INSAT-3D TPW agrees well with RS and NOAA TPW below 23∘ N, whereas the difference is higher above 23∘ N.

Table 3 shows the statistics for the comparison of TPWs from INSAT-3D, RS, and NOAA at the subdivisions over India. INSAT-3D coefficients of variation are similar to those of RS, but in the case of NOAA they are higher with respect to INSAT-3D and RS. This might be attributed to the lower number of correlated points in the data set. The coefficient of skewness values was found to be negative for INSAT-3D, RS, and NOAA measurements. The correlation coefficients of TPWs were found to be in good agreement for INSAT-3D and NOAA (0.96) as well as for INSAT-3D and RS (0.87) from June to September 2016.

The box–whisker plot shown in Fig. 6 compares the actual rainfall observation and INSAT-3D TPW for different values from June to September 2016. This figure is constructed from the daily rainfall observations of between 0 to 140 mm occurring over the 34 stations and collocated mean INSAT-3D TPW values between 0 and 90 mm over the entire Indian region. It can be seen from Fig. 6 that TPW is binned for the ranges 0–20, 21–40, 41–60, 61–81, and > 80. As seen from the whiskers, the rainfall has the least scatter for the 0–20 bin, while for TPW > 80 it shows the most scatter. The mean and median are almost the same for all the TPW bins, except for the TPW > 80. There exists exponential behaviour between rainfall amounts with higher INSAT-3D TPW values. However, further analysis with a higher number of observations is required for the quantification of a non-linear or exponential relationship. Atmospheric constituents and synoptic scale of monsoon conditions are important factors when considering the occurrence of rainfall and satellite-derived TPW. It is demonstrated from Fig. 6 that the heavy and heavy-to-very-heavy rainfall correspond to the higher TPW values (60–80 and above 80 mm). The TPW corresponds to the cloud-free observations and rainfall measurements are for cloudy atmosphere. This can be obviously related to the fact that the heavy rainfall occurs in the presence of higher TPW values (Wu et al., 2003). However, for the light-to-moderate rainfall amount (less than 40 mm) INSAT-3D TPW is comparable. The moisture convergence, advection of moisture over geographical locations of the subdivisions occasionally receive heavy-to-very-heavy rainfall due to synoptic-scale monsoon circulations or its topography. The highly mountainous regions like the north-east, Jammu–Kashmir and parts of the Western Ghats (on the western coast of India), have less evaporation and higher rainfall as the moisture-laden air mass is transported over the regions (Venkat Ratnam et al., 2016). Similarly, it is also observed that the rainfall is overestimated in dry conditions because the falling raindrops evaporate before coming to the surface in dry conditions, resulting in the overestimation of rainfall.

Keeping in the mind the potential impact of TPW, hourly INSAT-3D sounder derived TPW, and GNSS TPW were analysed for thunderstorm events that occurred in Pune (18.52∘ E 73.85∘ N, 12:00 UTC, 3 June 2017), Kochi (9.93∘ E 76.26∘ N, 06:00 UTC, 6 June 2017), and Dibrugarh (27.47∘ E 94.91∘ N, 00:00 UTC, 9 June 2017).

The purpose of using the IMD GNSS instrument is that it has a maximum coverage of India and access to multiple satellites, redundancy and availability at all times. Further details can be found at http://gnss.imd.gov.in/TrimblePivotWeb/ (last access: 18 October 2018).

Figure 7 shows the hourly comparison between TPW derived from INSAT and GNSS during thunderstorm events. The grey bar shows the time of occurrence (i.e., 12:00 UTC) of the thunderstorm over Pune city. It was observed from the satellite images (not shown here) that the initial convection development starts at 06:00 UTC with multiple significant convections. It can be seen from Fig. 7 that the INSAT-3D TPW shows higher TPW values around 53 mm in comparison with GNSS TPW of 54 mm at 06:00 UTC. The higher TPW of INSAT-3D continues up to 11:00 UTC, which is in agreement with GNSS TPW. The thunderstorm was reported at 12:00 UTC. Since the INSAT-3D retrieval cannot be made over a cloudy region, the TPW observation was not available after 12:00 UTC.

In the case of the event at Kochi city, the grey bar shows the time of occurrence of the thunderstorm at 06:00 UTC. It was observed from the satellite imageries that initial convection development starts at 01:00 UTC. INSAT-3D TPW shows the higher TPW values around 58 mm in comparison with GNSS TPW of 51 mm at 01:00 UTC. The TPW observation was not available after the 03:00 UTC due to cloudy conditions. The higher TPW of INSAT-3D continues up to 03:00 UTC in agreement with GNSS TPW and a thunderstorm was observed at 06:00 UTC.

At 00:00 UTC a thunderstorm over Dibrugarh city was reported. The initial convection development started at 18:00 UTC with values around 53 mm in comparison with GNSS TPW of 58 mm at 18:00 UTC. It can be very easily seen from Fig. 7 that a fall in TPW occurs at 14:00 UTC from 50 to 24 mm. This is due to less precipitation in Dibrugarh, while from 14:00 to 18:00 UTC, no precipitation was noted due to the cloudy sky. The higher TPW of INSAT-3D continues up to 20:00 UTC, which is in agreement with GNSS TPW. The thunderstorm was reported at 00:00 UTC on 9 June 2017.

The case studies show that during the thunderstorm events, INSAT-3D-derived TPW compares reasonably well with GNSS TPW observations, indicating the potential of INSAT-3D-derived TPW for the studies on thunderstorm events. Along with other meteorological parameters (e.g., CAPE; convective available potential energy), instability indices with INSAT-3D TPW and the mesoscale activities can be very easily detected and utilized for weather forecasts. However, the above case studies confirm the usefulness of INSAT-3D-derived TPW prior to the event and it can be considered one of the precursors for convective activities.