Daily Rainfall for the Indian Monsoon Region from Merged Satellite and Rain Gauge Values: Large-Scale Analysis from Real-Time Data

A. K. Mitra National Centre for Medium Range Weather Forecasting, New Delhi, India

Search for other papers by A. K. Mitra in
Current site
Google Scholar
PubMed
Close
,
M. Das Gupta National Centre for Medium Range Weather Forecasting, New Delhi, India

Search for other papers by M. Das Gupta in
Current site
Google Scholar
PubMed
Close
,
S. V. Singh National Centre for Medium Range Weather Forecasting, New Delhi, India

Search for other papers by S. V. Singh in
Current site
Google Scholar
PubMed
Close
, and
T. N. Krishnamurti Department of Meteorology, The Florida State University, Tallahassee, Florida

Search for other papers by T. N. Krishnamurti in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A system for objectively producing daily large-scale analysis of rainfall for the Indian region has been developed and tested by using only available real-time rain gauge data and quantitative precipitation estimates from INSAT-1D IR data. The system uses a successive correction method to produce the analysis on a regular latitude–longitude grid. Quantitative precipitation estimates from the Indian National Satellite System (INSAT) operational geostationary satellite, INSAT-1D, IR data are used as the initial guess in the objective analysis method. Accumulated 24-h (daily) rainfall analyses are prepared each day by merging satellite and rain gauge data. The characteristics of the output from this analysis system have been examined by comparing the accumulated monthly observed rainfall with other available independent widely used datasets from the Global Precipitation Climatology Project (GPCP) and Climate Prediction Center Merged Analysis of Precipitation (CMAP) analyses. The monthly data prepared from the daily analyses are also compared with the subjectively analyzed India Meteorological Department (IMD) monthly rainfall maps. This comparison suggests that even with only the available real-time data from INSAT and rain gauge, it is possible to construct a usable large-scale rainfall map on regular latitude–longitude grids. This analysis, which uses a higher resolution and more local rain gauge data, is able to produce realistic details of the Indian summer monsoon rainfall patterns. The magnitude and distribution of orographic rainfall near the west coast of India is very different from and more realistic compared to both the GPCP and CMAP patterns. Due to the higher spatial resolution of the analysis system, the regions of heavy and light rain are demarcated clearly over the Indian landmass. Over the oceanic regions of the Arabian Sea, Bay of Bengal, and the equatorial Indian Ocean, the agreement of the analyzed rainfall at the monthly timescale is quite good compared to the other two datasets. For NWP and other model verification of large-scale rainfall, this dataset will be useful. In the field of rainfall monitoring within weather and climate research, this technique will have real-time applications with data from current (METSAT) and future (INSAT-3A and INSAT-3D) Indian geostationary satellites.

Current affiliation: Department of Meteorology, The Florida State University, Tallahassee, Florida

Corresponding author address: Dr. A. K. Mitra, Visiting Research Associate, Department of Meteorology, The Florida State University, Tallahassee, FL 32306. Email: akmitra@earl.met.fsu.edu

Abstract

A system for objectively producing daily large-scale analysis of rainfall for the Indian region has been developed and tested by using only available real-time rain gauge data and quantitative precipitation estimates from INSAT-1D IR data. The system uses a successive correction method to produce the analysis on a regular latitude–longitude grid. Quantitative precipitation estimates from the Indian National Satellite System (INSAT) operational geostationary satellite, INSAT-1D, IR data are used as the initial guess in the objective analysis method. Accumulated 24-h (daily) rainfall analyses are prepared each day by merging satellite and rain gauge data. The characteristics of the output from this analysis system have been examined by comparing the accumulated monthly observed rainfall with other available independent widely used datasets from the Global Precipitation Climatology Project (GPCP) and Climate Prediction Center Merged Analysis of Precipitation (CMAP) analyses. The monthly data prepared from the daily analyses are also compared with the subjectively analyzed India Meteorological Department (IMD) monthly rainfall maps. This comparison suggests that even with only the available real-time data from INSAT and rain gauge, it is possible to construct a usable large-scale rainfall map on regular latitude–longitude grids. This analysis, which uses a higher resolution and more local rain gauge data, is able to produce realistic details of the Indian summer monsoon rainfall patterns. The magnitude and distribution of orographic rainfall near the west coast of India is very different from and more realistic compared to both the GPCP and CMAP patterns. Due to the higher spatial resolution of the analysis system, the regions of heavy and light rain are demarcated clearly over the Indian landmass. Over the oceanic regions of the Arabian Sea, Bay of Bengal, and the equatorial Indian Ocean, the agreement of the analyzed rainfall at the monthly timescale is quite good compared to the other two datasets. For NWP and other model verification of large-scale rainfall, this dataset will be useful. In the field of rainfall monitoring within weather and climate research, this technique will have real-time applications with data from current (METSAT) and future (INSAT-3A and INSAT-3D) Indian geostationary satellites.

Current affiliation: Department of Meteorology, The Florida State University, Tallahassee, Florida

Corresponding author address: Dr. A. K. Mitra, Visiting Research Associate, Department of Meteorology, The Florida State University, Tallahassee, FL 32306. Email: akmitra@earl.met.fsu.edu

1. Introduction

Rainfall is the most important meteorological parameter affecting India's economy and other activities of strategic importance. Observations of rainfall are needed to support a range of services extending from the real-time monitoring and prediction of severe weather to climatological studies of drought. Precipitation also plays an important role in the global energy and water cycle. One of the most critical issues in modeling the global atmosphere and climate by general circulation models (GCMs) is the simulation and initialization of precipitation processes. GCM results need to be verified based on observational precipitation data to monitor the progress made in models and observation techniques (Janowiak 1992). The global and regional climate modeling community will need regional rainfall data for verifying the regional details derived from model simulations. The Asian monsoon system, in terms of the circulation and rainfall, is of great importance to climate researchers for understanding and predicting the general circulation and its variability at different space and time scales. In order to better understand and predict the monsoon, different atmospheric models and coupled ocean–atmosphere models are being used to simulate both the seasonal mean rainfall patterns and their interannual variability. It is also felt that the interannual variation of the Asian monsoon is linked to intraseasonal variations through various feedback mechanisms. Therefore, realistic daily and monthly rainfall datasets will be useful for studying the intraseasonal and interannual variations in the Indian summer monsoon (Kripalani et al. 1991). Observational and model simulations of the Madden–Julian oscillation (MJO), which includes large-scale organization and propagation (eastward and northward) of convective cloud clusters within the equatorial waveguide during the northern summer monsoon, are challenging scientific problems related to the understanding of the monsoon system. Daily analyses of rainfall data will be required to study these aspects of the monsoon.

From the daily data, it is easy to produce monthly data for a longer period, which will aid in studying the seasonal cycle and the interannual variability of the monsoon. These daily and monthly representative datasets of the observed rain will be ultimately used for different model verifications. Various major operational weather forecasting centers in the United States, Europe, Australia, and Japan have started issuing extended range (monthly to seasonal) dynamical predictions of rainfall for different geographical regions of the world. Considerable effort is put into refining the dynamical seasonal to interannual predictions by using ensemble methods, including multimodel ensembles. Research on these aspects of climate variability and predictability on seasonal to interannual timescales and the interaction of seasonal effects with monsoonal flow, is continuing. The dynamical, seasonal prediction study with the objective of assessing the predictability of the mean circulation and rainfall up to a season in advance is carried out. In order to make progress in the area of dynamical extended range prediction of monsoon rainfall, the first requirement is obtaining observed rainfall data (used in verifying the models) and understanding the processes involved. Due to the previously mentioned reasons, in recent years there has been considerable interest among different scientific groups in the preparation and intercomparison of gridded rainfall datasets (Huffman et al. 1997; Xie and Arkin 1997; Janowiak and Xie 1999; Gruber et al. 2000; Beranger et al. 2001; Grassl et al. 2000). Rainfall prepared under the Global Precipitation Climatology Project (GPCP) and the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) are the most widely used and helpful datasets for various scientific research conducted in the area of weather and climate. These datasets are derived from remote sensing observations and in situ measurements from rain gauges. Short-range forecasts from atmospheric numerical models are used in the CMAP analysis. However, the contribution from a forecast model may be marginal. These new continuous sets of data are able to produce new climatologies of rainfall for different geographical regions of the world. In spite of differences in terms of their magnitudes, the seasonal cycle and interannual variability are produced adequately by these datasets. But there are still regions where the disagreement between the datasets may be of high order. The Asian monsoon region is one such important area where more detailed comparisons are required to assess the strengths and weaknesses of these popular datasets.

In India, different types of numerical weather prediction models are being used for operational and research purposes by various organizations. Rainfall prediction by such numerical forecast models for different Indian domains is of crucial importance to activities like agriculture and other economic activities. Observed rainfall datasets are needed on latitude–longitude grids to verify the rainfall forecasts from these operational numerical models. Various research scientists and organizations also require these verifying rainfall datasets in order to continuously improve the forecast models—by improving model resolution, physical parameterizations, and improved satellite data usage during data assimilation. At the Indian National Centre for Medium Range Weather Forecasting (NCMRWF), a global operational T80/L18 analysis/forecast system is being run daily to produce real-time medium-range weather forecasts. These are disseminated to different agro-meteorological field units located in different agro-climatic zones of the country for agriculture-related operations. The global forecast model produces large-scale rainfall forecasts 5 days in advance, in roughly 1.5° by 1.5° latitude–longitude grid boxes. Verification of rainfall forecasts from global weather forecast models requires suitable observations, upscaling, and interpretation techniques (Cherubini et al. 2001). Local observations play an important role in preparing such verifying datasets (Mills et al. 1997; Weymouth et al. 1999). To find out the performance of the NCMRWF global numerical model in terms of the rainfall for different weather events, it was necessary to develop a rainfall analysis system on a 1.5° latitude–longitude grid, representative of the model's resolvable scales. In this study, with available real-time rainfall data only from satellite and rain gauges, daily objectively analyzed rainfall datasets have been prepared on regular latitude–longitude grids for the full summer monsoon season of 2001 (June–September) for the Indian region. The monthly accumulated rainfall from this daily analysis is then compared with the other available popular datasets for the monsoon region. Inferences about the quality of these merged data are made in relation to the existing popular datasets.

2. Method and data

In this section, the daily rainfall analysis procedure using the available real-time data is described. Because the monthly datasets are compared with GPCP and CMAP analyses, the data used and the analysis procedures of GPCP and CMAP are briefly mentioned, along with their characteristics as reported by some recent studies.

a. Analysis procedure and data used

The first guess used in the successive correction method for rainfall analysis is taken from the satellite meteorology division of the India Meteorological Department, New Delhi. The estimated rainfall from the Indian National Satellite System geostationary satellite INSAT-1D (positioned at 74°E longitude) IR data is based on the Geostationary Operational Environmental Satellite (GOES) precipitation index (GPI) technique (Arkin and Meisner 1987; Arkin et al. 1989). The Very High Resolution Radiometer (VHRR) onboard INSAT-1D included a visible channel operating in the spectral wavelengths of 0.55–0.75 μm and an infrared channel operating at 10.5–12.5 μm. The spatial resolution of the visible and infrared channels are 2.75 and 11 km, respectively. The domain selected for the rainfall analysis is from 40° to 118°E longitude and 39°S to 39°N latitude, cast in a 1.5° × 1.5° grid. This domain was chosen because the daily 24-h accumulated (valid at 0300 UTC) rainfall estimates from INSAT were available in real time for the defined region on 1° × 1° regular grids. The satellite-derived rainfall estimates are interpolated bilinearly to the 1.5° analysis grid and are used as the first guess in the objective analysis procedure. The main reason for performing analysis on a 1.5° grid is that at NCMRWF, a global T80/L18 (equivalent horizontal resolution of 1.5°) model is used for issuing medium-range predictions in real time. The model rain can now be verified against this merged rain analysis, which has the same resolution. The other data used are the 24-h accumulated (valid at 0300 UTC) rainfall values from rain gauge observations available through the global telecommunication system (GTS) network in real time for the same domain. Figure 1a shows the geographical distribution of such gauge observations on a typical day. Figure 1b shows the amount of daily available data from such sources during the monsoon season of 2001. All the data used here are available in real time and the present study is aimed at testing the technique for real-time rainfall analysis production in operational use for the Indian region. In Fig. 1b, the solid line indicates the availability of data for the whole analysis region, and the dashed line indicates data for the Indian region (5°–40°N, 65°–100°E). On average, 257 observations were available per day in the whole analysis domain, and out of that, 212 per day were from the Indian box. India being surrounded by oceans, we notice that most of the real-time data for the full analysis domain comes from the Indian box. We can say that on average 82% of the total rain gauge data were from the Indian box region.

The objective analysis technique for rainfall used here is based on the successive correction method of Cressman (Tripoli and Krishnamurti 1975; Krishnamurti et al. 1983; Mitra et al. 1997). This involves the successive modification of an initial guess field (satellite estimates) based on observed station data (rain gauge). Presuming that the gauges are perfect, the error (bias) correction for the satellite estimate at each grid point is derived. First, the satellite estimates are interpolated to station locations to form a first guess. Their differences from the observed station values provide an error estimate at the station locations. This set of irregularly spaced values is used to derive corrections at the desired grid point using successive iterative corrections. Consider the guess field (satellite rainfall in mm day−1) rg(ν)ij defined at each grid point (i, j), on the υth iterative guess. From the two-dimensional initial guess field array, the guess value at each station location s is computed by a standard nine-point Lagrangian interpolation formula. The Cressman weight function used in the analysis is defined by
i1525-7541-4-5-769-e21
where R is the influence radius and d is the distance of the station from the grid point. During the successive correction, we used four scan radii of 1.5°, 1.4°, 1.2°, and 1.0°. Since the intention is to represent the observed large-scale monsoon rainfall at a 1.5° grid box, we intentionally select the scan radii slightly larger than the grid box size to accommodate and account for the continuity of the large-scale rainfall in relation to the processes happening in the neighboring grids. Next, the error (observation minus guess at station location) E(ν)s between the interpolated value rg(ν)s and the actual observation (station value) r(ν)s, given by
E(ν)sr(ν)s(ν)s
is computed. The value of the E(ν)s gives a correction to a nearby grid point for the (new + 1) iterative value. This correction is given by
i1525-7541-4-5-769-e23
where W(ν)s is a weighting factor defined as
W(ν)sω(ν)dβγ
Here, ω(ν) in Eq. (2.4) is the Cressman weighting function at the iteration level (new) as described earlier. The radius of influence R is a function of the iteration, which decreases with each scan, from 1.5° in first scan to 1° in the fourth scan. The time weighting function β is kept equal to one as all the rain gauge data used in this study belong to the same valid 24-h time window ending at 0300 UTC. The reliability estimate of each observation type, described by γ, is also kept as 1 as only one type of observation (rain gauge) is used now. In the future, different data types (rain radar, other satellite estimates) appearing in the same grid box can be assigned different reliability estimates within the respective weights.
The next iterative guess is obtained by the formula,
(ν+1)ij(ν)ijαC(ν+1)ij
where α is a smoothing factor related to the center of gravity of the influencing data. The smoothing factor is used to force the resultant field to press smoothly into the guess field while approaching data-void regions. Accepting the first guess field over data-void regions is believed far more reliable than extrapolation of distant existing data.

b. GPCP and CMAP analyzed data

GPCP (Huffman et al. 1997) and CMAP (Xie and Arkin 1997) are the two published and widely used datasets that contain analyses of global precipitation derived by merging rain gauge estimates with multisatellite estimates of precipitation. The GPCP version 2 final analysis (on a 2.5° latitude–longitude global grid) uses oceanic satellite estimates of precipitation derived from the Special Sensor Microwave Imager (SSM/I) emission algorithm. Over land, the SSM/I scattering algorithm estimates are used. Then the magnitudes of the IR-based estimates are adjusted in relation to the microwave-based (SSM/I) precipitation estimates (Adler et al. 1994). Poleward of 40° latitude where IR-based estimates are not usable, remotely sensed estimates from the Television Infrared Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) are incorporated. This combination of SSM/I, IR, and TOVS forms the multisatellite GPCP analysis. Over land, the bias in the multisatellite analysis is adjusted by rain gauge analyses, but no adjustment is made over ocean. In the final step, the multisatellite and rain gauge analyses are merged using an inverse error-weighting scheme to form the final GPCP analysis. The temporal and spatial scales of the CMAP datasets are the same as for GPCP. In the CMAP analysis, the first step is to combine the various satellite estimates based on IR, SSM/I (scattering algorithm over land and ocean, emission algorithms over ocean), and Microwave Sounding Unit (MSU) data using a maximum likelihood approach in which the weighting coefficients are inversely proportional to the squares of the individual random errors. These errors are assessed over land areas by comparing the satellite estimates with rain gauge analyses and over oceans by comparing them with rain gauge precipitation over atolls that are located in the western Pacific. The latter are assumed representative of open ocean conditions. Over land, the resulting satellite-based analyses are then merged with rain gauge analyses via a blending technique in which the satellite estimates provide the field shape, which is anchored to the available rain gauge data. Over ocean, the gauge adjustment is a bulk scheme, with latitude dependence tapering away from the equator. CMAP analyses use the NWP rainfall forecasts, which are found to be useful for middle and higher latitudes.

While there is good large-scale agreement between the two datasets, considerable regional differences also exist (Gruber et al. 2000; Janowiak et al. 2001). Since both datasets use similar gauge stations, their agreement is good over land. Over ocean, large-scale regional disagreements are reported. Over many of the oceanic tropical areas, CMAP is 20%–50% wetter compared to GPCP. In contrast, GPCP is wetter than CMAP over all oceanic areas poleward of about 45°. The differences that are noted are due primarily to differences in the input data and methodologies by which the various input data are merged. Although regional differences occur in the mean fields, the analyses compare well in terms of monthly anomalies. Seasonal variations are rather similar (correlation > 0.7) in most of the Tropics and midlatitudes. It has been shown (Adler et al. 2001) that merged precipitation datasets such as GPCP and CMAP provide superior results compared to other published and available products.

3. Performance of the analysis system during summer monsoon 2001

2001 had a normal monsoon, as discussed in a special issue report from the India Meteorological Department (IMD; 2001b). Rainfall for the country as a whole was 92% of its long-period average. During this year, the rainfall was more evenly distributed compared to the previous year. During the 2001 monsoon season (June–September), up to the third week of August, the rainfall was above normal and thereafter it was persistently below normal. Figure 2a indicates different aspects of the observed rainfall variability during the monsoon season as inferred from rain gauge data only. The dark solid line shows the maximum rainfall (in mm day−1) reported from any station on different days, roughly indicating the increase or decrease of convective activity. After 85 days (during September) the rainfall activity is reduced. The light solid line shows the daily variance of observed rainfall computed from rain gauge data during the season. We notice that the two curves have a good correlation. This shows that whenever the rainfall activity increases, the spatial variability of rainfall also increases. The dashed line shows the computed daily root-mean-square error (rmse) of the objectively analyzed rain with respect to the rain gauge observations. For convenience of plotting and visualization, the rmse reported in millimeters is multiplied by a factor of 10. We notice that in general the rmse of the final analysis is around 6.5 mm. The day-to-day changes in the rmse curve also agree very well with the other two curves. The quality of the final analysis improves (deteriorates) with a(n) decrease (increase) in rainfall activity in the region during the monsoon. Figure 2b shows the amount of spatial variability in the rainfall observations and in the final analysis. The dashed line indicates the spatial variance of the daily rainfall computed from the reported rain gauge data only. In the rain gauge data, stations with “no-rain” cases are also included. The light solid line shows the daily spatial variance of the objectively analyzed daily rain. The dark solid line shows the ratio of the variance of the analyzed rain to the variance of rainfall observations in terms of percentage. As expected, the monsoon rainfall has a high spatial variability as seen from the gauge data. The average of the daily variance during the season from rain gauge data is 195 mm2. However, the variance seen in the analyzed field is much lower, around 56 mm2 on average. In a mean sense (average of 122 days) the variance of the objectively analyzed rain is 36% of the original variance of the gauge data.

By adding the daily analyzed values, monthly total rainfall maps were prepared for four different months (June, July, August, and September) for the summer monsoon of 2001. These are now compared with the GPCP and CMAP analyses for finding the usability of this new dataset At the India Meteorological Department's research section, monthly rainfall maps for the Indian land region are prepared (and published regularly since 1996) by subjectively analyzing the good quality monthly rainfall data in the form of hand-drawn contours (India Meteorological Department 2001a). These contour maps of monthly analyzed rainfall depict the large-scale distribution of monsoon rainfall during each month (Figs. 3a–d). Figures 4a–d and 5a–d show the objectively analyzed rainfall from GPCP and CMAP, respectively, for the corresponding monsoon months of 2001. Figure 6a–d shows the objectively analyzed monthly total rainfall prepared by the present method (NCMRWF) during the 2001 monsoon.

June is normally the month when the summer monsoon rainfall regime over India advances in a north-northwestward direction to cover the whole country. Figure 3a indicates that most of the country had more than 10 cm of rainfall. The west coast of India shows rainfall amounts from 20 to 120 cm. Compared to the IMD analysis, the GPCP analysis (Fig. 4a) underestimates the rainfall over the Indian landmass. The high-rainfall regions along the west coast, east coast, and in the central region are missing in the GPCP analysis. The rainfall distribution in the CMAP analysis (Fig. 5a) for June is closer to the IMD analysis than to the GPCP analysis. In the CMAP analysis, the west coast rain (50 cm) is higher (closer to the IMD) than GPCP. In CMAP, another area of heavy rainfall is seen near the head region of the Bay of Bengal (Bangladesh coast), where the highest contour reaches up to 110 cm. This high-rainfall region has to be verified from other independent datasets. Over oceanic regions of the Arabian Sea, Bay of Bengal, and the equatorial Indian Ocean, the rainfall distribution pattern seen in CMAP and GPCP generally agrees well, except when CMAP indicates slightly higher values. The NCMRWF analysis for June (Fig. 6a), which was performed at a higher resolution and with more local data, naturally shows more details compared to both GPCP and CMAP analyses. In the NCMRWF analysis, areas with rainfall amounts less than 10 cm match well with the IMD analysis and are close to CMAP. As seen in the IMD analysis, we find in the NCMRWF analysis two zones of maxima on the west coast of India. The IMD data for June indicate a station named Mangalore reporting 129 cm, while the maximum value of the contour in NCMRWF analysis is seen to be between 100 and 130 cm for the same region. The sharp gradient of rainfall between the west coast heavy-rain region and the rain shadow region to the east is brought out realistically in the NCMRWF analysis.

During the month of July, generally, the steady monsoon flow is well established, and the whole country receives more rainfall compared to June. In the IMD analysis (Fig. 3b), contours of 20 cm or more cover the whole Indian region. On the west coast, we find rainfall between 20 and 100 cm. We find another region with a maxima of 40 cm of rainfall along the monsoon trough (contours running in the southeast to northwest direction). The northeast part even shows a contour of 60 cm. The west coast rainfall amount in the GPCP (Fig. 4b) is only 30 cm compared to 100 cm in the IMD. In the CMAP analysis (Fig. 5b) the west coast rainfall goes up to 60 cm. The NCMRWF analysis produces realistically high values of rainfall along the west coast, as also reported in the IMD analysis. Rainfall along the monsoon trough in the NCMRWF is also seen realistically extending in the northwest to southeast direction. In the NCMRWF analysis, we get another region of maximum rainfall (up to 70 cm) along the foothills of the Himalayas mountain range. Contours in the IMD map indicate a region of high rainfall in that area, but the values are lower compared to the NCMRWF. This may be because the IMD does not include data from neighboring Nepal in its analysis. In the NCMRWF analysis, data from all neighboring countries received through the global telecommunication system are included. This high-rainfall region near the Himalayan foothills is also not seen in the GPCP and CMAP analyses.

Like other previous months, in August, the GPCP (Fig. 4c) rain over the west coast is underestimated (20 cm) compared to the IMD analysis (Fig. 3c). CMAP (Fig. 5c) is better than GPCP, as it reports a higher value of 30 cm. In the NCMRWF analysis (Fig. 6c), the 20-cm contour agrees well with the IMD analysis. On the west coast in the NCMRWF analysis, we find two maxima of 50 and 80 cm, which agree well with the IMD pattern. This realistic rainfall was not seen in the GPCP and CMAP analyses. In the NCMRWF analysis the rainfall along the trough and the rain-shadow regions are also reproduced realistically when compared to the IMD analysis. Due to the higher resolution of the grid selection in the NCMRWF analysis, it has more features compared to the GPCP and CMAP patterns. Similar to July, August also has a very heavy rainfall region along the foothills of the Himalayas with 100 cm of rainfall. This high-rain region is noticed in the IMD and CMAP, but their magnitudes are seen to be much lower.

During September, the typical monsoon flow system generally starts to weaken and the associated reduced rainfall pattern withdraws from the northwest part of India toward the south and southeast (Fig. 3d). GPCP (Fig. 4d), CMAP (Fig. 5d), and NCMRWF (Fig. 6d) also show similar patterns of a withdrawn monsoon. The coastal Kerala region in the NCMRWF analysis shows 40-cm rains matching with the IMD values. In the NCMRWF analysis, another region of high rainfall in northeast India is also reproduced. Heavy rainfall up to 50 cm along the foothills of the Himalayas in the NCMRWF analysis is missing in the GPCP and CMAP analyses. This rainfall along the foothills is also noticed in the IMD analysis. Over the oceanic regions, the NCMRWF analysis agrees with the CMAP and GPCP patterns in capturing the oceanic monsoon rainfall regime during all four months. However, the maximum values are realistically higher in the CMAP and GPCP, as estimates from microwave data were used there. When a nine-point smoother is applied to the NCMRWF analyzed data (Figs. 7a–d), it loses many details, as seen in Figs. 6a–d. For June, the smoothed version of the NCMRWF analysis generally agrees very well with CMAP over both land and ocean. The former, however, shows higher values on the west coast, which is also closer to the IMD. The July chart retains more regional details as compared to CMAP and GPCP. The August chart looks very similar to the GPCP and CMAP patterns, except along the west coast, where it shows higher values (closer to IMD). The September smoothed field is very similar to the CMAP analysis.

The observed seasonal total rainfall from the IMD, NCMRWF, GPCP, and CMAP analyses are shown in Figs. 8a–d. Very good agreement between the IMD and NCMRWF analyses is seen over the west coast of India (300-cm contour). Along the foothills of the Himalayas, several pockets of 200-cm rainfall are seen in both analyses. In the GPCP and CMAP analyses, the maximum west coast rainfall reaches only 80 and 160 cm, respectively, which is very low compared to the IMD observations. Over the oceanic region, the areas of heavy and light rains match well in all three (NCMRWF, GPCP, and CMAP) analyses. The location of the equatorial maximum in rainfall (5°S, 92°E) is brought out in all the three analyses. However, in the maxima there is a 20-cm difference in the rainfall values among the three analyses. Both GPCP and CMAP use microwave-based rain data. However, the NCMRWF analysis is only from IR-based estimation, which leads to an underestimation. Figure 9 shows the quantitative comparison of rainfall amounts from different analyses at important (synoptic) locations within the Indian monsoon region. The regions selected are the northern part of the west coast (NWC) and the southern part of the west coast (SWC), which are regions of heavy rainfall associated with active phases of the monsoon. The eastern coast near the head of the northern Bay of Bengal (ECHNB) is another region associated with the formation of monsoon lows and depressions. The fourth region is the central part of India associated with the monsoon trough (CIT). For the west coast region, we find that the NCMRWF analysis agrees well with IMD and is better than CMAP and GPCP. For the remaining two regions, all four analyses agree well with each other.

4. Summary

Daily large-scale monsoon rainfall data have been prepared by objective analysis using available real-time rainfall estimates from satellites and rain gauges. The characteristics of the output from this analysis system have been examined by comparing the accumulated monthly observed rainfall from other available independent popular datasets like the GPCP and CMAP analyses. The monthly data prepared from daily analyses are also compared to IMD monthly rainfall maps produced by subjective analysis. It is seen that with available real-time data from INSAT-1D and rain gauges, it is possible to construct a usable large-scale rainfall map on regular latitude–longitude grids. This high-resolution NCMRWF analysis uses more local rain gauge data, and it is able to more realistically produce details of the typical features of the Indian summer monsoon rainfall patterns. The magnitude and distribution of the orographic rainfall near the west coast of India is very different (realistic) when compared to both the GPCP and CMAP patterns. The regions of heavy and light rain are demarcated clearly over the Indian landmass. Over the oceanic regions of the Arabian Sea, Bay of Bengal, and the equatorial Indian Ocean, the agreement of the analysed rainfall at the monthly scale is quite good compared to the other two datasets. For NWP and other model simulation verification of large-scale rainfall, this dataset will be useful.

5. Future work

The satellite rainfall estimates used here are from the INSAT-1D geostationary satellite, the last one in the INSAT-1 series. It was operational in July 1990 and it continued to provide IR data until April 2002. The Indian satellite program has a plan for continued launching of geostationary satellites with improved meteorological satellite data capabilities. The recently launched METSAT has an IR channel at 8-km spatial resolution. During 2003–04, two other Indian satellites, INSAT-3A and INSAT-3D, will also be launched and will have IR resolutions of 8 and 4 km, respectively. The geostationary satellite METEOSAT-5 located at 63°E is expected to continue operating over the Indian Ocean sector until the end of year 2003. There is a plan for moving METEOSAT-6 or -7 to replace METEOSAT-5 over the Indian Ocean region. Rainfall estimates from these current and future satellites will be useful for producing merged daily rainfall data sets on regular latitude–longitude grids. Some recent rainfall algorithms from IR data (Falkovich et al. 2000) remove the cirrus cloud influence on estimated rain. These types of techniques have to be tested for the Indian Ocean region. The IR-based rains from geostationary satellites should be calibrated by obtaining relationships between microwave rain rates and IR properties of cloud systems (Jobard and Desbois 1994; Adler et al. 1994). New knowledge of satellite rain estimation from the European project EURAINSAT will help in rainfall estimation on the geostationary scale with METEOSAT Second Generation (MSG) Spinning Enhanced Visible and Infrared Image (SEVIRI).

The India Meteorological Department has 556 surface observatories (one in each district) where rainfall is measured regularly. Of these, only around 35% of the station data is received in real time digitally via computer. Apart from this, there are 3540 rain gauge stations maintained by state governments, and another 206 agro-meteorological stations take rainfall observations. They are, however, kept in manuscript form and later transferred to computer. As the communication capabilities have improved in last few years, the requirement to get the rest of the data in real time in digital form should become possible soon. When this huge amount of data is used, the quality of the analyzed rain will improve substantially. In order to obtain more real-time rain gauge data, the reporting of daily (24-h collection) rain gauge data (especially zero precipitation amounts) needs to be standardized around the globe. The organized, constructed 24-h rainfall database from subdaily reports of precipitation (3-hourly, 6-hourly, etc.) coming regularly from good stations can contribute significantly as an additional data source. This will require coordination efforts at the national and international levels. Techniques developed to delineate areas of no-rain will be useful where less rain gauge data are available (Ebert and Weymouth 1999). Useful data are likely to come from the international Global Precipitation Measurement (GPM) mission and the Indian–French satellite Megha-Tropiques, where the experience and knowledge gained from the Tropical Rainfall Measuring Mission (TRMM) and SSM/I Defense Meteorological Satellite Program (DMSP) will be very useful.

Acknowledgments

We are grateful to all the authors and creators of the two datasets from the GPCP and CMAP projects used in this study for providing the monthly total values through ftp. The Satellite Meteorology division of the India Meteorological Department provided satellite rainfall estimates from INSAT in real time. We are grateful to Dr. Rajeswar Rao of the SATMET division, IMD for providing satellite rainfall data from the archives for the few missing days. We are thankful to Mr. Brian Mackey for many useful suggestions. The work of Prof. T. N. Krishnamurti was partially supported by NASA TRMM Grant NAG5 9662.

REFERENCES

  • Adler, R. F., Huffman G. F. , and Keehn P. R. , 1994: Global tropical rain estimation from microwave-adjusted geosynchronous IR data. Remote Sens. Rev., 11 , 125152.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Adler, R. F., Kidd C. , Petty G. , Morissey M. , and Goodman H. M. , 2001: Intercomparison of global precipitation products: The third precipitation intercomparison project (PIP-3). Bull. Amer. Meteor. Soc., 82 , 13771396.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and Meisner B. N. , 1987: The relationship between large-scale convective rainfall and cold cloud over the Western Hemisphere during 1982–84. Mon. Wea. Rev., 115 , 5174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., Rao A. V. R. K. , and Kelkar R. R. , 1989: Large-scale precipitation and outgoing longwave radiation from INSAT-1B during the 1986 southwest monsoon season. J. Climate, 2 , 619628.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beranger, K., Barnier B. , and Crepon M. , 2001: Comparison of 10 precipitation climatologies over the global ocean. Proc. WCRP/SCOR Workshop on Intercomparison and Validation of Ocean–Atmosphere Flux Fields, Potomac, MD, World Climate Research Program, Scientific Committee on Ocean Research, and Cosponsors, WCRP-115, WMO/TD 1083, 215–218.

    • Search Google Scholar
    • Export Citation
  • Cherubini, T., Ghelli A. , and Lalaurette F. , 2001: Verification of precipitation forecast over the Alpine region using a high density observation network. ECMWF Tech. Memo. 340, 18 pp.

    • Search Google Scholar
    • Export Citation
  • Ebert, E. E., and Weymouth G. T. , 1999: Incorporating satellite observations of “no rain” in an Australian daily rainfall analysis. J. Appl. Meteor., 38 , 4456.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Falkovich, A., Lord S. , and Treadon R. , 2000: A new methodology of rainfall retrieval from indirect measurement. Meteor. Atmos. Phys., 75 (3–4) 217232.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grassl, H., Jost V. , Schulz J. , Kumar R. , Bauer P. , and Schlussel P. , 2000: A climatological atlas of satellite derived air–sea interaction parameters over the worlds ocean. Rep. 312, Max Planck Institute for Meteorology, Hamburg, Germany, 120 pp.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., Su X. , Kanamitsu M. , and Schemm J. , 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and and Coauthors, 1997: The GPCP combined precipitation datasets. Bull. Amer. Meteor. Soc., 78 , 520.

  • India Meteorological Department, 2001a: Climate Diagnostic Bulletin of India—June, July, August, September 2001: Near Real Time Analysis. No. 64–67, Rep. of National Climate Centre, ADGM(R), IMD, Pune, India, 74 pp.

    • Search Google Scholar
    • Export Citation
  • India Meteorological Department, 2001b: Climate Diagnostic Bulletin of India—Seasonal Monsoon Season (June–September) 2001: Near Real Time Analysis. No. 22, Rep. of National Climate Centre, ADGM(R), IMD, Pune, India, 32 pp.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., 1992: Tropical rainfall: A comparison of satellite-derived rainfall estimates with model precipitation forecasts, climatologies, and observations. Mon. Wea. Rev., 120 , 448462.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., and Xie P. , 1999: CAMS–OPI: A global satellite–rain gauge merged product for real-time precipitation monitoring applications. J. Climate, 12 , 33353342.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., Gruber A. , Curtis S. , Huffman G. J. , Adler R. F. , and Xie P. , 2001: Evaluation and comparison of the CMAP and GPCP oceanic precipitation analyses. Proc. WCRP/SCOR Workshop on Intercomparison and Validation of Ocean–Atmosphere Flux Fields, Potomac, MD, World Climate Research Program, Scientific Committee on Ocean Research, and Cosponsors, WCRP-115, WMO/TD 1083, 215–218.

    • Search Google Scholar
    • Export Citation
  • Jobard, I., and Desbois M. , 1994: Satellite estimation of tropical precipitation using the METEOSAT and SSM/I data. Atmos. Res., 34 , 285298.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kripalini, R. H., Singh S. V. , and Arkin P. A. , 1991: Large scale features of rainfall and OLR over India and adjoining regions. Contrib. Atmos. Phys., 64 , 159161.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., Cocke S. , Pasch R. , and Low-Nam S. , 1983: Precipitation estimates from rain gauge and satellite observations, summer MONEX. FSU Rep. 83-7, Department of Meteorology, The Florida State University, 373 pp.

    • Search Google Scholar
    • Export Citation
  • Mills, G. A., Weymouth G. , Jones D. , Ebert E. E. , Manton M. , Lorkin J. , and Kelly J. , 1997: A national objective daily rainfall analysis system. Techniques Development Rep. 1, BMRC, Melbourne, VIC, Australia, 30 pp.

    • Search Google Scholar
    • Export Citation
  • Mitra, A. K., Bohra A. K. , and Rajan D. , 1997: Daily rainfall analysis for Indian summer monsoon region. Int. J. Climatol., 17 , 10831092.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tripoli, G. J., and Krishnamurti T. N. , 1975: Low-level flows over the GATE area during summer 1972. Mon. Wea. Rev., 103 , 197216.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weymouth, G., Mills G. A. , Jones D. , Ebert E. E. , and Manton M. J. , 1999: A continental scale daily rainfall analysis. Aust. Meteor. Mag., 48 , 169179.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and Arkin P. A. , 1997: Global Precipitation: A 17-year monthly analysis based on observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

(a) Distribution of available real-time rain gauge data stations in the analysis region on a typical day. (b) Time series indicating number of daily available rain gauge data during 122 days (1 Jun–30 Sep) of the 2001 monsoon season. Solid line is data from the full region; dashed line is data from Indian region

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 2.
Fig. 2.

(a) Different aspects of variability of observed rainfall during the 2001 monsoon season. Solid dark line is daily maximum rain reported by rain gauge in mm day−1; solid light line is daily variance of rain computed from reported rain gauges; dashed line is daily computed rmse of analyzed rain in mm × 10. (b) Daily variance of rainfall in observed data and final analysis during the 2001 monsoon season. Dashed line is daily variance computed from rain gauge data only; solid light line is daily variance computed from analyzed rain data; solid dark line is ratio of variance of analyzed and observed rain expressed in percentage

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 3.
Fig. 3.

IMD monthly observed rainfall during the 2001 monsoon season

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 4.
Fig. 4.

GPCP monthly observed rainfall during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 5.
Fig. 5.

CMAP monthly obs rainfall during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 6.
Fig. 6.

NCMRWF monthly observed rainfall during the 2001 monsoon season: (a) Jun, (b) Jul, (c) Aug, and (d) Sep. Contours are at 10, 20, … , 70, 100, 130 cm in (a) and (b), and every 10 cm in (c) and (d). Shading is between 5 and 10 cm in all panels

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 7.
Fig. 7.

NCMRWF monthly observed rainfall (smoothed) during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 8.
Fig. 8.

2001 monsoon seasonal total rainfall of (a) IMD, (b) NCMRWF, (c) GPCP, and (d) CMAP (Jun–Sep 2001 total rainfall)

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Fig. 9.
Fig. 9.

Quantitative comparison of rainfall amounts from various analyses of different regions for (a) Jun, (b) Jul, (c) Aug, and (d) Sep of the 2001 monsoon season

Citation: Journal of Hydrometeorology 4, 5; 10.1175/1525-7541(2003)004<0769:DRFTIM>2.0.CO;2

Save
  • Adler, R. F., Huffman G. F. , and Keehn P. R. , 1994: Global tropical rain estimation from microwave-adjusted geosynchronous IR data. Remote Sens. Rev., 11 , 125152.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Adler, R. F., Kidd C. , Petty G. , Morissey M. , and Goodman H. M. , 2001: Intercomparison of global precipitation products: The third precipitation intercomparison project (PIP-3). Bull. Amer. Meteor. Soc., 82 , 13771396.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., and Meisner B. N. , 1987: The relationship between large-scale convective rainfall and cold cloud over the Western Hemisphere during 1982–84. Mon. Wea. Rev., 115 , 5174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arkin, P. A., Rao A. V. R. K. , and Kelkar R. R. , 1989: Large-scale precipitation and outgoing longwave radiation from INSAT-1B during the 1986 southwest monsoon season. J. Climate, 2 , 619628.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beranger, K., Barnier B. , and Crepon M. , 2001: Comparison of 10 precipitation climatologies over the global ocean. Proc. WCRP/SCOR Workshop on Intercomparison and Validation of Ocean–Atmosphere Flux Fields, Potomac, MD, World Climate Research Program, Scientific Committee on Ocean Research, and Cosponsors, WCRP-115, WMO/TD 1083, 215–218.

    • Search Google Scholar
    • Export Citation
  • Cherubini, T., Ghelli A. , and Lalaurette F. , 2001: Verification of precipitation forecast over the Alpine region using a high density observation network. ECMWF Tech. Memo. 340, 18 pp.

    • Search Google Scholar
    • Export Citation
  • Ebert, E. E., and Weymouth G. T. , 1999: Incorporating satellite observations of “no rain” in an Australian daily rainfall analysis. J. Appl. Meteor., 38 , 4456.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Falkovich, A., Lord S. , and Treadon R. , 2000: A new methodology of rainfall retrieval from indirect measurement. Meteor. Atmos. Phys., 75 (3–4) 217232.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grassl, H., Jost V. , Schulz J. , Kumar R. , Bauer P. , and Schlussel P. , 2000: A climatological atlas of satellite derived air–sea interaction parameters over the worlds ocean. Rep. 312, Max Planck Institute for Meteorology, Hamburg, Germany, 120 pp.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., Su X. , Kanamitsu M. , and Schemm J. , 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., and and Coauthors, 1997: The GPCP combined precipitation datasets. Bull. Amer. Meteor. Soc., 78 , 520.

  • India Meteorological Department, 2001a: Climate Diagnostic Bulletin of India—June, July, August, September 2001: Near Real Time Analysis. No. 64–67, Rep. of National Climate Centre, ADGM(R), IMD, Pune, India, 74 pp.

    • Search Google Scholar
    • Export Citation
  • India Meteorological Department, 2001b: Climate Diagnostic Bulletin of India—Seasonal Monsoon Season (June–September) 2001: Near Real Time Analysis. No. 22, Rep. of National Climate Centre, ADGM(R), IMD, Pune, India, 32 pp.

    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., 1992: Tropical rainfall: A comparison of satellite-derived rainfall estimates with model precipitation forecasts, climatologies, and observations. Mon. Wea. Rev., 120 , 448462.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., and Xie P. , 1999: CAMS–OPI: A global satellite–rain gauge merged product for real-time precipitation monitoring applications. J. Climate, 12 , 33353342.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Janowiak, J. E., Gruber A. , Curtis S. , Huffman G. J. , Adler R. F. , and Xie P. , 2001: Evaluation and comparison of the CMAP and GPCP oceanic precipitation analyses. Proc. WCRP/SCOR Workshop on Intercomparison and Validation of Ocean–Atmosphere Flux Fields, Potomac, MD, World Climate Research Program, Scientific Committee on Ocean Research, and Cosponsors, WCRP-115, WMO/TD 1083, 215–218.

    • Search Google Scholar
    • Export Citation
  • Jobard, I., and Desbois M. , 1994: Satellite estimation of tropical precipitation using the METEOSAT and SSM/I data. Atmos. Res., 34 , 285298.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kripalini, R. H., Singh S. V. , and Arkin P. A. , 1991: Large scale features of rainfall and OLR over India and adjoining regions. Contrib. Atmos. Phys., 64 , 159161.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., Cocke S. , Pasch R. , and Low-Nam S. , 1983: Precipitation estimates from rain gauge and satellite observations, summer MONEX. FSU Rep. 83-7, Department of Meteorology, The Florida State University, 373 pp.

    • Search Google Scholar
    • Export Citation
  • Mills, G. A., Weymouth G. , Jones D. , Ebert E. E. , Manton M. , Lorkin J. , and Kelly J. , 1997: A national objective daily rainfall analysis system. Techniques Development Rep. 1, BMRC, Melbourne, VIC, Australia, 30 pp.

    • Search Google Scholar
    • Export Citation
  • Mitra, A. K., Bohra A. K. , and Rajan D. , 1997: Daily rainfall analysis for Indian summer monsoon region. Int. J. Climatol., 17 , 10831092.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tripoli, G. J., and Krishnamurti T. N. , 1975: Low-level flows over the GATE area during summer 1972. Mon. Wea. Rev., 103 , 197216.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weymouth, G., Mills G. A. , Jones D. , Ebert E. E. , and Manton M. J. , 1999: A continental scale daily rainfall analysis. Aust. Meteor. Mag., 48 , 169179.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and Arkin P. A. , 1997: Global Precipitation: A 17-year monthly analysis based on observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    (a) Distribution of available real-time rain gauge data stations in the analysis region on a typical day. (b) Time series indicating number of daily available rain gauge data during 122 days (1 Jun–30 Sep) of the 2001 monsoon season. Solid line is data from the full region; dashed line is data from Indian region

  • Fig. 2.

    (a) Different aspects of variability of observed rainfall during the 2001 monsoon season. Solid dark line is daily maximum rain reported by rain gauge in mm day−1; solid light line is daily variance of rain computed from reported rain gauges; dashed line is daily computed rmse of analyzed rain in mm × 10. (b) Daily variance of rainfall in observed data and final analysis during the 2001 monsoon season. Dashed line is daily variance computed from rain gauge data only; solid light line is daily variance computed from analyzed rain data; solid dark line is ratio of variance of analyzed and observed rain expressed in percentage

  • Fig. 3.

    IMD monthly observed rainfall during the 2001 monsoon season

  • Fig. 4.

    GPCP monthly observed rainfall during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

  • Fig. 5.

    CMAP monthly obs rainfall during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

  • Fig. 6.

    NCMRWF monthly observed rainfall during the 2001 monsoon season: (a) Jun, (b) Jul, (c) Aug, and (d) Sep. Contours are at 10, 20, … , 70, 100, 130 cm in (a) and (b), and every 10 cm in (c) and (d). Shading is between 5 and 10 cm in all panels

  • Fig. 7.

    NCMRWF monthly observed rainfall (smoothed) during the 2001 monsoon season (contour interval is 10 cm; shading from 5 to 10 cm)

  • Fig. 8.

    2001 monsoon seasonal total rainfall of (a) IMD, (b) NCMRWF, (c) GPCP, and (d) CMAP (Jun–Sep 2001 total rainfall)

  • Fig. 9.

    Quantitative comparison of rainfall amounts from various analyses of different regions for (a) Jun, (b) Jul, (c) Aug, and (d) Sep of the 2001 monsoon season

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 567 150 10
PDF Downloads 377 103 9