Rainfall analysis for Indian monsoon region from merged satellite and dense raingauge observations: Evaluation of monsoon rain

Revise Title: Rainfall analysis for Indian monsoon region from merged dense rain gauge observations and satellite estimates: Evaluation of monsoon rainfall features (DS0601)

Point wise reply to comments of Referee 1

Comment 1:It is surprising that the authors have not chosen to compare their ‘product’ with the TRMM data, say 3B42, which is 3 hourly and 0.25 deg, Granted that TRMM observations over the Western Ghats are not adequate to resolve the variability caused by orography. What about the rest of India ?

Reply 1: In the present work TRMM data is not used for comparison. The TRMM data has got some limitations over the land areas due to factors like topography, meso scale convection etc. The new analysis is compared with the standard global product of GPCC.

Comment 2: Importance of using INSAT data is not completely clear to me. It is used for filling gaps over mostly sea and neighboring countries. All the data over land in India appears to come from their mapping procedure, Could a merging procedure over land have helped extract the advantages of both sets of observations, rather than be limited by the choice only one data source ?. The impression one gets from the title, the abstract and conclusion is that gauge data and INSAT data have been combined (which have been more fantastic) and that, from main body of the manuscript is not adequate.
Reply 2: INSAT data is used for filling gaps over the sea and other data gap areas. The main purpose of this study is to generate daily rainfall analysis which could be used for real time real-time operational application as well as for validation of NWP models over Indian monsoon region. Previous studies suggest (references are given in the manuscript) that the performance of INSAT QPE over the oceanic areas are reasonably good but performance deteriorates over the land areas. In order to extract the advantage of raingauge data over the land areas and satellite estimates over the ocean, the gridded raingauge data over the land areas are merged with the QPE (at the same resolution) over the ocean and over the raingauge gap areas. In section 2 of the revised version, a detailed discussion on the objective analysis procedure is added and very clearly described how gridded rainfall data is merged with the satellite estimates.
To avoid confusion regarding merged or combined, the title of the paper is accordingly changed. Abstract and other relevant portions of the manuscript are also revised.

Comment 3: Given that the mapping of data from rain gauges to a grid remains a open problem, more space needs to be allocated to discussing the use of the weighting function, namely to address issues such as: (i) What other method have the author tried? (ii) why this particular weighting function ? (iii) What are the error bars for their estimation/mapping procedure ? are they more along/across the mountains and less along the flats ? Surely the density of stations (fig 1) plays an important role in governing the error. The choice of R=200 km seems to on higher side for mappings on to I deg box. True that there is a reference provided, but a brief description of this choice would be helpful for readers, and perhaps make the manuscript more self sufficient. On these aspects of (and several other related to) point to area mapping the presentation is less than satisfactory.

Reply 3: The methodology (section 2) is revised addressing the concerns of the referee. Reference of other works on rainfall analysis is suitably cited in section 1 and 2. In a very recent study, Roy Bhowmik et al. (2005) applied the technique of correlation co-efficient (CC) as the function of distance for objectively determination of radius of influence of rainfall over Indian monsoon region. The method is tested for different synoptic situation as well as for the different sectors of the country to investigate its geographical variation. The study showed that, in general, the radius of influence for Indian monsoon region is 200 km. The result of this study is referred in the revised version (in section 2)

Comment 4: It would have been great to have seen the analysis for more years. Just one year (and that too only about 90 days) is too small a sample to make any objective assessment about the reported comparison. Perhaps this persistent thought in my mind of wishing to see the analysis and comparison made over a large sample, made section 5 appear a tad bit forced.

Reply 4: During monsoon 2002, INSAT 1D was defunct and no INSAT QPE data was available. The analysis for monsoon 2003 with the INSAT (KALPANA-1) QPE is included in the revised version (in section 4). A new section (section 3) is added where characteristics of monsoon 2001 and 2003 are briefly described.

Comment 5: As a minor comment, the author say that the analysis was performed over a large area, but most of their inferences are about the Western Ghats.

Reply 5: The discussion covered Western Ghats (orographic rainfall ) as well as rainfall associated with other monsoon synoptic systems. The discussion on orographic rainfall has got more importance in the manuscript due to the fact that significant improvement is noticed for orographic rainfall in the new analysis.

Point wise reply to comments of Referee 2

Comment 1: The description of the methodology is very poor. I am not clear how the satellite observation is brought into the analysis. The estimated rainfall value at a grid pint is the weighted average of observation at N sites as given by the first equation. The satellite data is available on a 1 deg grid, how is this included in the analysis ? Are the satellite observations are also considered as one of the N site observation. This neds clear description. Methodology

Reply 1: The methodology (section 2) is revised where a detailed discussion on the objective analysis procedure is added and very clearly described how gridded rainfall data is merged with the satellite estimates.

Comment 2: The radius of influence is fixed as 200 km. How is this determined? And more importantly how sensitive are the final estimates to the choice of this radius? And are their objective method to obtain this?

Reply 2: In a very recent study, Roy Bhowmik et al. (2005) applied the technique of correlation co-efficient (CC) as the function of distance for objectively determination of radius of influence of rainfall over Indian monsoon region and found that radius of influence for Indian monsoon region is 200 km. The results of above study is referred in the revised version (in section 2).

Comment 3: The comparison of the proposed method with other products is done visually. While this is good, I think to make a case the author have to provide quantitative measures of comparison, e.g., RMSE. This is one of the shortcomings of the [paper.
Reply 3: In a recent study Rajeevan et al. (2005) made an quantitative assessment of the high resolution ( 1 ^o x1 ^o lat/long ) daily gridded rainfall data developed for Indian region. The study considered daily land station observations for the period from 1951 to 2003 covering the domain from 6.5 ^o N to 37.5 ^o N, 67.5 ^o E to 101.5^o E. The study reported that the difference between the IMD and GPCP data are positive along the west coast implying that GPCP data set underestimates the heavy rainfall amounts along the west coast. Similarly over the eastern parts of central India, GPCP data underestimates the heavy rainfall amounts. Over the most parts of the country the differences are of the order of 50 mm. The correlations between GPCP and IMD rainfall data exceeds 0.4 over central and northwest India. In the revised version, section 1 is modified referring the work of Rajeevan (2005). No further quantitative assessment of these data set is made in the present work. In order to extract the advantage of station observations and to cover the data gap over the sea area, attempt has been made in this paper to generate daily rainfall analysis for the domain from 0 ^o N to 40 ^o N, 60 ^o E to 100^o E over Indian monsoon region merging satellite estimates and land station observations. This is the monsoon domain for which validation of operational NWP models is very important. In this paper, we subjectively examine the potential of this merged gridded dataset to capture characteristic features (large scale as well as mesoscale) of Indian summer monsoon.

For the validation purpose, maps of monthly rainfall analysis) are also compared with the station rainfall plots.

Comment 4: In the process of merging, what is the contribution of satellite data? In other words, if one performs the analysis using the weighted average equation based on only on the observation data and only on the satellite data how different wil they be ? This will provide a sense of relative importance of satellite observations.

Reply 4: This issue was addressed in our very recent study (Roy Bhowmik and Sen Roy, 2006) where an quantitative assessment of QPE in relation to gridded rainfall (based on raingauge observation only ) over the land areas has been documented. The result of this study is referred in the revised version (in section 2).

Comment 5: The Methodology proposed is a nonparametric ‘local’ functional estimation method.

Reply 5: This point has been discussed in section 2 referring the results of our previous study (Roy Bhowmik et al., 2005) regarding computation of radius of influence. Reference of other works on rainfall analysis is also suitably cited in section 1 and 2.

The manuscript is revised in-corporating comments and suggestions of both the referees. The language is further improved. Title of the manuscript is suitably changed as “Rainfall analysis for Indian monsoon region from merged dense rain gauge observations and satellite estimates: Evaluation of monsoon rainfall features”.