Dear Bagher,

You are asking a very pertinent but quite difficult question: if your study area is not equipped with rain gages (or has only one), it will be difficult to properly evaluate the performance of TRMM measurements because precipitations are expected to be highly dependent on altitude in mountainous environments.

However, as you point out, even if such local observations were available, the main difficulty would be to deal with the major difference in the scale of representativity between the satellite-based TRMM instrument, which only measures average radiation signals for large areas (25 by 25 km, corresponding to individual instrument pixels), and rain gages that generate in situ, quite local measurements. Moreover, the reliability of observations from both instrumental approaches are constrained by the accuracy of the models used to interpret the space-based measurements, and by intrinsic limitations in the effectiveness and accuracy of rain gages, especially in mountainous environments, respectively.

In any case, since you also indicate that you only have one precipitation ground station in the area anyway, you may want to look at alternative approaches. One possibility would be to delimit the boundary of a large enough river basin (say, 50,000 square km or more) for which you do have hydrographic data (runoff) as a function of time. The main advantage of runoff is that it is also an integral estimator, since it accumulates water over the watershed.

You could then feed a model of the water balance for that watershed with inputs from TRMM and other sources to estimate runoff, infiltration, and evapotranspiration in order to compute the outflow predicted by this model and compare it to the actual river discharge. Of course, this approach involves multiple sources of input data, as well as a model which can have its own limitations, but it would at least give you a sound basis for establishing the coherency and consistency (or lack thereof) of your tools and data sets. It would also permit you to conduct sensitivity analyses to determine the relative importance of each input, and indeed the role of TRMM data in this context.

Such an approach would probably not be applicable (or reliable) at the scale of 4 TRMM pixels because the actual size and shape of your study area are unlikely to closely match the region observed by remote sensing. If you deal with a larger area, these differences in coverage tend to reduce in proportion to the ratio of the actual basin area over the area observed by the satellite.

Another independent approach might consist in not trying to "validate" the TRMM data itself, but to investigate to what extent using those results yields meaningful results. For instance, if the area is at least partially exploited for agricultural production, you might develop a model of crop productivity using generic climatological data to predict yields, and then establish whether using TRMM data instead of some measured or assumed precipitation climatology actually improves the forecasting skill of the model. If the answer is positive, you will have shown that TRMM data may not be actually validated, but still very useful for practical applications.

A less satisfactory but simpler approach might be to establish whether there is any correlation between TRMM-derived precipitation levels and these (or some other) precipitation-dependent measurable quantities, and to explore whether these variables vary synchronously over long periods of time. This exercise could perhaps serve as a starting point for the second approach, as it would generate preliminary results while imposing fewer and less demanding requirements on the inputs.

Best regards, Michel.

If there are no in situ data, other researches from literature or book may be considered to identify the precipitation data from TRMM. However, during my study, which maybe macro, the TRMM have a high correlation with ground data. Hope it will help you.

Dear Jianbo

Thank you for your comment.

As mentioned a bow, we have one rain gauge station nearest to the study area outlet at the height of 1729 meter which most likely is not capable to catch most of precipitation because of the study area high expansion. On the other hand, according to the climate variability, I think a ground station cannot be supported a more expanded area, especially in the area with the highest elevation change. So, what is important is that the significant difference between both of them results.

Regards

Bagher 

Dear Bagher,

You are asking a very pertinent but quite difficult question: if your study area is not equipped with rain gages (or has only one), it will be difficult to properly evaluate the performance of TRMM measurements because precipitations are expected to be highly dependent on altitude in mountainous environments.

However, as you point out, even if such local observations were available, the main difficulty would be to deal with the major difference in the scale of representativity between the satellite-based TRMM instrument, which only measures average radiation signals for large areas (25 by 25 km, corresponding to individual instrument pixels), and rain gages that generate in situ, quite local measurements. Moreover, the reliability of observations from both instrumental approaches are constrained by the accuracy of the models used to interpret the space-based measurements, and by intrinsic limitations in the effectiveness and accuracy of rain gages, especially in mountainous environments, respectively.

In any case, since you also indicate that you only have one precipitation ground station in the area anyway, you may want to look at alternative approaches. One possibility would be to delimit the boundary of a large enough river basin (say, 50,000 square km or more) for which you do have hydrographic data (runoff) as a function of time. The main advantage of runoff is that it is also an integral estimator, since it accumulates water over the watershed.

You could then feed a model of the water balance for that watershed with inputs from TRMM and other sources to estimate runoff, infiltration, and evapotranspiration in order to compute the outflow predicted by this model and compare it to the actual river discharge. Of course, this approach involves multiple sources of input data, as well as a model which can have its own limitations, but it would at least give you a sound basis for establishing the coherency and consistency (or lack thereof) of your tools and data sets. It would also permit you to conduct sensitivity analyses to determine the relative importance of each input, and indeed the role of TRMM data in this context.

Such an approach would probably not be applicable (or reliable) at the scale of 4 TRMM pixels because the actual size and shape of your study area are unlikely to closely match the region observed by remote sensing. If you deal with a larger area, these differences in coverage tend to reduce in proportion to the ratio of the actual basin area over the area observed by the satellite.

Another independent approach might consist in not trying to "validate" the TRMM data itself, but to investigate to what extent using those results yields meaningful results. For instance, if the area is at least partially exploited for agricultural production, you might develop a model of crop productivity using generic climatological data to predict yields, and then establish whether using TRMM data instead of some measured or assumed precipitation climatology actually improves the forecasting skill of the model. If the answer is positive, you will have shown that TRMM data may not be actually validated, but still very useful for practical applications.

A less satisfactory but simpler approach might be to establish whether there is any correlation between TRMM-derived precipitation levels and these (or some other) precipitation-dependent measurable quantities, and to explore whether these variables vary synchronously over long periods of time. This exercise could perhaps serve as a starting point for the second approach, as it would generate preliminary results while imposing fewer and less demanding requirements on the inputs.

Best regards, Michel.

Hi - Michel has written a comprehensive answer to which I need not add much.  It is the case that several researchers in Iran have asked about TMPA in specific localities, and the results are somewhat mixed.  You might want to ask around your colleagues and see what their results are.  One general comment is that mountainous areas are difficult because the passive microwave retrievals have a hard time with a) estimating orographic enhancement, and b) making any estimates when there is snow on the surface.  I can say that looking around Zabol seemed to show a questionable trend in TMPA, perhaps due to changes in the satellite constellation that provides the data.  To reiterate Michel's excellent point, in the absence of validating gauges, you have to fall back on consistency with other data sets, with consistency likely being better for larger areas.  By the way, if there's snowfall involved, you need to consider the ability of the particular precipitation gauges to accurately record the snowfall.

Good luck with the study!  Ask as you have other questions.

George

Dear Prof. Michel and George

Thank you very much for your professional comments.

Unfortunately the variable climate, especially in the local scales (man-made climate), leads to the phenomena hard to understand and the results be not reliable. More importantly, the lack of well situated rain gauges, especially in the arid and semi-arid areas, caused to we are faced with a serious issue in our research. Formerly (before doing this study), I have thought that using remote sensing techniques could be a suitable alternative to that end. I hope that progress in the remote sensing instrument will be overcome to these limitations.

With Kind Regards

Bagher