تهیه نقشه‌های هم‌دما و هم‌رطوبت جوّ با استفاده از باندهای گرمایی تصاویر ماهواره‌ای مادیس

In recent years, extraction of atmospheric temperature and humidity profiles from thermal bands of satellite images is a common practice. The most deployed images for this task have been acquired by Moderate Resolution Imaging Spectroradiometer (MODIS). This sensor which is onboard of Terra and Aqua platforms consists of a spectroradiometer with 36 bands in visible (VIS), near infra-red (NIR) up to thermal infra-red (TIR) region (0.405 – 14.385 ( )). The combination of sixteen infrared spectral channels is suitable for sounding temperature and humidity profiles in the atmosphere with a relatively acceptable precision. Different daily atmospheric products of MODIS images are in access on MODIS site. Atmospheric temperature and humidity profiles are two of these products that were used in this work. These products are named MOD07 and MYD07 for Terra and Aqua platforms respectively. Other MODIS products used in this work were MOD35 and MYD35. These products are mainly used for detection of pixel cloud contamination as well as for detection of aerosol concentration in pixel.
Presently, maps of temperature and humidity isopleths from MODIS images are being extracted by applying some global algorithms. In these algorithms, global profiles of temperature, moisture, and ozone are being used in the calculations, where it is believed that these algorithms are not usually valid in regional scales. It is believed that the regional algorithms can boost the precision of aforementioned maps. In this work, 5 MODIS images of 8 and 11, June 2004 and 8, 15 and 22 June 2007 and their corresponding temperature and moisture profiles were used for modeling and  4 MODIS images of June 2, 6, 7 and 21 of 2007 were used for evaluation. In order to prepare a temperature and moisture profile as an initial guess in model, a 5 year (2004–2008) radiosonde data set consisting of 240 temperature profiles all interpolated for the times of satellite overpass in June were averaged. The aforementioned radiosonde measurements were acquired from the vicinity of synoptic station. This station is located at 51^o, 21’E and 35^o, 41’N in the south of Tehran at an altitude of 1191 (m) from mean sea level. Also some data from other synoptic stations including Kermanshah, Tabriz, Shiraz, Ahwaz, Bandarabbass, Zahedan and Mashhad stations were used for model evaluation.
A practical method for extraction of temperature from MODIS measurements is to use the predefined statistical relationship between measured or modeled radiance flux densities and the corresponding temperature and moisture profile in the atmosphere. In statistical extraction method, the regression between flux density radiated from CO₂ and water vapor in corresponding absorbing bands are modeled. This method is usually used in producing the first guess profile to be used in physical models later on. A method named Localized Statistical Regression Profile Retrieval (LSRPR) for extraction of temperature profiles from MODIS images based on statistical regressions is introduced in this work. In this method an approach for improving the clear sky temperature profile calculation is presented, where using local atmospheric profiles collected by Radiosondes and corresponding MODIS images, some regression coefficient matrix is calculated locally. Then by applying this matrix to other MODIS images, one can calculate temperature profiles with a precision better than what is achieved by MODIS research team i.e. MOD07/MYD07.
Here radiosonde data along with concurrent MODIS images were used and the relevant regression coefficients were calculated. The average RMSE between temperature profile calculated from LSRPR and the one measured by radiosonde in the selected stations around the country was about 3.43K. This for MODIS products was 4.66K. The average RMSE between humidity profiles calculated from LSRPR and the one measured by Radiosonde in the selected stations was 1.27 g/kg and this value for MODIS products was 1.41 g/kg. As can be seen LSRPR model shows improvement in the isopleths of temperature compared to MODIS products. This improvement for humidity isopleths was not as good as for temperature. This could be due to the low amount in humidity in June. On the other hand considerable improvements in the precision of temperature isopleths extraction can be due to the use of local temperature profiles as the first guess in extraction algorithm.
Moreover the results show that in the stations other than Mehrabad airport, the RMSE between temperature and humidity profiles extracted from LSRPR algorithm and those calculated from radiosonde measurements is increased.
Based on the achievements in this research, it seems that LSRPR algorithm can enhance the precision compared to MODIS and the values calculated by this method are well comparable with the radiosonde collected data. So this algorithm can be used as an efficient method in regions such as Iran with relatively low number of radiosonde stations and irregular radiosonde measurements for producing temperature and humidity isopleths maps at different pressure levels in the atmosphere. It is hoped that by using these maps the accuracy of weather and climate prediction in the regional scale can be increased.