Ali Chavoshian; P.S. Katiraie-Boroujerdy
Abstract
Introduction: Precipitation has an important role not only in the variety of scientific applications including climate change, climate simulations, weather modeling, and forecasting but also in decision making such as water management, hydrology, agriculture, drought, and crisis management. Different ...
Read More
Introduction: Precipitation has an important role not only in the variety of scientific applications including climate change, climate simulations, weather modeling, and forecasting but also in decision making such as water management, hydrology, agriculture, drought, and crisis management. Different temporal resolutions and coverages of data are required for this and other applications. For example, long term meteorological data are needed for monitoring the climate variability and trends and for climate simulation assessments in local and global scales. Also, present data are used to assimilate into forecast models to improve the predictions. Historical and present precipitation data are the main requirements to monitor and predict droughts which help to early warning system and water management decisions in a country. The recent rainfall data are also the primary input of hydrological models to flood forecast in a basin. The accurate estimation of precipitation amount is vital for these applications.
Materials and Methods: However, rainfall is discontinuous and varies greatly both in time and space which makes it parallel with difficulties in the actual measurements. The two main sources of observational precipitation datasets are ground-based rain gauge measurements and space-based remote sensing satellite estimations each one with its own limitations and strengths. Historically, rain-gauge measurements have been considered as the “ground truth”, but they have mostly limited to land surface, the measurements are sparse or nonexistent in some regions like deserts or high topographic areas. Although rain gauges measure rainfall directly, their data are only representative for a limited spatial extent and may be subjected to some errors caused by local effects such as topography or wind-induced undercatch. An alternative approach which can provide relatively homogenous estimates with complete coverage over most of the globe is based on using satellite observations. Therefore, satellite data are capable to estimate precipitation over the oceans and over remote areas where few or no ground measurements are available. The satellite-based precipitation estimates are derived mainly from visible, infrared (IR) and passive microwave (PMW) radiances which are measured by satellites. Although the visible channels cannot be used at night, the IR data are available in fine spatial resolution (about 3-4 km) with high temporal sampling (15 min) which are provided by geosynchronous satellites. Another source of data is PMW that can be used to estimate rainfall more directly. Low-altitude polar-orbiting satellites serve to measure the PMW data. Although, the microwave sensors can penetrate into the clouds and provide more information about the cloud characteristics such as water vapor, cloud particles, and structure of hydrometeors, but at the expense of temporal sampling. In recent years, different algorithms have been developed using the combination of the IR, Visible (VIS) and PWM observations to provide more accurate rainfall estimations in high spatial and temporal resolutions. To demonstrate the similarities and differences between the spatial distribution of different satellite-based and gauge-based precipitation datasets over Iran we compared seven different datasets. For comparisons all datasets are regridded to 0.25-degree latitude longitude spatial resolutions. Then the spatial distribution of the mean and relative standard deviations of annual precipitation of these datasets have been calculated. We also used more than 2000 rain gauges to evaluate the selected datasets. To reduce error only 228 pixels, include at least 3 rain gauges are used for comparisons of spatial average of monthly, seasonal and annual precipitation of gauge and seven datasets.
Results and Discussion: The results showed a large amount of differences in annual precipitation between seven selected datasets. The most differences pronounce in wet areas in the north of Alborz Mountain, in the semi-arid and arid regions of the central desert and in the high mountainous areas of the southern Zagros. The reason for these differences is that not only satellite-based but gauge-based datasets have large uncertainties estimating areal precipitation in such high topographic areas. The satellite products are prone to some errors arising from not fully understood physical process, sampling error and parameter estimation. Therefore, verification of precipitation datasets is one of the most important parts of the data development and refinements. In this paper, the spatial distribution of seven different global-observational precipitation datasets over Iran are compared for the period 2003-2007. At first all datasets were regridded to 0.25° spatial resolutions using linear interpolation method. Then, the mean and relative standard deviation of annual precipitation of the datasets were calculated to analyze the spatial discrepancies between datasets. The areal average of annual precipitation and the contribution of seasonal precipitation were calculated for comparison purposes. The results showed that areal average of annual and seasonal precipitation for 228 selected pixels for PERSIANN-CDR, TRMM, and GPCP which are satellite-based and gauge adjusted datasets are more similar to the rain gauge data than other datasets. The results for the above datasets are even better than CRU and APHRODITE which are gauge-based datasets.
Conclusion: The results showed that the satellite estimates are not capable to show the precipitation (detection and amount) over the coast of Caspian Sea and the high areas of the Zagros Mountain as well as other parts of the country. There are some useful recommendations for data users at the end of this paper. In fact, in this paper our spatial focus is on Iran and we introduced a web address which data users can access freely from one of the most popular and widely used satellite-based products in easy-to-use format only for Iran. The results show considerable differences between the datasets. The difference is about 0.8 times of mean annual precipitation (about 300 mm in a year) for the coast of Caspian Sea. The satellite-based estimations were less accurate over the coast of Caspian Sea and high mountainous area of the southwest of Zagros comparing to other parts of the country. While spring precipitation shows maximum contributions in annul precipitation for in-situ datasets, winter precipitation shows maximum contribution in annual precipitation for other datasets. The results showed that areal average of monthly, seasonal and annual precipitation over 228 selected pixels for PERIANN-CDR, TRMM and GPCP were consistent with rain gauge data. CMORPH and PERSIANN underestimate areal average of monthly and seasonal precipitation over the pixels.
