In order to mitigate with the natural hazards (surface water, floods, etc.) of precipitation origin it is necessary to measure, model and predict the precipitation for water resources and basin management. Due to high spatial and temporal variability of precipitation measurement networks, there are difficulties in determining small scale intensive rainfall events. Currently available weather prediction models yield reliable results, but they cannot catch the spatio-temporal patterns of convective rainfall events. For this reason, the satellite based meteorological information is used as input in weather prediction models. One of the most significant subjects in meteorology domain is the determination of precipitation pattern types and quantities through remote sensing. The instruments used for this purpose are radars and satellites. Satellite systems have the most refined resolution among all these instruments, which provide them significant superiority. For convective structures the cloud top temperature is a key parameter that can be related to precipitation. The basic idea herein is that the cold cloud top temperatures generate more precipitation than hot or warm cloud top temperatures. Keeping this point in mind, first currently available techniques in the literature are explained in detail including Arkin, GOES precipitation index, adjusted GOES precipitation index, Griffith-Woodley, Negri-Adler-Wooley, convective-stratiform, enhanced convectivestratiform, automatic estimator, and advanced automatic estimator techniques. They include empirical equations in their structure and the parameters of these equations must be determined for our country. On the other hand, these methods include linear relationships. In this paper, one convective precipitation event in 2000 is examined each within time period as 5-7 September. These are explained by considering Meteosat 7 infrared channel data for 6-hour total rainfall amounts in such a manner that the data at the upper troposphere vertical levels are transferred to grid points from NCEP/NCAR. The application of the methodology is presented for 26 Marmara region raingauge stations. On the other hand, as effective methodology Artificial Neural Network (ANN) is used and, it has the analogy such that the inputs are taken from the environment through the neurons and transmitted to the brain. ANNs can be thought as a black box model, which processes inputs and produces convenient outputs for inputs. As a first stage this black box is trained and after training system reaches level of decision for inputs. ANNs have ability of learning, due to their training stages. This is feature that there is not available in any classical method. The ANN model can learn by training similar to a human and it has a non-linear structure. Such a non-linearity provides a distinctive possibility in the domain of artificial intelligence. For the model ANN and its analysis, various statistical criteria are used as detection probability (POD), bias, wrong alarm ratio (FAR), critical success index (CSI), target ratio (HR), and transferring of grid points to station points by using a method called inverse distance square. These statistical criteria are explained in this paper. By following the methodology explained above, as a result of ANN properties, convenient architecture for total precipitation amount prediction is proposed that constitutes single input, hidden and output layers with 37, 19 and 1 neurons, respectively. Furthermore, this model is used for precipitation prediction by considering each within time period as 5- 7 September. The correlation value of ANN model is 0.89 in testing. Moreover, statistical values are calculated as POD = 0.61, FAR = 0.30, CSI = 0.49, BIAS = 0.86 and HR = 0.77. For this period, rainy events are predicted with success. As it is seen from the descriptions and application results, the cloud top temperature is more related to precipitation by considering low level data of troposphere. Under the light of the proposed prediction model this is the result of key parameter, which reflects the cloud top temperature significant relationship to precipitation. Although other parameters have also significance to lesser degrees the final prediction model take into consideration few of the input variables. In the mean time different combinations of the prediction are investigated throughout the study.