Optimal Development of Rain Gauge Network using Kriging and Entropy in Geographic Information System (GIS), (Case Study of Karkhe Basin)

In the water resources planning and management, rainfall is considered as the main representative factor of region's hydroclimate system. There is a high level of spatial and temporal variability in rainfall events. Rain gauges are used for precipitation measurement in different points of the region. Due to limitations of rain gauges development, it is impossible to measure the exact distribution of rainfall over an area, so it is needed to select the optimal number and locations of the rain gauges in a region. There are several factors that affect the optimal number and location of stations, thus, development of an optimization model for development of rain gauges network is necessary. In this study an optimization model is proposed for determining location of the rain gauges considering the transformation entropy and the regional rainfall estimation error. In this model, the points with the minimum transformation entropy and the maximum rainfall estimation error are considered as candidates for development of new stations. The results of two considered criteria are combined to determine the location of the new stations. Geogrpahic information system (GIS) is used to better represent the results of rainfall spatial analysis and explain results of the proposed optimization model. A sub basin of Karkhe basin in South West of Iran has been considered as the case study of this study. Results show that using new rain gauges which are determined to be 17 using the proposed approach could improve accuracy of spatial analysis of rainfall siginficiantly.
Keywords:
Variance estimation, entropy, rainguage network optimization, GIS, Karkheh Basin
Methodolgoy
This proposed methodology in this study involves the following steps: (1) Data collection and analysis (2) application of kriging to existing rainfall data to calculate the rainfall spatial analysis variance (3) calculating the transformation entropy in the basin surface (4) selection of candidates points for rain gauge development considering the minimum transformation entropy and the maximum rainfall estimation error (5) presentation of rain gauge network final map.
In this paper, the best combination of sampling stations in a monitoring network is selected using the entropy theory by considering the maximum uncertainty (minimum redundant information in the system) and the maximum rainfall estimation Kriging error. Hence, in this study, a new model composed of variance estimation and entropy is proposed to relocate the rainfall network and to obtain the optimal design with the minimum number of rain gauges.
Results and Discussion
The rainfall data of the 49 stations in the study region are for the period of October to April are utilized.The correlation coefficient higher than 0.6 in rainfall and height, Cokriging method was used to analyze the spatial rainfall. Kolmogorov Smirnov test (K-S) with a confidence level of 95% of normal monthly precipitation data is verified. In case of non-normal data conversion Cox - Box or log normal distribution, the data are close to normal distribution. The estimated variance is calculated for each month. After calculating variance estimates for each month, the layers can be weighted according to the average rainfall. The final layer of the overlapping layers are obtained and as a measure of the objective function be considered. The transformation entropy layer such as variance estimation layer obtained. A new model composed of variance estimation and entropy is proposed to relocate the rainfall network to obtain the optimal design with the minimum number of rain gauges. As a case study, the application of the proposed method to an existing rain network over the Karkhe catchment region under a minimum transformation entropy of 30% and maximum Kriging error of 60% resulted in 17 new rain stations to be added to the original network.
Cnclusions
In this study a methodology is proposed to suggest new locations for rain gauges development using kriging and entropy methods. On the basis of the rainfall data from the current rain gauge stations, the rainfall of the candidate rain gauge stations are generated by estimation Kriging error. The information entropy is based on the concept of probability to measure uncertainties. A network optimization model based on minimizing the estimated variance and by rain gauge data suggest that the implementation of this new model, 17 stations were added to the network location. Most of the stations in the eastern and north-eastern border of the basin, in the highlands and in places where the space station is too high, they were located. The results show that using the theory of Entropy with geostatistical methods, a higher accuracy in rainguage network development, can provid. By combining the two methods can be used to determine the best places established stations, so that the two factors cover each other. Spatial design using model proposed in this paper, the best combination for rainguage stations using the minimum transformation entropy and the maximum rainfall estimation Kriging error is selected.