Determining sedimentation trend on flood spreading system using satellite image change detection technique

Background and objectives: Flood spreading system (FSS) is one of a variety of methods for artificial aquifer recharge. The system's main problem is sedimentation on the FSS area, which reduces its effectiveness for aquifer recharge. In this regard, satellite image change detection technique is a proper method for accurate sediment distribution pattern diagnosis with the FSS. A relevant study has been conducted by Sarreshtehdari (2005), where classified images of 1987 and 2001 were compared and revealed a sediment distribution pattern and variations. In the present study, which was conducted on Jarmeh FSS in the north of Khuzestan Province, sedimentation area and distribution pattern in several periods was determined by satellite image change detection technique.
Materials and methods: The changes were detected for three periods: 1999 (FSS construction year) to 2000, 2000 to 2006 and 2006 to 2013. Change detection follows three main steps: 1- Data pre-processing including geometry and radiometry corrections 2- Selection of change detection method 3- Results accuracy assessment. In the data pre-processing step, image to image method was used for geometry correction and Dark Object Subtraction method was used for radiometry correction. Post-Classification comparison algorithm, change detection statistics method was selected for the present study. The method is capable of preparing a complete change matrix with details. Some indices including overall accuracy, Omission, Commission and kappa coefficient extracted from error matrix, were used to assess the accuracy of results. Maximum likelihood algorithm was selected and used for image classification. Also, Optimum Index Factor (OIF) was used to improve image color composite selection, for the preparation of training samples.
Results: Based on the FSS area properties, three regions including sediments, disturbed and undisturbed terrains were separated on the FSS and the images were classified accordingly. Classification overall accuracy for 2000, 2006 and 2013 images was 78, 85 and 83%, respectively. The results of post-classification comparison algorithm by change detection statistics showed that, in the 2000 to 2006 period, sediment area increased from 8.5 to 27.5 ha. In 2006 to 2013, due to drought and the spread of flood on the FSS, the sediment area increased a little and reached 29.5 ha.
Conclusion: The results showed that, the sediment area was well extended in the first period. Proper amounts of precipitation and flooding on the one hand and no sediment initial surfaces on the other hand resulted in a good change detection. However, in the second period, the sediment area slightly decreased. Such a result is logically unexpected but drought in recent years and no flooding especially on the middle area of FSS, result in old sediment surfaces with different reflection properties. In this case, the classification algorithm was unable to enhance the entire sedimentation area. Hence, not only is there no reduction in sedimentation area, but the area slightly increased in the upper FSS for the second period.