USING OF INTELLIGENT ARTIFICIAL NEURAL NETWORK PREDICTIVE MODEL FOR IRAQI MARSHES RESTORATION

Abstract

This paper focuses one of the problems that challenges the continuation of inhabitant and aquatic organisms' life in the Mesopotamians wetlands, it is the dryness of marshes. Artificial Neural Networks (ANN) approach is applied to forecast and suggest a future release policy to restore Al-Huweizah Marsh. The suggested ANN model used to predict three different long-term (5, 10 and 15-years) policies to increase marsh water level from 2 to 7 meter above sea level. The results showed that the application of ANN for Al-Huweizah Marsh restoration using a network structure of 9:9:1 (input: hidden: output) has the ability to simulate marsh restoration process successfully with a regression coefficient of 99.8% and root mean square error of 0.88. Linear increment, high inflows months and month inflows weight restoration policies are applied to restore the marsh using each of optimistic, pessimistic and abstemious expected inflows. The ANN model and suggested policies are built in such a way to produce the same target head at the end of any year, this strategy will not cause a large variation between the resulted outflows for the suggested policies; at same time, the strategy enables the decision maker to move between policies according to the available inflows without changing the target level needed to be reached at the end of operating policy. The shortages in expected inflows to complete the suggested policies are little and may be overcame.