Identification Seagrass Condition from ALOS AVNIR-2 using Artificial Neural Network at Pari Island

Abstract

Seagrass beds have important roles in marine life, but the unavailable of information about the condition of seagrass causes difficulties in managing coastal areas properly. Regularly updated and accurate information on the percentage cover of seagrass is an essential component of the knowledge required to monitor, understand and manage this resource. Artificial Neural Network (ANN) was applied to ALOS AVNIR-2 to identify seagrass condition. Twenty two classification scenarios were done to compare the result of accuracy. There are three class of seagrass condition. Seagrass cover more than 60% indicates the condition of good seagrass. Seagrass cover from 30% to 59,9% indicates the condition of medium seagrass. Seagrass cover below 29,9% indicates the condition of poor seagrass. Best accuracy was obtained by scenario H by entering combination of blue (0.42 to 0.50 μm) and NIR (0.76 to 0.89 μm) wavelength plus water depth data as input parameters, with a value of 71.43% overall accuracy. However, looking at individual class, Scenario C, which is 58.33% of overall accuracy by using combination of blue (0.42 to 0.50 μm), green (0.52 to 0.60 μm), NIR (0.76 to 0.89 μm) wavelength plus water depth achieved higher producer and user accuracy. Overall, the result of identification seagrass condition from ALOS AVNIR-2 using artificial neural network at Pari Island in 2010 is dominated by poor seagrass, while good seagrass and medium seagrass were found in small area.

Padang lamun memiliki peran penting dalam kehidupan laut, namun tidak tersedianya informasi tentang kondisi padang lamun menyebabkan kesulitan dalam mengelola kawasan pesisir dengan benar. Memperperbaharui secara teratur dan akurat tentang informasi luas tutupan lamun merupakan hal yang penting dari pengetahuan yang dibutuhkan untuk memantau, memahami dan mengelola sumberdaya ini. Artificial Neural Network (ANN) telah diterapkan pada ALOSAVNIR 2 untuk mengetahui kondisi lamun. Dua puluh dua skenario klasifikasi dilakukan untuk membandingkan hasil akurasi. Ada tiga kelas kondisi lamun. Tutupan lamun lebih dari 60% menunjukkan kondisi padang lamun baik. Tutupan lamun dari 30% sampai 59,9% menunjukkan kondisi lamun sedang. Tutupan lamun dibawah 29,9% menunjukkan kondisi lamun jelek.