Depth-Integrated Estimation of Dissolved Oxygen in a Lake


Akkoyunlu A., Altun H., Cığızoğlu H. K.

JOURNAL OF ENVIRONMENTAL ENGINEERING-ASCE, cilt.137, sa.10, ss.961-967, 2011 (SCI-Expanded) identifier identifier

Özet

The majority of variable estimation studies in water resources investigate the temporal variation of the variable. In this study, we examined the depth-dependent estimation of a lake's dissolved oxygen (DO) using two artificial neural network (ANN) methods: (1) the radial basis functions (RBFs) and the feed forward back-propagation (FFBP), and (2) the multilinear regression (MLR). We tested two different input layer configurations. In the first case, we employed all other available lake parameters-total dissolved solids (TDS), pH, conductivity, lake depth, and lake temperature-to estimate DO. In the second case, we considered only depth and temperature to estimate DO. The performance evaluation criteria of these two cases were close. ANN estimation performances were noticeably superior to those of MLR, as reflected in the performance evaluation criteria and DO lake depth plots. We saw that the spatial variation of the lake's DO can be captured by ANNs satisfactorily, even if available measurements are quite limited. DOI: 10.10611(ASCE)EE.1943-7870.0000376. (C) 2011 American Society of Civil Engineers.