Estimation of cooling degree days (CDDs) from AVHRR data and an MLF neural network

Abstract

Cooling degree day (CDD) is the most common and simplest measure used in the ventilating and air-conditioning industry to estimate cooling energy requirements. In this study, multilayer feed-forward (MLF) neural networks with one hidden layer were used to estimate CDDs in Khuzestan Province (in southwest Iran) using Advanced Very High Resolution Radiometer (AVHRR) images. The input variables for the networks were the four bands of the AVHRR image (B1, B2, B4, and B5), altitude, and Julian day. The study examined various combinations of these parameters as inputs to the artificial neural network (ANN) models to evaluate the degree of effectiveness of each of these variables on CDD. Bands 4 and 5, which are used for retrieval of surface temperature, were the most critical components for the estimation. The contribution of Julian day to the precision of estimated CDD was much more superior to that of altitude. The network using four bands, Julian day, and altitude provided the best results, with 12 nodes in the hidden layer. The results of this study showed that daily CDD values can be estimated with acceptable levels of the statistical indicators from AVHRR data and from the two geographic parameters using the ANN model.