Methodology for optimizing a photosynthetically active radiation monitoring network from satellite-derived estimations: A case study over mainland Spain

• Published in: Atmospheric research 2018-11, Vol.212, p.227-239
• Main Authors: Vindel, J.M., Valenzuela, R.X., Navarro, A.A., Zarzalejo, L.F.
• Format: Article
• Language: English
• Subjects: Clustering analysis; K-means algorithm; Photosynthetically active radiation
• ISSN: 0169-8095; 1873-2895
• DOI: 10.1016/j.atmosres.2018.05.010

A methodology is presented for determining optimal locations to install photosynthetically active radiation (PAR) measurement stations. Initially, a cluster analysis was performed from PAR satellite-derived estimations over mainland Spain. Once the optimal number of clusters was obtained, the total number of locations included in the monitoring network was distributed among different groups, according to the size and variability of each group. Finally, the specific locations for measurement stations placement was determined using an iterative technique: The largest region within each cluster was split into two new sub-regions, providing two new sites for substituting the initial location. Clustering analysis has previously been applied to determine locations to monitor solar radiation. However, this is the first implementation for PAR stations in mainland Spain. Another novelty developed in this work is the distribution employed for specific sites within each cluster. The outcome achieved using clustering analysis was compared to those obtained using three other methods: two methods without clustering analysis and the third where clustering is performed but not optimizing the number of clusters. In one technique without clusters, the largest region is split into two new sub-regions, similar to the clustering analysis with optimization. In the second without clustering analysis, since the data variability was not previously addressed, the region divided is those with the largest combined effect of variance and size. The results fully justify using a clustering process; however, clustering without optimization is the worst performing method. •A new methodology to determine optimal locations for PAR measurements is presented.•The method is based on a cluster analysis from PAR satellite-derived estimations.•The specific locations are determined by an iterative technique.