Bushra Zaman^{1, 2,}, Mac McKee^{1, 2}

Unmanned Aerial Vehicle (UAV) data and artificial intelligence (AI) techniques have often been separately used for wetland management applications. Existence of native or invasive weed species which are a threat to the biodiversity of the wetland ecosystem is a common problem. High resolution multispectral image data from UAV and AI analysis together may prove to be a very robust combination for solving some of the wetland management problems. AggieAir, a UAV platform developed at Utah State University (USU), was used to acquire very high-resolution, multispectral aerial images of the study area in the summers of 2010 and 2011. The AggieAir data and an AI model have been used to classify and detect the weed P. australis in the Bear River Migratory Bird Refuge (BRMBR) wetland ecology. The images were classified based on the reflectance values in red, green and NIR bands using the multiclass relevance vector machine (MCRVM). The total P. australis cover was calculated and results indicated a decrease of 6.6% in total P. australis cover between June and September 2010 but an increase of 43.85% between June 2010 to July 2011. This study provided useful information about the extent and spread of P. australis P. australis which is a priority as regards to invasive plant treatment and control in the BRMBR. The results are easy to interpret and can contribute to management advising. Given the high spatial and temporal resolution of the UAV and excellent performance of the MCRVM model, we propose their further use for wetland management applications.

wetland management, water management, relevance vector machine, unmanned aerial vehicles, classification, P. australis P. australis, modeling, artificial intelligence