Satellite Mapping of the Demolition of the Rocky Flats Nuclear Weapons Plant Marco Chini Department of Physics Alma Mater Studiorum University Bologna, Italy chini@ingv.it William J. Emery CCAR University of Colorado at Boulder Boulder, Colorado, USA emery@colorado.edu Fabio Pacifici Earth Observation Laboratory Tor Vergata University Rome, Italy f.pacifici@disp.uniroma2.it Abstract— We present two different change detection techniques to monitor surface changes that occurred at the Rocky Flats nuclear weapons facility located immediately to the North West of the city of Denver, Colorado, USA. The site started being cleaned up and dismantled in 1998 and was completed in 2005. The first Change Detection method is based on a Maximum Likelihood classifier, while the other is an approach based on a Neural Network architecture called NAHIRI (Neural Architecture for HIgh-Resolution Imagery) to produce change detection maps from very high-resolution satellite imagery. NAHIRI simultaneously exploits spectral and temporal information by adding a filter, directly stemming from the multi- temporal information, to the classification changes derived from the multi-spectral data. In fact, the distinctive feature of this method is that the NNs exploit both the multi-spectral and the multi-temporal information in parallel that are associated with the changed values of the pixel spectral reflectances. The quantitative results are analyzed in order to single out advantages and shortcomings of the two different approaches. Keywords- Change detection, very high spatial resolution optical imagery, neural networks, maximum likelihood I. INTRODUCTION In the past decametric spatial resolution satellite imagery were able to detect urban texture, building densities, roof colour associated to building age and main road directions [1]. This information was useful within demographic and statistical analysis, local transport management, urban growth monitoring [2][3] and damage detection in urban areas caused by destructive events [4]. The advent of very high spatial resolution optical satellite imagery has greatly increased our ability to monitor land cover changes from space. Satellite observations, which are carried out regularly and continuously, provide a great deal of insight into the temporal changes of land cover use. This enhanced spatial resolution provides more precise information on land use and land cover changes and is being increasingly used to carry out detailed characterization of the trends in urban areas. In fact, the detection of fine-scale physical changes in individual objects, such as single buildings, houses or roads, is greatly enabled by these systems. Although land cover changes in the urban environment usually apply to surface transitions from bare soil or vegetated/pastoral surfaces to man-made structures (buildings, roads, etc.), in this paper we detect and monitor land surface changes in the opposite direction (from man-made structures to bare soil or naturally vegetated surfaces). The test area considered for this study is Rocky Flats, a site for the production of nuclear weapons located immediately to the North West of the city of Denver, Colorado, USA. From 1952 to 1989, the primary mission of Rocky Flats was to build plutonium triggers for nuclear bombs. In 1993, the U.S. Secretary of Energy announced that the site’s nuclear weapons production was officially over and the site started being cleaned up and dismantled in 1998. About 800 buildings, some of them very large, have been taken down to bare soil. This demolition was completed in mid-2005. In this paper we carry out a comparative study between two change detection techniques to monitor land cover changes in the opposite direction as described above. The first one is a Change Detection (CD) method based on a Maximum Likelihood (ML) classifier, where temporal changes of reflectance and different response of multi-spectral data for distinct objects are simultaneously taken into account in the same classifier to identify the different classes of change in the scene. The other is an approach based on a Neural Network (NN) architecture called NAHIRI to produce change detection maps from very high-resolution satellite imagery [5]. NAHIRI simultaneously exploits spectral and temporal information by adding a filtering effect, directly stemming from the multi- temporal information, to the classification changes yielded by the multi-spectral data. In fact, the distinctive feature of this method is that the NNs exploit both the multi-spectral and the multi-temporal information in parallel associated with the changed values of the pixel spectral reflectances. Temporal changes over the Rocky Flats site have been mapped using both ML CD and NAHIRI applied to two multi- spectral QuickBird (QB) images. The two methods have been able to automatically identify the land cover changes that occurred between 2003 and 2005 as the demolition of the buildings, the conversion of asphalt parking lots to soil, the creation of drainage channels and the relocation of water bodies. The quantitative results are critically analyzed and discussed to single out advantages and disadvantages of the two different approaches in view of an operational tool for the analysis of high resolution optical images for urban change detection.