How does data accuracy influence the reliability of digital viewshed models? A case study with wind turbines

Abstract Viewshed analysis is a GIS tool commonly used in a number of research and practical spatial analyses. Input data and their spatial uncertainty are important aspects affecting analysis reliability. Given that inappropriately selected input geodata can produce imprecise visibility models and as a result cause incorrect spatial decisions, quantifying the effect of this uncertainty on resulting visibility models is important for the models' subsequent use. The objective of our study was to evaluate the suitability of digital surface models with varying levels of detail (a LiDAR-based model and models based upon vector data at differing scales) for simple (binary) viewshed analysis of wind turbines (three wind parks each containing 3–6 turbines). Visibility models based upon this input data were compared with actual visibility from 150 control points at random locations. The study results confirmed the prediction that the viewshed model based on more precise input data corresponded more closely to reality. Moreover, our study is the first to demonstrate that only the number of false positives (where the model predicts that an object is visible while in reality it is not) depended on input data precision, while input data did not affect the false negatives. In addition, all vector-based models had far more false positives than false negatives, while the opposite was true for the LiDAR-based model. When we considered the same number of modeled and actually visible wind turbines as a model's matching of reality, there were matches at 83.6–93.7% of control points (95% confidence interval) for the LiDAR-based model. For models based upon vector maps of various scales, the intervals were 68.4–82.2% (1:10,000), 59.1–74.2% (1:25,000), and 48.1–63.9% (1:500,000). We recorded false positives in 6 cases with the LiDAR-based model and 26, 39, and 59 cases, respectively, for vector-based models.