Cloud base height estimates from sky imagery and a network of pyranometers

Abstract

Cloud base height (CBH) is an important parameter for physics-based high resolution solar radiation modeling. In sky imager-based forecasts, a ceilometer or stereographic setup is needed to derive the CBH; otherwise erroneous CBHs lead to incorrect physical cloud velocity and incorrect projection of cloud shadows, causing solar power forecast errors due to incorrect shadow positions and timing of shadowing events. In this paper, two methods to estimate cloud base height from a single sky imager and distributed ground solar irradiance measurements are proposed. The first method (Time Series Correlation, denoted as “TSC”) is based upon the correlation between ground-observed global horizontal irradiance (GHI) time series and a modeled GHI time series generated from a sequence of sky images geo-rectified to a candidate set of CBH. The estimated CBH is taken as the candidate that produces the highest correlation coefficient. The second method (Geometric Cloud Shadow Edge, denoted as “GCSE”) integrates a numerical ramp detection method for ground-observed GHI time series with solar and cloud geometry applied to cloud edges in a sky image. CBH are benchmarked against a collocated ceilometer and stereographically estimated CBH from two sky imagers for 15 min median-filtered CBHs. Over 30 days covering all seasons, the TSC method performs similarly to the GCSE method with nRMSD of 18.9% versus 20.8%. A key limitation of both proposed methods is the requirement of sufficient variation in GHI to enable reliable correlation and ramp detection. The advantage of the two proposed methods is that they can be applied when measurements from only a single sky imager and pyranometers are available.