Examining the Potential of a Random Forest Derived Cloud Mask from GOES-R Satellites to Improve Solar Irradiance Forecasting

Abstract

In order for numerical weather prediction (NWP) models to correctly predict solar irradiance reaching the earth’s surface for more accurate solar power forecasting, it is important to initialize the NWP model with accurate cloud information. Knowing where the clouds are located is the first step. Using data from geostationary satellites is an attractive possibility given the low latencies and high spatio-temporal resolution provided nowadays. Here, we explore the potential of utilizing the random forest machine learning method to generate the cloud mask from GOES-16 radiances. We first perform a predictor selection process to determine the optimal predictor set for the random forest predictions of the horizontal cloud fraction and then determine the appropriate threshold to generate the cloud mask prediction. The results show that the random forest method performs as well as the GOES-16 level 2 clear sky mask product with the ability to customize the threshold for under or over predicting cloud cover. Further developments to enhance the cloud mask estimations for improved short-term solar irradiance and power forecasting with the MAD-WRF NWP model are discussed.

Exploring the Potential of Statistical Modeling to Retrieve the Cloud Base Height from Geostationary Satellites: Applications to the ABI Sensor on Board of the GOES-R Satellite Series

“…Each tree in the forest is a set of rules, or decisions, that is used in order to minimize the variance or impurity of the response variable, which in this case was the CBH [ 25 ]. More details of the random forest machine learning algorithm can be found in [ 26 , 27 ].…”

Assessment of the GOES-16 Clear Sky Mask Product over the Contiguous USA Using CALIPSO Retrievals

“…The cloud retrieval process starts with the determination of the cloud mask. Approaches based on statistical models and threshold based models have been proposed e.g., [ 4 , 5 ]. The operational cloud mask algorithms use threshold-based algorithms.…”

“…The multi-year assessment presented herein provides the most comprehensive evaluation of the ACM product over CONUS to date. The previous studies assessing the ACM performance that we are aware of have used also CALIPSO retrievals, but the studies have focused on a different region [ 14 ], or provided a basic evaluation of the product performance [ 5 ]. The algorithm theoretical basis [ 14 ] quantifies the performance of the ACM algorithm using retrievals from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on board of the Meteosat.…”

“…The assessment used eight weeks of data distributed over the four seasons and found a probability of detection, or accuracy (numer of correct retrievals divided by the number of retrievals), for all sky conditions (clear or cloudy) of 91.4%. The other study [ 5 ], focused on the development of an alternative retrieval algorithm based on statistical modeling, and only provides a basic evaluation of the ACM product as a baseline for comparison. Using about two months of data in each season, the study found an accuracy under all sky conditions of 85.7% with higher accuracy for cloud detection, 93.0%, than for clear sky detection, 67.7%.…”

“…Using about two months of data in each season, the study found an accuracy under all sky conditions of 85.7% with higher accuracy for cloud detection, 93.0%, than for clear sky detection, 67.7%. The study used the same grid under consideration herein, and thus the present work can be considered to a certain extent as an extension of the evaluation presented in [ 5 ]. The novel aspects of this study include a better pairing of GOES-16 and CALIPSO retrievals, and the use of a multi-year evaluation period.…”