Landsat 8 OLI images have become important data sources;however

they are usually covered by clouds and cloud shadows

which reduce data availability.Therefore

a rapid method for detecting clouds and cloud shadows in a single image is necessary for followup data recovery applications.Threshold setting has been a commonly used method;however

it is difficult to use because the same threshold value usually indicates different objectives and varies across different data sources

as well as across different period images.Thus

the relative position relationship between a cloud shadow and a cloud is the key to detect a cloud shadow.First

a cloud index（CI） with a cirrus band and coastal/aerosol band was established to distinguish the thin and thick cloud pixels by setting thresholds.Second

a normalized dark pixel index（NDPI） with coastal/aerosol band and short-wave infrared 2 band was established.Furthermore

a ratio of shadow index（RSI） was established based on the NDPI and normalized differential vegetation index.RSI was used to distinguish the potential cloud shadow pixels in thin-cloud and non-cloud areas also by using thresholds.Third

an azimuth search method based on the solar azimuth angle and an appropriate searching distance were adopted to detect the cloud shadow from the potential cloud shadow pixels in the images.Two Landsat 8 OLI images were selected for the case study;the image captured on April 22

2014 was used for the test

whereas the image captured on July 12

2015 was used for validation.For each type（e.g.

thick cloud

water

cloud shadow

or other kinds of shadow）

200 random sampling points were used to assess the detection accuracy.Results showed that CI could quickly distinguish cloud from on-cloud pixels by statistics according to the cloud coverage ratio in the header files

with thresholds of CI ≥ 0.0011 and CI ≥ 0.0048 in the test and validation OLI images for the thick cloud type.Both user accuracy rates of detection for the thick cloud samples were over 99%.RSI could enhance the difference among water

cloud shadow

and other kinds of shadow

thereby facilitating the differentiation among the different types.The pixels with 0.45 ≤ RSI < 0.76 in the test image and 0.36 ≤ RSI < 0.76 in the validation image belong to potential cloud shadows.Using the solar azimuth angle at the time of imaging as the searching azimuth angle and a reasonable searching distance（500-2200 m in the test image and270-800 m in the validation image）

the azimuth searching method simplified the models of the relative position relationship of a cloud shadow to a cloud and accurately distinguished cloud shadow from water and other shadows.The detection precision levels for both test and validation reached more than 93%

which compensated for the limitation of the threshold method.The proposed shadow detection method combines threshold with simplified relative relationship model and leverages band difference.The method is feasible and rapid when applied to a single image

further improving the utilization accuracy of OLI images.