Dataset Information

Cloud Masking for Landsat 8 and MODIS Terra Over Snow-Covered Terrain: Error Analysis and Spectral Similarity Between Snow and Cloud.

ABSTRACT: Automated, reliable cloud masks over snow-covered terrain would improve the retrieval of snow properties from multispectral satellite sensors. The U.S. Geological Survey and NASA chose the currently operational cloud masks based on global performance across diverse land cover types. This study assesses errors in these cloud masks over snow-covered, midlatitude mountains. We use 26 Landsat 8 scenes with manually delineated cloud, snow, and land cover to assess the performance of two cloud masks: CFMask for the Landsat 8 OLI and the cloud mask that ships with Moderate-Resolution Imaging Spectroradiometer (MODIS) surface reflectance products MOD09GA and MYD09GA. The overall precision and recall of CFMask over snow-covered terrain are 0.70 and 0.86; the MOD09GA cloud mask precision and recall are 0.17 and 0.72. A plausible reason for poorer performance of cloud masks over snow lies in the potential similarity between multispectral signatures of snow and cloud pixels in three situations: (1) Snow at high elevation is bright enough in the "cirrus" bands (Landsat band 9 or MODIS band 26) to be classified as cirrus. (2) Reflectances of "dark" clouds in shortwave infrared (SWIR) bands are bracketed by snow spectra in those wavelengths. (3) Snow as part of a fractional mixture in a pixel with soils sometimes produces "bright SWIR" pixels that look like clouds. Improvement in snow-cloud discrimination in these cases will require more information than just the spectrum of the sensor's bands or will require deployment of a spaceborne imaging spectrometer, which could discriminate between snow and cloud under conditions where a multispectral sensor might not.

SUBMITTER: Stillinger T

PROVIDER: S-EPMC6988483 | biostudies-literature | 2019 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Cloud Masking for Landsat 8 and MODIS Terra Over Snow-Covered Terrain: Error Analysis and Spectral Similarity Between Snow and Cloud.

Stillinger Timbo T Roberts Dar A DA Collar Natalie M NM Dozier Jeff J

Water resources research 20190729 7

Automated, reliable cloud masks over snow-covered terrain would improve the retrieval of snow properties from multispectral satellite sensors. The U.S. Geological Survey and NASA chose the currently operational cloud masks based on global performance across diverse land cover types. This study assesses errors in these cloud masks over snow-covered, midlatitude mountains. We use 26 Landsat 8 scenes with manually delineated cloud, snow, and land cover to assess the performance of two cloud masks: ...[more]

PMID: 32025064

Similar Datasets

Project description:Understanding and managing the health effects of ambient temperature (Ta) in a warming, urbanizing world requires spatially- and temporally-resolved Ta at high resolutions. This is challenging in a large area like France which includes highly variable topography, rural areas with few weather stations, and heterogeneous urban areas where Ta can vary at fine spatial scales. We have modeled daily Ta from 2000 to 2016 at a base resolution of 1 km2 across continental France and at a 200 × 200 m2 resolution over large urban areas. For each day we predict three Ta measures: minimum (Tmin), mean (Tmean), and maximum (Tmax). We start by using linear mixed models to calibrate daily Ta observations from weather stations with remotely sensed MODIS land surface temperature (LST) and other spatial predictors (e.g. NDVI, elevation) on a 1 km2 grid. We fill gaps where LST is missing (e.g. due to cloud cover) with additional mixed models that capture the relationship between predicted Ta at each location and observed Ta at nearby weather stations. The resulting 1 km Ta models perform very well, with ten-fold cross-validated R2 of 0.92, 0.97, and 0.95, mean absolute error (MAE) of 1.4 °C, 0.9 °C, and 1.4 °C, and root mean square error (RMSE) of 1.9 °C, 1.3 °C, and 1.8 °C (Tmin, Tmean, and Tmax, respectively) for the initial calibration stage. To increase the spatial resolution over large urban areas, we train random forest and extreme gradient boosting models to predict the residuals (R) of the 1 km Ta predictions on a 200 × 200 m2 grid. In this stage we replace MODIS LST and NDVI with composited top-of-atmosphere brightness temperature and NDVI from the Landsat 5, 7, and 8 satellites. We use a generalized additive model to ensemble the random forest and extreme gradient boosting predictions with weights that vary spatially and by the magnitude of the predicted residual. The 200 m models also perform well, with ten-fold cross-validated R2 of 0.79, 0.79, and 0.85, MAE of 0.4, 0.3, and 0.3, and RMSE of 0.6, 0.4, and 0.5 (Rmin, Rmean, and Rmax, respectively). Our model will reduce bias in epidemiological studies in France by improving Ta exposure assessment in both urban and rural areas, and our methodology demonstrates that MODIS and Landsat thermal data can be used to generate gap-free timeseries of daily minimum, maximum, and mean Ta at a 200 × 200 m2 spatial resolution.

Dataset Information

Cloud Masking for Landsat 8 and MODIS Terra Over Snow-Covered Terrain: Error Analysis and Spectral Similarity Between Snow and Cloud.

Publications

Cloud Masking for Landsat 8 and MODIS Terra Over Snow-Covered Terrain: Error Analysis and Spectral Similarity Between Snow and Cloud.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets