Skip all navigation and jump to content Jump to site navigation
About MODIS News Data Tools /images2 Science Team Science Team Science Team

   + Home
MODIS Publications Link
MODIS Presentations Link
MODIS Biographies Link
MODIS Science Team Meetings Link



Wu, W, Hall, C, Zhang, LJ (2006). Predicting the temporal and spatial probability of orographic cloud cover in the Luquillo Experimental Forest in Puerto Rico using generalized linear (mixed) models. ECOLOGICAL MODELLING, 192(4-Mar), 473-498.

We predicted the spatial pattern of hourly probability of cloud cover in the Luquillo Experimental Forest (LEF) in North-Eastern Puerto Rico using four different models. The probability of cloud cover (defined as the percentage of the area covered by clouds in each pixel on the map in this paper) at any hour and any place is a function of three topographic variables: aspect, slope and the difference between elevation and lifting condensation level. We chose the best models based on multiple statistics including the Akaike Information Criterion (AIC), scaled deviance and extra-dispersion scale. As a result, the generalized linear model (GLM) and one generalized linear mixed model (GLMM) with exponential spatial structure were the best candidate models. The probabilities of cloud cover in both our simulations and the observations increased with elevation, and were higher at night. They decreased in the morning after the sun rose until early afternoon, and then increased again for the rest of the day until night, apparently in response to the movement of the lifting condensation level. Two types of satellite images were available to calibrate our models: the higher spatial resolution, but expensive and infrequent Landsat-7 Enhanced Thermal Mapper plus (ETM+) images and the frequent, free, but low spatial resolution Moderate Resolution Imaging Spectroradiometer (MODIS) images. The derived probabilities of cloud cover when calibrated to the two types of remote sensing images were very similar, which justifies our using the free MODIS images instead of the Landsat images to calibrate the models. We applied the model to all months and the results indicated in agreement with the data that the probability of cloud cover is less during the dry season, higher in wet season and moderate for the rest of the year. We found our models could usually predict the probability of cloud cover for each 100 m in elevation level at a certain time with an index of agreement (IoA) of 0.560-0.919 and at a certain location over a day with an IoA of 0.940-0.994, indicating a medium to good model simulation at that particular time or location. (c) 2005 Elsevier B.V. All rights reserved.



NASA Home Page Goddard Space Flight Center Home Page