We use data from the Multispectral Thermal Imager (MTI) to evaluate the effects of spatial resolution on the accuracy of aerosol optical depth (AOD) retrieval. Our results show that increasing the pixel size by itself has little effect on AOD retrieval accuracy at our chosen study site. However, increased pixel size does increase the error in AOD retrieval as a result of clouds. High-resolution sensors like MTI are able to avoid most clouds, but as the pixel size increases, subpixel clouds avoid the cloud mask, creep into selected pixels, and add a positive bias to the retrieved value of AOD. In accompanying work, we show that increasing pixel size has a small but noticeable impact on the normalized difference vegetation index (NDVI) and the 2.2-mu m reflectance, both used in the retrieval algorithm. We also examine the uniformity of the aerosol layer and show that the AOD varies by less than 0.02 in optical depth units over a 2.3 x 3.8 km(2) area. An analysis of the temporal variability of Aerosol Robotic Network-retrieved AOD shows a standard deviation of 0.02 on partly cloudy days and 0.004 on clear days.
