Impact of BRDF Spatiotemporal Smoothing on Land Surface Albedo Estimation

Round 1

Reviewer 1 Report

The authors address the use of a priori BRDF knowledge of land cover to produce albedo with high resolution, specifically, the notion that BRDF smoothed both temporally and spatially could produce changes in both magnitude and shape, and therefore impact to albedo retrieval.  They investigate albedo retrieval variation caused by BRDF smoothing both temporally and spatially over various land cover.  Their results show that as temporal smoothing increases from daily to monthly, and spatial smoothing increases from 500m to 5600m, the magnitude and shape of the smoothed BRDF do change, and demonstrate detectable absolute differences of 0.012-0.037 in subsequent albedo retrieval, depending on spectral band, landscape and vegetation.

This paper appears well conceived and executed.  However, poor English throughout makes it difficult to read at times.  Recommend editing for English usage.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This study investigates the impacts of spatial and temporal scaling of BRDF in surface albedo estimation with MODIS (and Landsat 8/OLI) data over a few example sites in Southwestern US. The methodology is straightforward and the results are reasonable. The presentation of paper is clear, though it can be further improved (e.g., with professional English editing services).

My specific comments are as follows:

1 (somewhat major).  The analysis uses MODIS BRDF at the higher temporal/spatial resolutions as the reference in the comparisons, which is fine. However, it must be borne in mind that the higher resolution BRDF does not necessarily have the better quality. Various errors could be introduced in the estimates of BRDF at daily steps, for instance. Temporal averaging, therefore, may be a reasonable way to suppress the impacts of such transient errors. Therefore, I would like the authors explicitly emphasize this point (i.e., high resolution satellite products do not necessarily represent surface truth) in the revised paper.

2. L214, L243-246: MODIS sinusoidal projection is a pretty bad choice for Western US. Why don't you reproject all the data layers to a common coordinates (e.g., lat/lon or UTM) instead? This is only a suggestion for the future, I guess, as the analyses are already done.

3. L256-257: What do you mean by "pushed forward and back"? It is confusing. I would suggest to totally remove the which-clause.

4. Fig. 3: as far as I understand, MODIS BRDF is estimated based on multi-day composite reflectance. Therefore the "daily" BRDF may not be truly "daily". You may want to clarify this here.

5. L266: Please provide a reference to the AMBLARS algorithm here. 

6. Eq. 6:  Note that the sampling scheme here by the zenith angles (-70deg to 70deg) put more weights on larger zenith angles. An alternative scheme is to sample the space by cos(SZA).

7. Fig. 13: What are causes of the spikes of the differences shown in the figures (e.g., panel a for EBF and f for ENF)?

8. L603-605: Can you provide a few references to previous studies on this topic?

Author Response

Please see the attachment.

Author Response File: Author Response.docx