Forest Canopy Water Content Monitoring Using Radiative Transfer Models and Machine Learning

Round 1

Reviewer 1 Report

This paper presents an evaluation of radiative transfer models for estimating Canopy Water Content (CWC) in complex forest canopy. They compare the performance of three coupled radiative transfer models, parameterizing them with three existing datasets, training and validating the models using Random Forests-based model inversion on MODIS bands, and using MODIS (from Google Earth Engine) products to bring in the spatiotemporal variation of CWC over the continental US (CONUS). As far as my knowledge of radiative transfer models and forests permits, the methods used are valid and clearly presented. The comparison of radiative transfer models’ performance is a needed contribution as the estimation of CWC in forests, which present complex canopy structures, currently needs improvement to be able to monitor, respond, and better understand forest growth and mortality under drought conditions and other risks that come with climate change. The results demonstrate the better performance of the 3D radiative transfer model given the broad category of plant functional type, the low sample size of the LOPEX93 and ANGERS parameterizing datasets compared to application (CONUS), and other generalizing model assumptions (example: grass background).  The authors provide improved methods (using Gaussian copula) and using RF-based model-inversion, compared model performance between three radiative transfer models while recognizing some of the still-existing limitations which come from a gross abstraction of forest complexity and heterogeneity.

I would like to thank the authors for a well written paper and important contribution. I offer a few minor editorial notes, a few questions, and suggestions that I hope would improve this manuscript.

lines 45-46 – I do not understand this sentence (starts at the end of line 45). Maybe you wanted to say “as” instead of “and”. Please revise.

line 58 – I suggest you specify that radiative transfer models are physical models.

line 61 – define what an ill-posed problem – example from the internet that would suffice “In an inverse problem we draw conclusions about the cause from its observed (measured) effect. This kind of task is typically ill-posed, that is, small changes (noisy measurements) in the effects lead to dramatic changes in the corresponding causes.”

line 97 – Artemisia tridentata is a shrub, not a tree species. It does not occur within forest, but in gaps in very dry non-continuous canopy forests (not in ENF, EBF nor in DBF).

lines 31-169 - Please specify here the number of ENF, EBF and DBF NEON site that you use in your parameterization. Also, this description of the NEON datasets could be shortened (example: do we need to specify that the samples are sealed and places in coolers? It suffices to say that Physical, chemical, and stable isotope data are collected for plant foliage).

line 221 – “reflectance of grass instead of bare soil” is an improvement but grass is not often a ground cover in forests. Herbaceous layers and shrub dominate the ground cover in most of the CONUS. This point could be added to the discussion.

lines 239-245 – why is this paragraph in bold?

line 245 – grass is not the ground cover in most forest in the CONUS. The ground cover is more frequently a shrub and herbaceous layer. Please change the title in S1 to herbaceous as opposed to “herb” and change the grass reference on this line to match.

line 247 – “tree” does that mean all trees or was this done by plant functional type? Is there not enough data to do this by plant functional type?

line 253 – “exhibit better prediction ability” - than what? other empirical methods? Please complete the thought.

lines 261-262 - Can you show the resulting CWC by PFT (PFT as per the map presented in Figure 1)?

lines 389-401 - There are no results presented that support this discussion of the CWC by PFT. Could you add some supporting tables or figures?

Line 400 – “Due to the limitation of the data and study area,…” how is the study area limited? Do you mean the limited available data given the size of the study area?

Discussion point

Plant functional types are a coarse representation of forests. I suggest that you add some discussion regarding how we would more towards more species-specific characterization of CWC.

Both the LOPEX93 and ANGERS dataset are a very small sample of forests when considering the CONUS. Please add some discussion of what we are missing because of this low sample rate, and what error this brings to our estimates.

Author Response

We appreciate your insightful comments and suggestions to the manuscript. We have carefully revised the manuscript following your comments.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The article "Forest Canopy Water Content Monitoring Using Radiative

Transfer Models and Machine Learning", deals with the water content that forests store, and relates it to climate change. The authors highlight the danger that forest ecosystems have in the face of drought.

The article is within the scope of Forest magazine, and is well structured. The results and discussion are consistent with the proposed objectives.

Author Response

We appreciate your insightful comments and suggestions to the manuscript. And we have carefully revised the manuscript following your comments.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Dear Authors,

Please find my comments in the attached file.

Also i would like to highlight that you need  the improval in the results part. I think for each prediction you have to estimate additionally bias, precision and accuracy.

Regards

Comments for author File: Comments.pdf

Author Response

We appreciate your insightful comments and suggestions to the manuscript. And we have carefully revised the manuscript following your comments.

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Dear Authors,

Thank you for taking my comments into account.

Regards