Predicting Habitat Properties Using Remote Sensing Data: Soil pH and Moisture, and Ground Vegetation Cover

Round 1

Reviewer 1 Report

Review of Remote Sensing manuscript 1934004 – “Predicting habitat properties using remote sensing data: soil pH and moisture, and ground vegetation cover.”

 

Manuscript Recommendation

This manuscript presents scientifically valid analyses of ecological parameters derived via remote sensed data and techniques.  The data are well-presented, receive appropriate statistical analyses, and conclusions are consistent with the results obtained. The manuscript needs minor revision to address key details in the material and methods that require greater explanation.  The manuscript is generally well-written, but is unclear or awkward in places. , Detailed review comments follow that outline suggested additions and and/or edits. The abstract is rephrased as an example to improve written clarity. The manuscript is appropriate for publication in Remote Sensing following revision.

 

Detailed review comments

 

Abstract and Introduction

 

See attached document that contains rephrased abstract and suggested edits to the introduction.

 

 

Materials and Methods

 

The site description indicates that both study areas contain more than one forest type or ecosystem (Area 1, coniferous forest and mixed forest, Area 2, temperate deciduous and smaller areas dominated by spruce and pine).  Figure1 shows the sampling point locations within each study area and the related discussion (lines 134-135) states that sampling locations were randomly generated. 

 

Sample locations in Figure 1 occur in clusters (not a completely random point generation), which suggests that sample clusters occur within a specific forest type with adjacent areas of different forest types or land use.   

 

The site description, map representation, and sampling locations of the forest types needs additional details, discussion, and/or figure representation. 

 

In specific:

Why are the sample points clustered?

Do the clusters represent areas of a certain forest type? If so, identify the area and forest type.

 How many specific forest types in total are included in the study?

Are sample locations random within each separate area of a forest type?   

 

The Materials and Methods need greater detail on soil depth

The authors use soil depth as a model variable with five soil depth measurements averaged for each  10 m plot.  How was soil depth measured?  Hand auger? Hydraulic probe? What criteria were used to determine soil depth? Depth to bedrock? Thickness of the solum (A and B horizon)? Depth to a limiting restriction (e.g., fragipan or water table)? Provide more info on these details.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

It´s of great interest to predict habitat poperties with RS and to analyse RS-indicators which have causal relationships to the landscape spatial differentiation, clear predictors are needed. It´s challenging to predict with models based on RS-data not only vegetation ground cover, but also highly variabel soil parameters as pH and soil moisture. In this study linear and multiple regression analysis could be successfull, because pH and soil moisture was grouped in rough classes (Ellenberg pH indicator, and drought tolerance classes with vegetation types. Limits of these estimations must be more discussed.

In Introduction importance of RS for monitoring biodiversity and habitats is well adressed. Predictors must be selected where correlational relationships are well understood - but in the case of vegetation type and soil moisture it´s not clear in the text. Field data from two areas were sampled with coniferous and deciduous forest. In methods missing a table, where sampling plots are grouped for vegetation types - was this in total analysed by statistics (P-package) or grouped by vegetation types? In area 2 sample plots looks lumped. Are plots excluded from regression an alysis for evaluation of the models? (can not be seen in the paper). With Lidar data 5 indices for topography and 4 indices for light conditions were build as predictor variables (see table 1). Rough maps for sediments were included and show in the statistical analysis high influence on pH and soil moisture. Limits of this additional information for spatial correlation must be more discussed. With R-package linear (pH) and regression models (soil moisture) were set up and in second step with multiple regression best predictors and their combinations were extracted. But for the best regression models a verification with some plots excluded from statistical analysis is missing!

The results describes the best regression models - very dry and wet sites are not predicted well - why? It´s wondering why as in other studies wetness index SWI is not included in the predictive models? - In discussion influence of sediment type with clay content is discussed in a plausible way. But missing a comparision with other studies where with DEV and SWI and sediment type soil moisture classes were detected in spatial prediction. What is the main favor to do this with RS and not only with DEM (SRTM-data) and vegetation type and sediment types?

Also it´s necessary to explain more the analysis steps of Lidar-data, how predictors are gained?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Added corrections and text are fine.