Characterization of Background Temperature Dynamics of a Multitemporal Satellite Scene through Data Assimilation for Wildfire Detection

Round 1

Reviewer 1 Report

The authors give very detailed description of the theory and methods for remote sensing data processing considering estimation of pixel background brightness temperature and ensemble forecasting-based change detection.

Lines 163-572: These sections take significant portion of the manuscript. Possibly, some of these materials could be moved to the Supplementary Materials section without reduction of overall manuscript quality. This could help to emphasize the outcomes from the research in the context of the research objectives: “...a method that assimilates brightness temperatures acquired from the Geostationary Earth Orbit (GEO) sensor MSG-SEVIRI into a Diurnal Temperature Cycle (DTC) model” (Line 5-6).

Some notes:

What is the accuracy of correspondence between the recorded (simulated) temperature and a specific image pixel considering possible spatial discrepancy of data at the level of spatial resolution of remote sensing system?

Is the proposed approach applicable for monitoring fires at high latitudes? What satellite systems can be an alternative to data from geostationary satellite systems?

Most of equations (e.g. 10-24, 27-33) are not cited in the text. Could they be excluded from the main sections of manuscript?

Why Listing 1 (Line 432) and Listing 2 (Line 599) are not placed into the Supplementary section?

Currently the manuscript is arranged as a review of Ph.D thesis. This is confirmed by authors as well. See Line 1028: "The research was conducted under the doctoral study of G.U.". Pre-processing of the manuscript into a scientific article format can significantly improve the understanding of results by readers. It is likely that the discussion of practical results and the possibility of their application is more significant for remote sensing and fire monitoring specialists.

Author Response

Please see the attachment.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper develops several data assimilation techniques for a Diurnal Temperature Cycle (DTC) to aid wildfire detection. The primary focus of this is improving the characterisation of background temperature dynamics based on DA. Using a DA framework allows for the DTC forecasting model to be combined with observations from SEVIRI in an optimal manner.  Three DA methods are considered 1) weak-constraint 4D-Var is time-only which provides a point estimate of background brightness temperature and 2) classical EnKF and 3) SIR particle filter which provides the full posterior pdf. The paper then develops a fire detector algorithm based on the background temperature assimilation forecasts using a contrast threshold which adapts to limit the false positive rate. The method is shown to provide good detective performance for small/cool fires occurring at any time of day. 

General comments

The methods are rigorously laid out. I enjoyed the care the authors took in relating different DA methods, for example, the explanation of the EnKF in relation to a Kalman filter. This link between different methods and the reasons for choosing methods (normality assumptions, linearity, computation) arent always explained well in papers which contrast several methods. 

The link between the estimation procedures for the DTC with DA and then the fire detector is done nicely.  For example, the link between the variance of the t-1 DTC solution and the CFAR threshold for t is a nice method to link the distribution/uncertainty outputs of the DA directly into the detection method. Demonstrating that the DA approaches improve the ability to rapidly detect fire start and cool fires is an interesting result. 

The work is a novel and I have no issues with the MS. 

Corrections:

577 “using anomaly detection” → “using an anomaly detection”

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is very interesting proposing a new method that assimilates brightness temperatures acquired from the GEO sensor MSG-SEVIRI into a DTC model. This method is more adequate to determine background temperature of a pixel without overestimating the results in the omission of a fire event. The method has a high detection rate and has been successfully validated. Thus the implementation of this new method in wildfire detection is very much possible.

However, the main problem of this paper is that it is far too long. Authors should move a part of Results and Discussion and also a part of Methods to the Supplementary Materials. Likewise, about a half of all the figures could be moved to the Supplement.

Thus the paper needs at least minor revision regarding the structural changes. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors taken into account and commented all my questions in the response.

The Response 3 and 4 (see Coverletter) could be interesting to readers of the article as well. Perhaps a discussion of these points could be included in the text of the manuscript.

In my opinion, the article should be recommended for publication.

Comments for author File: Comments.pdf