Improving an Urban Cellular Automata Model Based on Auto-Calibrated and Trend-Adjusted Neighborhood
Round 1
Reviewer 1 Report
After reading the paper again and checking that the authors have implemented the suggestions that I mentioned in the first review, I consider that the paper is ready to be published.
Reviewer 2 Report
Excellent manuscript, almost ready, I spotted just one typo:
line 225 should be "weights" not "wights"
Suggest just to give a final very careful proof read through to ensure no further typos.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
In this paper the authors present an algorithm that try to improve the method for calculating the neighborhood of the urban expansion model with the goal of enhancing the urban expansion simulation accuracy.
In my opinion the paper must be change the following:
- In general, there are quite a few grammatical and expression errors that need to be corrected. It would be convenient, for the paper, to be reviewed by a native English.
- The title is too long, should be shortened.
- The abstract must be change because of it do not is a summary of the paper. In my opinion, an abstract should convey the main results and conclusions of a scientific paper because it communicates complex investigations and can act as an independent entity instead of an entire article. In the case of this article, clearly, this does not happen. It is the very extensive, there are many sentences left over. There are phrases that must be removed as the last sentences: “The average overall accuracy, Kappa,…, respectively”.
- Regarding to the Introduction, it does not put in perspective the motivation and objectives. What is the purpose of the paper? What motivates the authors to write this paper? These two questions should be clearly answered in the first section of the paper. In addition, there is no state of the art in this section. There are very long and intelligible sentences and quite a few grammatical and syntax errors. For instance: in page 2, first paragraph, the last sentence has words with the first letter in capital letters. The second paragraph contains sentences that are not understood. In short, after reading the introduction it is not clear what the authors want to do, what the objective is, etc. The motivation is missing, why is the article written? What does it mean that they don't have other articles already written? What are the objective of the paper?
- In my opinion the section 2 should be remove or, better, included in the methodology. Subsection 2.1 does not make sense on its own, it is a simple table or figure. Figure 1 would sufficient to explain this section. Subsection 2.2 must be explain with other simple table.
- The main section of the article is the methodology. Firstly, the authors have to explain what they want to do, including a flow chart at the beginning of the section. Then is mandatory to explain step by step what you intend to do and its bases. This section should be redone from start to finish. There are errors in the formulas, the notation is very confusing since normally capital letters usually represent matrices and in this case it seems that they do not, the elements of the formulas are not aligned, there are references to appendices that do not exist (Appendix S1), etc.
- The Results section is very bad, it is a simple table with almost no explanation.
- A discussion of the results should be more clearly explaining, what is the improvement of this method compared to others?.
The theoretical part of the article has potential, but it must be presented much better, abstract, introduction, methodology and experimental results must be redone or greatly improved.
Author Response
Issue 1: In general, there are quite a few grammatical and expression errors that need to be corrected. It would be convenient, for the paper, to be reviewed by a native English.
Response: Done as suggested. The manuscript was edited for proper English language, grammar, punctuation, spelling, and overall style by one or more of the highly qualified native English speaking editors at SpringerNature Author Services (SNAS). This can be verified on the SNAS website (https://secure.authorservices.springernature.com/certificate/verify) using the verification code D060-19EB-C576-89B1-654P.
Issue 2: The title is too long, should be shortened.
Response: Done as suggested. The title was shorten as “Improving an urban cellular automata model based on au-to-calibrated and trend-adjusted neighborhood”.
Issue 3: The abstract must be change because of it do not is a summary of the paper. In my opinion, an abstract should convey the main results and conclusions of a scientific paper because it communicates complex investigations and can act as an independent entity instead of an entire article. In the case of this article, clearly, this does not happen. It is the very extensive, there are many sentences left over. There are phrases that must be removed as the last sentences: “The average overall accuracy, Kappa,…, respectively”.
Response: Done as suggested. The abstract was revised as “Accurately simulating urban expansion is of great significance for promoting sustainable urban development. The calculation of neighborhood effects is an important factor that affects the ac-curacy of urban expansion models. The purpose of this study is to improve the calculation of neighborhood effects in an urban expansion model, i.e., the land use scenario dynamics-urban (LUSD-urban) model, by integrating the trend-adjusted neighborhood algorithm and the automatic rule detection procedure. Taking eight sample cities in China as examples, we evaluated the accuracies of the original model and the improved model. We found that the improved model can increase accuracy of simulated urban expansion in terms of both the degree of spatial matching and the similarity of urban form. The increase of accuracy can be attributed to that such integration comprehensively considers the effects of historical urban expansion trends and the influences of neighborhoods at different scales. Therefore, the improved model in this study can be widely used to simulate the process of urban expansion in different regions.”
Issue 4: Regarding to the Introduction, it does not put in perspective the motivation and objectives. What is the purpose of the paper? What motivates the authors to write this paper? These two questions should be clearly answered in the first section of the paper. In addition, there is no state of the art in this section. There are very long and intelligible sentences and quite a few grammatical and syntax errors. For instance: in page 2, first paragraph, the last sentence has words with the first letter in capital letters. The second paragraph contains sentences that are not understood. In short, after reading the introduction it is not clear what the authors want to do, what the objective is, etc. The motivation is missing, why is the article written? What does it mean that they don't have other articles already written? What are the objective of the paper?
Response: Clarified and revised. The motivation of this study is that the calculation of neighborhood effects in CA-based UE models still needs further improvement, and two new approaches (i.e., the trend-adjusted neighborhood algorithm and the automatic rule detection procedure) were developed to improve such calculation of neighborhood effects. Thus, we performed this study to integrate these two approaches to further improve the calculation of neighborhood effects. We revised the Introduction to clarify the motivation and the objective of this study. Please refer to the revised Introduction part.
Issue 5: In my opinion the section 2 should be remove or, better, included in the methodology. Subsection 2.1 does not make sense on its own, it is a simple table or figure. Figure 1 would sufficient to explain this section. Subsection 2.2 must be explain with other simple table.
Response: Done as suggested. We moved Section 2 into Methods and added a table to explain data sources. Please refer to Methods and Table 1.
Issue 6: The main section of the article is the methodology. Firstly, the authors have to explain what they want to do, including a flow chart at the beginning of the section. Then is mandatory to explain step by step what you intend to do and its bases. This section should be redone from start to finish. There are errors in the formulas, the notation is very confusing since normally capital letters usually represent matrices and in this case it seems that they do not , the elements of the formulas are not aligned, there are references to appendices that do not exist (Appendix S1), etc .
Response: Clarified and revised. We revised Methods according to the reviewer’s suggestions. Please refer to Figure 1 and Methods. Since our study was performed on the basis of pervious studies on LUSD-urban model, the trend-adjusted neighborhood algorithm, and the automatic rule detection procedure. The formulas and the capital letters were used to be consistent with these studies.
Issue 7: The Results section is very bad, it is a simple table with almost no explanation .
Response: We rewrote the Results section to highlight the improvement of accuracy of simulated urban expansion in terms of different indicators and various sample cities. Please refer to the revised Results.
Issue 8: A discussion of the results should be more clearly explaining, what is the improvement of this method compared to others?
Response: We revised the Discussion section to make this part more clearly explaining. In addition, the advantage of integrating the trend-adjusted neighborhood algorithm and automatic rule detection procedure is that it both considers the urban expansion trend and calibrates the parameters related to the neighborhood attenuation law. Please refer to the revised 4.1 section for the details.
Issue 9: The theoretical part of the article has potential, but it must be presented much better, abstract, introduction, methodology and experimental results must be redone or greatly improved.
Response: Done as suggested. We rewrote Abstract, Methods, and Results, and revised Introduction according to the reviewer’s suggestions.
Reviewer 2 Report
A well-written paper, presenting a clear objective which it sets out to acheive in a logical and coherent way.
Strengths.
The emphasis on neighbourhood as the most important factor in these kinds of models is correct, in my opinion. I agree with the authors that there is plenty of room for improvement in the calibration of these models, so the paper addresses a real problem and presents a worthwhile solution. The model presented is nice and straightforward, following the principles of ockham’s razor, it doesn’t add unnecessary complications or try to blind the reader with magical solutions obtained using obscure algorithms. As a result I recommend publication, once the following weaknesses have been addressed.
Weaknesses
Two key weaknesses:
1. Line 154-6 “We employed the adaptive Monte Carlo method for model calibration, that is, by repeatedly simulating the regional historical UE, using the kappa coefficient as the evaluation criterion” I fully support the workflow used, and it is clearly explained. But the focus on the kappa statistic as the only indicator of accuracy is questionable. Kappa is problematic, for various reasons, though its limitations are probably overstated (see various papers by R Gil Pontius vs The Rest of the World). But whether or not you think Kappa is good or not, it’s a single indicator, and it doesn’t, for example, tell you anything about pattern, which is rather important in simulating urban form. Another disadvantage to Kappa (shared by accuracy and AUC) is that it’s going to be very insensitive if most of your study area is white space. The easiest way to check this is to calculate hits, misses and correct rejects and see if the correct rejects are more than twice as large the hits and misses. If so, I suggest using a statistic that doesn't include white space. I'll make a suggestion in the next para.
2. Demand. In these models, calibration is usually carried out by simulating the number of cells difference between the start map and the target map i.e. the UE area, known as the land use demand. This means that you are using a constrained cellular automata model, such that at each time step the quantity of land to be allocated is determined in advance. This means that statistics which measure the quantity aspect of model accuracy are not really appropriate. So precision is out, as well as accuracy and AUC, because these statistics all use false alarms (FAs). FAs are not helpful in a constrained CA model because we already told the model how much land to allocate. The stat you want here is the inverse – ie. the number of times the model does not allocate into areas where there really was change, usually known as "misses". There is a very appropriate statistic for this which is called "recall" or "Hit rate" and is basically Hits/Hits+Misses. It's simple and obvious, not patented by some dogmatic idiot who just wants you to cite their papers and it tells you how good the model is at hitting the target. Nothing else. For pattern, you have a gigantic choice of mostly overlapping statistics, but patch level fractal dimension index, class level clumpiness and class level edge density are all good. So, how to address these 2 weaknesses?
(i). State how demand was obtained, and whether or not the CA model is a constrained CA model (I'm fairly certain that it is, but you should state this clearly)
(ii) In the adaptive monte carlo method, use Hit rate (recall) instead of kappa, and instead of just using only this, add fractal dimension, clumpiness, or edge density and choose the best fit from the two stats with a ranking procedure. I suggest a geometric mean to avoid giving too much weight to either statistic. Selecting the best-fitting comparison based on e.g. hit rate and fractal dimension will nicely optimise pattern against cell by cell accuracy. Note that the Roodposhti et al paper you cite uses a ranking procedure in this way.
Author Response
Issue 1: State how demand was obtained, and whether or not the CA model is a constrained CA model (I'm fairly certain that it is, but you should state this clearly).
Response: Done as suggested. In the processes of calibration and simulation, the urban land demand for each sample city was obtained from the long-term urban built-up area data set. In Lines 229-231, we clarified that “In the processes of calibration and simulation, the urban land demand for each sample city was obtained from the long-term urban built-up area data set released by Gong et al.”. In addition, the LUSD-urban model is a constrained CA model. In Line 91 in Introduction, we stated that “To achieve this goal, we first improved the method of calculating the neighborhood effects based on the LUSD-urban model, a constrained CA-based UE model, by integrating the trend-adjusted neighborhood algorithm and ARD procedure.”
Issue 2: In the adaptive monte carlo method, use Hit rate (recall) instead of kappa, and instead of just using only this, add fractal dimension, clumpiness, or edge density and choose the best fit from the two stats with a ranking procedure. I suggest a geometric mean to avoid giving too much weight to either statistic. Selecting the best-fitting comparison based on e.g. hit rate and fractal dimension will nicely optimise pattern against cell by cell accuracy. Note that the Roodposhti et al paper you cite uses a ranking procedure in this way.
Response: Done as suggested. We selected the recall coefficient and four landscape indices, including edge density (ED), landscape shape index (LSI), fractal dimension index (FDI) and clumpiness index (CLUMPY) to construct an evaluation system and redone our work according to the reviewer’s suggestions. Please refer to the revised Methods and Results for the details.
Reviewer 3 Report
The scopes of journal "Land" are land use/land change, land management, land system science, landscape, soil-sediment-water systems, urban contexts and urban-rural interactions, and land–climate interactions, etc. But the purpose of presented research is to integrate the trend-adjusted neighborhood algorithm and the ARD procedure to improve the simulation accuracy.
For example Algorithms (ISSN 1999-4893) is more suitable journal which provides an advanced forum for studies related to algorithms and their applications. Other example is Mathematical and Computational Applications (ISSN 2297-87) on the applications of the mathematical and/or computational techniques
Authors should add information about why this study is needed, how it can help solve problems land management, land system science, landscape, soil-sediment-water systems, urban contexts and urban-rural interactions, and land–climate interactions, etc. with appropriate examples
Author Response
Issue 1: The scopes of journal "Land" are land use/land change, land management, land system science, landscape, soil-sediment-water systems, urban contexts and urban-rural interactions, and land–climate interactions, etc. But the purpose of presented research is to integrate the trend-adjusted neighborhood algorithm and the ARD procedure to improve the simulation accuracy.
For example Algorithms (ISSN 1999-4893) is more suitable journal which provides an advanced forum for studies related to algorithms and their applications. Other example is Mathematical and Computational Applications (ISSN 2297-87) on the applications of the mathematical and/or computational techniques.
Response: Clarified. This study was submitted to the special issue “New Approaches to Land Use/Land Cover Change Modeling” (https://www.mdpi.com/journal/land/special_issues/LULCC_model) of the journal “Land”. Our topic “improvement of urban land modeling” fits the scope of this special issue.
Issue 2: Authors should add information about why this study is needed, how it can help solve problems land management, land system science, landscape, soil-sediment-water systems, urban contexts and urban-rural interactions, and land–climate interactions, etc. with appropriate examples
Response: Clarified and revised. The accurate modeling of urban land expansion is essential to evaluate the social and environmental effects of urban growth as well as sustainable urban land management and planning. We revised the Abstract and Introduction to clearly declare this point. Please refer to the Abstract and the first sentence in the Introduction.
