Global Sensitivity Analysis and Calibration by Differential Evolution Algorithm of HORTSYST Crop Model for Fertigation Management

Round 1

Reviewer 1 Report

Ms. Ref. No.: water-1123370

 

Title: Global sensitivity analysis and calibration by differential evolution algorithm of HORTSYST crop model for fertigation management

 

Journal: Water

 

The authors have a done a great job.

 However, the manuscript can be improved by considering the minor points further:

 

Specific comments

 

Currently, many of the statements are not supported by published works. Authors may like to find studies in line of their statements to add the scientific weight in their observations. For instance, at line 64-65: Authors can include some of the recent studies to discuss the application of genetic algorithm in the crop models. Some of the good reviews are given below which they can include to make their introduction better. Authors may like to supplement other paragraphs from the listed studies as well as other relevant studies.

 

Srivastava, A., Sahoo, B., Raghuwanshi, N. S., & Singh, R. (2017). Evaluation of variable-infiltration capacity model and MODIS-terra satellite-derived grid-scale evapotranspiration estimates in a River Basin with Tropical Monsoon-Type climatology. Journal of Irrigation and Drainage Engineering, 143(8), 04017028. https://doi.org/10.1061/(ASCE)IR.1943-4774.0001199

Srivastava, A., Sahoo, B., Raghuwanshi, N. S., & Chatterjee, C. (2018). Modelling the dynamics of evapotranspiration using Variable Infiltration Capacity model and regionally calibrated Hargreaves approach. Irrig Sci 36, 289–300 (2018). https://doi.org/10.1007/s00271-018-0583-y

Elbeltagi, A., Aslam, M. R., Malik, A., Mehdinejadiani, B., Srivastava, A., Bhatia, A. S., & Deng, J. (2020). The impact of climate changes on the water footprint of wheat and maize production in the Nile Delta, Egypt. Science of the Total Environment, 743, 140770. https://doi.org/10.1016/j.scitotenv.2020.140770

 

Line 77-79 how does the model take into account the albedo and atmospheric transmissivity into account as it will affect the net radiation components which in turn result in over or underestimation of the ET. Therefore, have you accounted for constant value of atmospheric transmissivity or does it changes based on different climatic regions.

 

Line 81-82 Is the LAI considered in the model is dynamic or constant and if its static then how the parameters will change accordingly. In addition, what are the soil moisture parameters taken into account and how does the seasonality of soil moisture is capture in the model.

 

Line 86-87 What are the advantages of the differential evolution algorithm technique over other existing state-of-art machine learning methods. It would be better if authors can provide a table detailing the selection criteria or scores of choosing this method over others.

 

 

 

 

Author Response

Dear: Reviewer 1. We want to thank you for the revision made to the manuscript water 1123370.

Point 1: Currently, many of the statements are not supported by published works. Authors may like to find studies in line of their statements to add the scientific weight in their observations. For instance, at line 64-65: Authors can include some of the recent studies to discuss the application of genetic algorithm in the crop models. Some of the good reviews are given below which they can include to make their introduction better. Authors may like to supplement other paragraphs from the listed studies as well as other relevant studies.

Answer 1: Other bibliography was added to support the application the technique to estimate parameters of the model and the introduction was improved.

Point 2. Line 77-79 how does the model take into account the albedo and atmospheric transmissivity into account as it will affect the net radiation components which in turn result in over or underestimation of the ET. Therefore, have you accounted for constant value of atmospheric transmissivity or does it changes based on different climatic regions.

Answer 2: The sub-model used in this research to determine crop transpiration (ETc) is a simplification of the Penman-Monteith model and does not consider the concept of albedo due to short- and long-wave radiation and does not consider the net radiation too. The model considers the global solar radiation (Rg) incident above crop. There is also no effect of atmospheric transmissivity because it is a crop growing under plastic cover.

Point 3. Line 81-82 Is the LAI considered in the model is dynamic or constant and if its static then how the parameters will change accordingly. In addition, what are the soil moisture parameters taken into account and how does the seasonality of soil moisture is capture in the model.

Answer 3: The Leaf Area index (LAI) variable is dynamic because its estimation depends on the photothermal time (PTI) which is a dynamic variable in the model. The model does not consider the substrate’ moisture variable, because it is not directly related to the estimation of crop transpiration (ETc). In addition, for development the HORTSYST model was considered a well-watered crop, without any type of stress, including water stress.  Although in later work to implement the model for irrigation and nutrient management, it is intended to couple the infiltration model such as Green & Ampt to estimate the moisture of the substrate.

Point 4. Line 86-87 What are the advantages of the differential evolution algorithm technique over other existing state-of-art machine learning methods. It would be better if authors can provide a table detailing the selection criteria or scores of choosing this method over others.

Answer 4. In the introduction is described the advantage of differential evolution (DE) technique over other methods, although it was not compared to the machine learning method because it is not the aim of the present research work.

Author Response File: Author Response.docx

Reviewer 2 Report

paragraph from line 54 to 60 must be clarified. it is repetitive in some way and not easy to follow.

Both the introduction and the conclusions do not clearly state what was the purpose of the work done. Hiw did the calibration of this model serve the purpose of predicting crop production. Evidence of the applicability of the model need to be further explained in the paper 

Author Response

Dear: Reviewer 2. We want to thank you for the revision made to the manuscript water 1123370.

Point 1. Paragraph from line 54 to 60 must be clarified. it is repetitive in some way and not easy to follow.

Answer 1: Paragraph 54-60 was rewritten to better understand the idea. Resulting as follow (L54-69):

The calibration of dynamic models is another important step in the development of models, which is carried out through an optimization problem, where an objective function is minimized or maximized [11, 12]. Model calibration is also required because not all parameters are directly measured [13]. And to solve this problem there are local and global optimization methods. The local method, using an iterative search starting from the parameter nominal value, may often trapped in a local minimum and prematurely terminate the search [14] and only allow exploring in a narrower domain of the nominal values of the parameters in the vicinity of a nominal value, regardless of whether the model has one or more optimal values of each parameter. On the other hand, the global optimization method for parameter estimation, the search range of the optimal values is extended, especially when the model being analyzed is little known for its recent development and it is unknown if it has a multimodal behavior (many optimal maximum and minimum optimal values).

This motivates the application of global parameter estimation methods in dynamics crop models, e.g., Covariance Matrix Adaptation Evolution Strategy (CMA-ES) [15], Genetic Algorithms (GA) [14], Particle Swarm Optimization (PSO) [16], Differential Evolution (DE) [17] and Artificial Bee Colony (ABC) [15]. These methods are heuristic optimization techniques that use bio-inspired concepts in biological evolution such as inheritance, mutation, selection and crossover [18].

Point 2. Both the introduction and the conclusions do not clearly state what was the purpose of the work done. How did the calibration of this model serve the purpose of predicting crop production. Evidence of the applicability of the model need to be further explained in the paper 

Answer 2: Paragraph from line 416 to 421in updated manuscript was ordered as follows:

The global sensitivity analysis based on Sobol’s method was an effective procedure for determining the most influential parameters in the HORTSYST model. With this procedure to evaluate de sensitivity analysis the number of parameters of the model were reduces from sixteen to nine, also it was realized that all parameters had a temporal variation in their sensitivity, which could be explained by each development stage of crop. Since it is known that dry matter production, nitrogen uptake, leaf area index and crop transpiration have nonlinear behavior.

Answer 2: Paragraph from line 430 to 439 was added to updated manuscript in order to link the goal, introduction with conclusion

The predictions of the model output variables during the calibration process were accurate, since the deviation or error between the simulated variables and the measurements were minimal. Therefore, the differential evolution (DE) technique used to estimate the parameters was effective and had good convergence, and its efficiency was computationally acceptable.

The HORTSYST model has a simple mathematical structure and due to the reduced number of influencing parameters (results of the sensitivity analysis) and its accurate prediction of the output variables during calibration and once validated, its implementation is feasible for irrigation and nutrient management in intensive crops. Since it uses of RUE (Radiation Use Efficiency) approach of some models mentioned above, which have been used as decision support systems (DSS) for the management of agricultural systems.

Author Response File: Author Response.docx