Abstract
Carbon dioxide (CO2) is considered one of the main greenhouse effect gases and contributes significantly to global climate change. In Brazil, the agricultural areas offer an opportunity to mitigate this effect, especially with the sugarcane crop, since, depending on the management system, sugarcane stores large amounts of carbon, thereby removing it from the atmosphere. The CO2 production in soil and its transport to the atmosphere are the results of biochemical processes such as the decomposition of organic matter and roots and the respiration of soil organisms, a phenomenon called soil CO2 emissions (FCO2). The objective of the study was to investigate the use of neural networks with backpropagation algorithm to predict the spatial patterns of soil CO2 emission during short periods in sugarcane areas. FCO2 values were collected in three commercial crop areas in the São Paulo state, southeastern Brazil, registered through the LI-8100 system during the years 2008 (Motuca), 2010 (Guariba city), and 2012 (Pradópolis), in the period after the mechanical harvesting (green cane). A neural network multilayer perceptron with a backpropagation algorithm was applied to estimate the FCO2 in 2012, using data from 2008 and 2010 as training for the neural network. The neural network initially presented a mean absolute percentage error (MAPE) of 18.3852 and a coefficient of determination (R2) of 0.9188. Data obtained from the observed and estimated values of FCO2 present moderate spatial dependence, and it is observed from the maps of the spatial pattern of the CO2 flow that the results from the neural network show considerable similarity to the observed data. The model results identify the higher and lower characteristics in sample points of CO2 emissions and produce an overestimation of the range of spatial dependence (0.45 m) and an underestimation of the interpolated values in the field (R2 = 0.80; MAPE = 12.0591), when compared to the actual soil CO2 emission values. Therefore, the results indicate that the artificial neural network provides reliable estimates for the evaluation of FCO2 from data of the soil’s physical and chemical attributes and describes the spatial variability of FCO2 in sugarcane fields, thereby contributing to the reduction of uncertainties associated with FCO2 accountings in these areas.
Similar content being viewed by others
References
Bicalho, E. S., Panosso, A. R., Teixeira, D. D. B., Miranda, J. G. V., Pereira, G. T., & La Scala, N. (2014). Spatial variability structure of soil CO2 emission and soil attributes in a sugarcane area. Agriculture Ecosystems & Environment, 189, 206–215. https://doi.org/10.1016/j.agee.2014.03.043.
Cerri, C. E. P., Sparovek, G., Bernoux, M., Easterling, W. E., Melillo, J. M., & Cerri, C. C. (2007). Tropical agriculture and global warming: impacts and mitigation options. Scientia Agricola, 64, 83–99. https://doi.org/10.1590/S0103-0162007000100013.
CONAB 2018. Available at: https://www.conab.gov.br/info-agro/safras/graos. Accessed date: 6 November 2018.
Dahikar, S. S., & Rode, S. V. (2014). Agricultural crop yield prediction using artificial neural network approach. International Journal of Innovative Research in Electrical, Electronics, Instrumentation and Control Engineering, 2(1), 683–686.
Deneshkumar, V., Kannan, S. and Manikandan, M. (2015). Designing of neural network models for agricultural forecasting. Available online at https://doi.org/10.1080/09720510.2015.1040237, pp. 547–559.
EMBRAPA. (1997). Brazilian Agricultural Research Corporation. In Manual of soil analysis methods (2nd. ed.). Brasília: MAPA (In Portuguese).
Epron, D., Bosc, A., Bonal, D., & Freycon, V. (2006). Spatial variation of soil respiration across a topographic gradient in a tropical rain forest in French Guiana. Journal of Tropical Ecology, 22, 565–574. https://doi.org/10.1017/S0266467406003415.
Figueiredo, E. B., & La Scala Júnior, N. (2011). Greenhouse gas balance due to the conversion of sugarcane areas from burned to green harvest in Brazil. Agricultura, Ecosystems & Environment, 141, 77–85. https://doi.org/10.1016/j.agee.2011.02.014.
Haykin, S. (1999). Neural networks: a comprehensive foundation. New Jersey: Prentice-Hall.
Herbst, M., Prolingheuer, N., Graf, A., Huisman, J. A., Weihermuller, L., & Vanderborght, J. (2009). Characterization and understanding of bare soil respiration spatial variability at plot scale. Vadose Zone Journal, 8, 762–771.
Herbst, M., Bornemann, L., Graf, A., Welp, G., Vereecken, H., & Amelung, W. (2012). A geostatistical approach to the field-scale pattern of heterotrophic soil CO2 emission using covariates. Biogeochemistry, 111, 377–392.
IPCC - Intergovernmental Panel on Climate Change. (2007). Climate change 2007: the physical science basis. Summary for policymakers, Geneva, Switzerland. Available in: http://www.ipcc.ch/publications_and_data/publications_ipcc_fourth_assessment_report_wg1_report_the_physical_science_basis.htm.
IPCC - Intergovernmental Panel on Climate Change. (2014). Climate change 2014: synthesis report for policymakers, Geneva, Switzerland. Available in: http://www.ipcc.ch/pdf/assessment-report/ar5/wg3/ipcc_wg3_ar5_full.pdf. Access in 23 abr. 2016.
Isaaks, E. H., & Srivastava, R. M. (1989). Applied geostatistics. Nova York: Oxford University Press.
Kartalopoulos, S. V. (1996). Understanding neural networks and fuzzy logic: basic concepts and applications. Piscataway, NJ, USA: IEEE Press.
Kim, M., & Gilley, J. E. (2008). Artificial neural network estimation of soil erosion and nutrient concentrations in runoff from land application areas. Computer and Electronics in Agriculture, 64, 268–275.
Konda, R., Ohta, S., Ishizuka, S., Arai, S., Ansori, S., Tanaka, N., & Hardjono, A. (2008). Spatial structures of N2O, CO2, and CH4 fluxes from Acacia mangium plantation soils during a relatively dry season in Indonesia. Soil Biology & Biochemistry, 40, 3021–3030.
Kosugi, Y., Mitani, T., Ltoh, M., Noguchi, S., Tani, M., Matsuo, N., Takanashi, S., Ohkubo, S., & Nik, A. R. (2007). Spatial and temporal variation in soil respiration in a Southeast Asian tropical rainforest. Agricultural and Forest Meteorology, 147, 35–47.
La Scala, N., Marques, J., Pereira, G. T., & Cora, J. E. (2000). Short-term temporal changes in the spatial variability model of CO2 emissions from a Brazilian bare soil. Soil Biology & Biochemistry, 32, 1459–1462.
La Scala, N., de Sá Mendonça, E., Vanir de Souza, J., Panosso, A. R., Simas, F. N. B., & Schaefer, C. E. G. R. (2010). Spatial and temporal variability in soil CO2-C emissions and relation to soil temperature at King George Island, maritime Antarctica. Polar Science, 4, 479–487.
Lawless, H. T., & Heymann, H. (2010). Data relationships and multivariate applications, sensory evaluation of food—principles and practices (pp. 433–449). Berlin: Springer.
Lentzsch, P., Wieland, R., & Wirth, S. (2005). Application of multiple regression and neural network approaches for landscape-scale assessment of soil microbial biomass. Soil Biology & Biochemistry, 37, 1577–1580. https://doi.org/10.1016/j.soilbio.2005.01.017.
Liu, H. (2010). On the Levenberg-Marquardt training method for feedforward neural networks. In Proceedings of the 2010 International Conference on Natural Computation, Icnc’10, volume 1.
Lopes, M. L. M., Minussi, C. R., & Lotufo, A. D. P. (2003). Electrical load forecasting formulation by a fast neural network. Engineering Intelligent Systems, 11(1), 51–57 http://hdl.handle.net/11449/9730.
Lotufo, A. D. P., Lopes, M. L. M., & Minussi, C. R. (2007). Sensitivity analysis by neural networks applied to power systems transient stability. Electric Power Systems Research, 77, 730–738. https://doi.org/10.1016/j.epsr.2005.09.020.
Luca, E. F., Feller, C., Cerri, C. C., Barthès, B., Chaplot, V., Campos, D. C., & Manechini, C. (2008). Evaluation of physical attributes and carbon and nitrogen stocks in soils with no burning of sugar cane. Brazilian Journal of Soil Science, 32, 789–800. (In Portuguese). https://doi.org/10.1590/S0100-06832008000200033.
Merdun, H. (2011). Self-organizing map artificial neural network application in multidimensional soil data analysis. Neural Computing and Applications, 20, 1295–1303. https://doi.org/10.1007/s00521-010-0425-1.
Moitinho, M. R., Padovan, M. P., Panosso, A. R., & La Scala, N. (2013). Effect of soil preparation and residue of the sugarcane harvest on the emission of CO2. Brazilian Journal of Soil Science, 37, 1720–1728. (In Portuguese). https://doi.org/10.1590/S0100-06832013000600028.
Moretti, J. F., Minussi, C. R., Melges, J. L. P., Akasakil, J. L., & Tashima, M. M. (2016). Prediction of modulus of elasticity and compressive strength of concrete specimens by means of artificial neural networks. Acta Scientiarum Technology, 38, 65–70. https://doi.org/10.4025/actascitechnol.v38i1.27194.
Ohashi, M., & Gyokusen, K. (2007). Temporal change in spatial variability of soil respiration on a slope of Japanese cedar (Cryptomeria japonica D. Don) forest. Soil Biology & Biochemistry, 39, 1130–1138.
Pandey, A., & Mishra, A. (2017). Application of artificial neural networks in yield prediction of potato crop. Russian Agricultural Sciences, 43(3), 266–272.
Panosso, A. R., Peillo, L. I., Ferraudo, A. S., Pereira, G. T., Miranda, J. G. V., & La, S. J. (2012). Fractal dimension and anisotropy of soil CO2 emission in a mechanically harvested sugarcane production area. Soil & Tillage Research, 124, 8–16. https://doi.org/10.1016/j.still.2012.04.005.
Rayment, M. B., & Jarvis, P. G. (2000). Temporal and spatial variation of soil CO2 efflux in a Canadian boreal forest. Soil Biology and Biochemistry, 32, 35–45. https://doi.org/10.1016/S0038-0717(99)00110-8.
Razafimbelo, T., Barthès, B., Larré-Larrouy, M. C., de Luca, E. F., Laurent, J. Y., Cerri, C. C., & Feller, C. (2006). Effect of sugarcane residue management (mulching versus burning) on organic matter in a clayey Oxisol from southern Brazil. Agriculture Ecosystem & Environment, 115, 285–289. https://doi.org/10.1016/j.agee.2005.12.014.
Santos, H. G., Jacomine, P. K. T., Anjos, L. H. C., Oliveira, V. A., Oliveira, J. B., Coelho, M. R., Lumbreras, J. F., & Cunha, T. J. F. (2013). Brazilian system of soil classification. Rio de Janeiro: Brazilian Agricultural Research Corporation (EMBRAPA) Soil.
Song, X., Peng, C., Zhao, Z. S., Zhang, Z., Guo, B., Wang, W., Jiang, H., & Zhu, Q. (2014). Quantification of soil respiration in forest ecosystems across China. Atmospheric Environment, 94, 546–551. https://doi.org/10.1016/j.atmosenv.2014.05.071.
Specht, D. F. (1991). A generalized regression neural network. IEEE Transactions on Neural Networks, 2, 568–576.
Stoyan, H., De-Polli, H., Bohm, S., Robertson, G. P., & Paul, E. A. (2000). Spatial heterogeneity of soil respiration and related properties at the plant scale. Plant and Soil, 222, 203–214.
Tavares, R. L. M., Siqueira, D. S., Panosso, A. R., Castioni, G. A. F., Souza, Z. M., & La Scala Jr., N. (2016). Soil management of sugarcane fields affecting CO2 fluxes. Scientia Agricola, 7, 543–551.
Tedeschi, V., Rey, A., Manca, G., Valentini, R., Jarvis, P. G., & Borghetti, M. (2006). Soil respiration in a Mediterranean oak forest at different developmental stages after coppicing. Global Change Biology, 12, 110–121.
Teixeira, D. B., Panosso, A. R., Cerri, C. E. P., Pereira, G. T., & La Scala, N. (2011). Soil CO(2) emission estimated by different interpolation techniques. Plant and Soil, 345, 187–194.
Teixeira, D. D. B., Bicalho, E. S., Panosso, A. R., Perillo, L. I., Iamaguti, J. L., Pereira, G. T., & La Scala, N. (2012). Uncertainties in the prediction of spatial variability of soil CO2 emissions and related properties. Brazilian Journal of Soil Science, 36, 1466–1475. https://doi.org/10.1590/S0100-6832012000500010.
Ussiri, A. N., & Lal, R. (2009). Long-term tillage effects on soil carbon storage and carbon dioxide emissions in continuous corn cropping system from an Alfisol in Ohio. Soil & Tillage Research, 104, 39–47. https://doi.org/10.1016/j.still.2008.11.008.
Raij, B. Van, Andrade, J. C., Cantarela, H., and Quaggio, J. A. (2001). Chemical analysis for the evaluation of tropical soils. Agronomic Institute of Campinas. (In Portuguese).
Wang, G., Gertner, G., Singh, V., Shinkareva, S., Parysow, P., & Anderson, A. (2002). Spatial and temporal prediction and uncertainty of soil loss using the revised universal soil loss equation: a case study of the rainfall-runoff erosivity R factor. Ecological Modelling, 153, 143–155. https://doi.org/10.1016/S0304-3800(01)00507-5.
Wasserman, P. D. (1989). Neural computing—theory and practice. New York: Van Nostrand Reinhold.
Webster, R., & Oliver, M. A. (1990). Statistical methods in soil and land resource survey. New York: Oxford University Press.
Werbos, P. J. (1974). Beyond regression: new tools for prediction and analysis in the behavioral sciences. [PhD thesis]. [Harvard]: Harvard University.
Widrow, B., & Lehr, M. A. (1990). 30 years of adaptive neural networks: perceptron, Madaline, and backpropagation. Proceedings of the IEEE, 78, 1415–1442. https://doi.org/10.1109/5.58323.
Funding
The authors gratefully thank the funding and support from CAPES, CNPq, and FAPESP (08/58187-0; 10/20364-9; and 13/24926-0).
Author information
Authors and Affiliations
Contributions
A.D.P.L., A.R.P., L.B.C., C.R.M., M.L.M.L., N.L.S.J., and R.L.B.F. contributed to the conception and/or design of the work. P.S.F., A.D.P.L., A.R.P., C.R.M., M.L.M.L., N.L.S.J., and R.L.B.F. participated in the conduction of the experiments and the acquisition of data. A.R.P. and A.D.P.L. performed data analysis. All authors contributed to the interpretation of data and to the drafting and revision of the manuscript, and approved the final version to be published.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Backpropagation algorithm
Backpropagation algorithm
The initial weights are usually adopted as random numbers (Widrow and Lehr 1990). The backpropagation (BP) algorithm consists of adapting the weights such that the network quadratic error is minimized. The sum of the instantaneous quadratic error of each neuron of the last layer (network output) is given by (Widrow and Lehr 1990):
where
- ε i :
-
di − yi;
- d i :
-
desired output of the ith element of the last layer;
- y :
-
output of the ith element of the last layer; and
- no :
-
number of neurons of the last layer.
Considering the ith network neuron and using the descent gradient method, the weight adjustments can be formulated by (Widrow and Lehr 1990):
where
- θi (r):
-
− γ [∇i (r)];
- γ :
-
stability control parameter or training rate;
- ∇i (r):
-
quadratic error gradient related to neuron i weights; and
- Γ i :
-
vector containing neuron i weights
- =:
-
[w0i, w1i, w2i… wni]T.
The adopted direction in Eq. (1012) to minimize the objective function of the quadratic error corresponds to the gradient opposite direction. The γ parameter determines the vector length ϕi(r). The sigmoid function is defined by (Widrow and Lehr 1990):
or
where
- λ :
-
constant that determines the slope of the function yi.
The variation of Eqs. (11) and (12) are (− 1,+ 1) and (0,+ 1), respectively.
Next, calculating the gradient as shown in Eq. (1012) and considering the sigmoid function defined by Eq. (11) or (12) and the momentum term, the following adaptation weight scheme is obtained (Lopes et al. 2003):
where
Πij weight corresponding to the connection with the ith and the jth neuron;
γtraining rate; and
ηmomentum constant (0 ≤ η < 1).
If the jth element is in the last layer, then:
where
If the jth element is in other layers, we have:
- Γ(j):
-
set of the element indices that are in the next layer to the jth element layer and are interconnected to the jth element.
The γ parameter that is used as a stability control for the iterative process is dependent on λ. The network weights are randomly initialized from the interval [0,1]. For convenience, the parameter γ (training rate) can be redefined by the following (Lopes et al. 2003):
Replacing Eq. (19) in Eq. (12) will “cancel” the amplitude dependency of σ related to λ. The σ amplitude will be maintained constant to every λ. This alternative is important considering that λ will only actuate in the left and right tails of σ. Equation (14) can then be written as the following:
The BP algorithm executes as follows (Widrow and Lehr 1990):
-
Step 1.
Present a pattern X to the network, which provides an output Y.
-
Step 2.
Calculate the error (difference with the desired value and the output) for each output.
-
Step 3.
Determine the backpropagated error by the network associated with the partial derivative of the quadratic error.
-
Step 4.
Adjust the weights of each element.
-
Step 5.
Present a new pattern to the network and repeat the process until the convergence is attained (according to a predefined tolerance).
Rights and permissions
About this article
Cite this article
Freitas, L.P.S., Lopes, M.L.M., Carvalho, L.B. et al. Forecasting the spatiotemporal variability of soil CO2 emissions in sugarcane areas in southeastern Brazil using artificial neural networks. Environ Monit Assess 190, 741 (2018). https://doi.org/10.1007/s10661-018-7118-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10661-018-7118-0