پهنه‌بندی رقومی شوری خاک سطحی با بکارگیری مدل جنگل تصادفی در اراضی شور دشت ایوانکی.

نوع مقاله : مقاله پژوهشی

نویسندگان

گروه علوم و مهندسی خاک، دانشکده کشاورزی، دانشگاه تهران، تهران، ایران.

چکیده

هدف این مطالعه بررسی تغییرات مکانی شوری خاک با استفاده از مدل RF در بخشی از دشت ایوانکی (استان سمنان، 1398) بود. تعداد 104 نمونه به روش شبکه (فاصله 100 متر)، از 105 هکتار خاک‌های، واقع بر روی مارن و آبرفت‌های سنگریزه‌دار و کاربری‌ پسته کاری با آبیاری جویچه‌ای و اراضی رها انجام شد. بیشترین EC خاک در اراضی رها شده و باغ پسته به ترتیب 2/173و 34 dSm-1 بود. عوامل شوری مواد مادری، کیفیت آب آبیاری، PET زیاد و خیز مویینه املاح بود. ضریب تبیین (R2) نقشه پیش‌بینی شوری توسط مدل RF مساوی 49/0 و مهم‌ترین شاخص‌های کمکی، شوری نرمال شده، خیسی توپوگرافی، سطح مبنای زهکش، پوشش گیاهی نرمال شده و پوشش گیاهی تعدیل شده خاک بودند. شاخص‌های نسبت‌ طیفی داده‌های لندست 8، در پیش‌بینی تغییرات شوری اهمیت زیادی داشتند. از 5 متغیر کمکی موثر در مدل، 3 متغیر مربوط به شاخص‌های نسبت‌ طیفی بود. دلیل اهمیت شاخص‌های نسبت‌ طیفی در مدل، تجمع نمک در سطح خاک، و کاهش سهم متغیرهای زمین‌نما به دلیل مسطح بودن منطقه بود. کاربرد NDVI به تنهایی برای مطالعات شوری کافی نیست و استفاده از شاخص‌های شوری و رطوبت برای پیش‌بینی صحیح ضروری است. بررسی همبستگی بین متغیر‌های کمکی و اجرای مدل حذف برگشتی نشان داد که متغیرهای کمکی زیاد، سبب افزایش پیچیدگی و خطا در پیش‌بینی می­شود. روش حذف برگشتی با شناسایی مهم‌ترین متغیرها به ساده‌سازی مدل کمک کرد. نقشه پیش‌بینی شوری با مدل جنگل تصادفی با مشاهدات میدانی تطابق داشت و منطقه بحرانی شوری را به خوبی مشخص نمود.

کلیدواژه‌ها

موضوعات


عنوان مقاله [English]

Soil salinity digital mapping using random forest model in saline lands of Eyvanekey plain

نویسندگان [English]

  • Leila Jahanbazi
  • Ahmad Heidari
  • Mohammad Hosein Mohammadi
Soil Science Department Faculty of Agriculture,, University of Tehran, Karaj,
چکیده [English]

This study aimed to investigate the spatial changes of soil salinity using RF model in a part of Eyvanekey Plain (Semnan Province 2018). Grid sampling with 100 m intervals (106 samples) was taken from 105 ha of soils developed on marl and gravely alluviums. The land uses were pistachio plantations with furrow irrigation and abandoned land. The maximum EC was (173.2 and 34 dS/m) in the abandoned and furrow irrigation pistachio plantations respectively. The main factors of salinization were saline marls, saline irrigation water, and high PET. The R2 for the salinity prediction map by RF model was 0.49, and the most important covariates were normalized difference salinity index (NDSI), topographic wetness index (TWI), Channel Network Base Level (CNBL), normalized difference vegetation index (NDVI), and modified soil vegetation index (SAVI). Spectral ratio indices derived from Landsat 8 contributed the most to the soil salinity prediction. Out of 5 main auxiliary variables, 3 variables are related to spectral ratio indices and the reason was the presence of salt on the soil in the studied area. Using NDVI with other salinity and moisture indices improved the salinity prediction model. Examining the results of covariates correlation and the implementation of recursive feature elimination showed that many covariates increase model complexity and prediction error. Recursive feature elimination helped to simplify the model by identifying the most important covariates. The salinity prediction map by random forest was consistent with the field observations and clearly defined the critical saline area.

کلیدواژه‌ها [English]

  • Digital soil mapping
  • Furrow irrigation
  • Soil properties
  • Spatial changes
  • Spectral ratio

Soil Salinity Digital Mapping using Random Forest Model in Saline Lands of Eyvanekey Plain

EXTENDED ABSTRACT

Introduction

Soil salinization is a prevalent form of land degradation in arid regions which threatens soil productivity, agricultural sustainability, and food security. Since, the salinity can be caused by anthropogenic activities, in addition to natural sources, many studies have been conducted to study the spatial and temporal changes and their constituents. Monitoring and evaluating spatiotemporal dynamics of soil salinity are important due to high variability of salinity. The classical methods of study of soil salinity are time and cost-consuming compared to digital soil mapping. Monitoring and lab works in digital soil mapping could be reduced.

Purpose

The purpose of this study was characterizing and mapping secondary salinity by using random forest model at a field scale in Eyvanekey plain for precision and sustainable management of saline soils in study area.

Research method

106 surface samples were taken in 2018 by grid with 100 m intervals from 105 ha of soils developed on marl and gravely alluviums parent materials and the necessary measurements were made on them. The land uses were pistachio tree with furrow irrigation and abandoned land. According to SCORPAN model, soil salinity data as target variable and spectral indices derived from satellite images and train attributes derived from DEM as environmental covariates were prepared in 30 m resolution. Random forest model was used to connect soil salinity data and environmental covariates to predict salinity map. Also recursive feature elimination method was used to determine the suitable amount and type of environmental covariates for avoiding prediction model complexity and error.

Results

The hot and arid climate combined with other environmental parameters demonstrate the unsuitability of the study area for agricultural purposes. All the study area was classified as hypersaline soil (ECe > 4 dS/m). EC amount range was between 4.25 to 173.2 dS/m with mean amount 51.98 dS/m. According to salinity map the highest values of ECe were at the middle of the field, in pistachio trees land use, followed by the abandoned area located on the downward slope that receives the runoff water which has passed through the higher saline area. The main factors of salinity in study area were saline parent materials, the quality of irrigation water, high evaporation and transpiration and capillary rise of solutes. The coefficient of determination (R2) of the salinity prediction map by random forest model was equal to 0.49 and the most important covariate for salinity mapping were normalized difference salinity index (NDSI), topographic wetness index (TWI), Channel Network Base Level (CNBL), normalized difference vegetation index (NDVI), and modified soil vegetation index (SAVI). Spectral ratio indices derived from Landsat 8 had an important contribution in the soil salinity prediction model and out of 5 main environmental covariates, 3 were related to spectral ratio indices. Also, the results showed that use of NDVI withother spectral ratio indices like salinity and moisture indices improved salinity prediction model. In general, the use of spectral ratio indices derived from satellite images are very useful for salinity studies due to provide an overview of the salinity variation in study area, but always the uncertainty of the data should be considered. It should be considered that the amount and mineralogy of salt, soil moisture, soil colour and roughness could effect on salt reflection and as a result on the remote sensing derived data. Examining the results of environmental covariates correlation and the implementation of recursive feature elimination showed that the presence of many environmental covariates increased model complexity and prediction error. Recursive feature elimination results helped to simplify the model by identifying the most important environmental covariates for making random forest model. The salinity prediction map by random forest was consistent with the field observations and clearly defined the critical saline area in middle part of the field.

Conclusion

The availability of suitable land for agriculture use in arid region is scarce and having information about limitation factors like salinity is essential to prevent soil degradation. To improve the existing fragile conditions, characterize saline soils and monitoring salinity changes is necessary. salinity map produced by RF was reliable and helped to see the extent and severity of salinity in each single pixel of study area. The salinity prediction map was in consistent with the field observations and showed the middle part of the study area as a critical saline area. The digital nature of these maps allows information to be updated at a lower cost and faster in future in the study area.

Abbas, A., Khan, S., Hussain, N., Hanjra, M. A., & Akbar, S. (2013). Characterizing soil salinity in irrigated agriculture using a remote sensing approach. Physics and chemistry of the Earth, Parts A/B/C, 55, 43-52.
Aksoy, S., Yildirim, A., Gorji, T., Hamzehpour, N., Tanik, A., & Sertel, E. (2022). Assessing the performance of machine learning algorithms for soil salinity mapping in Google Earth Engine platform using Sentinel-2A and Landsat-8 OLI data. Advances in Space Research69(2), 1072-1086.
Alarape, M. A., Ameen, A. O., & Adewole, K. S. (2021). Hybrid students’ academic performance and dropout prediction models using recursive feature elimination technique. In Advances on Smart and Soft Computing: Proceedings of ICACIn 2021 (pp. 93-106). Singapore: Springer Singapore.
Al-Khaier, F. (2003). Soil Salinity Detection Using Satellite Remote Sensing, ITC MSc. Thesis, Supervisor: Bastiaanssen, ITC, Netherlands.
Allbed, A., Kumar L., & Sinha, P. (2014). Mapping and Modeling Spatial Variation in Soil Salinity in the Al Hassa Oasis Based on Remote Sensing Indicators and Regression Techniques. Remote Sensing, 6, 1137-1157.
Dos Santos, E. P., da Silva, D. D., do Amaral, C. H., Fernandes-Filho, E. I., & Dias, R. L. S. (2022). A Machine Learning approach to reconstruct cloudy affected vegetation indices imagery via data fusion from Sentinel-1 and Landsat 8. Computers and Electronics in Agriculture194, 106753.
Elhag, M. (2016). Evaluation of different soil salinity mapping using remote sensing techniques in arid ecosystems, Saudi Arabia. Journal of Sensors2016.
Finke, P. A. (2012). On digital soil assessment with models and the Pedometrics agenda. Geoderma171, 3-15.
Gao, B. C. (1996). NDWI—A normalized difference water index for remote sensing of vegetation liquid water from space. Remote sensing of environment58(3), 257-266.
Guyon, I., Weston, J., Barnhill, S., & Vapnik, V. (2002). Gene selection for cancer classification using support vector machines. Machine learning46, 389-422.
Hammam, A. A., & Mohamed, E. S. (2020). Mapping soil salinity in the East Nile Delta using several methodological approaches of salinity assessment. The Egyptian Journal of Remote Sensing and Space Science23(2), 125-131.
Hassan, R., Ahmed, Z., Islam, M. T., Alam, R., & Xie, Z. (2021). Soil Salinity Detection Using Salinity Indices from Landsat 8 Satellite Image at Rampal, Bangladesh. Remote Sensing in Earth Systems Sciences4, 1-12.
Heim Jr, R. R. (2002). A review of twentieth-century drought indices used in the United States. Bulletin of the American Meteorological Society83(8), 1149-1166.
Huang, S., Tang, L., Hupy, J. P., Wang, Y., & Shao, G. (2021). A commentary review on the use of normalized difference vegetation index (NDVI) in the era of popular remote sensing [Erratum: December 2021, Vol. 32 (8), p. 2719].
Hunt, G. R., JW, S., & CJ, L. (1972). Visible and near-infrared spectra of minerals and rocks: V. Halides, phosphates, arsenates, vanadates and borates.
Huete, A. R., Jackson, R. D., & Post, D. F. (1985). Spectral response of a plant canopy with different soil backgrounds. Remote sensing of environment17(1), 37-53.
Jahanbazi, L., Mirkhani, R. & Qavami, M. S. (2018, September). Possibility of salinity changes detect by using remote sensing data. The focus of the article: Pedometry and Soil Evaluation, 16th Iran Soil Science Congress, Zanjan, Iran. (In Persian)
Jahanbazi, L., Heidari, A., Mohammadi, M. H., & Kuniushkova, M. (2023). Salt accumulation in soils under furrow and drip irrigation using modified waters in Central Iran. Eurasian Journal of Soil Science, 12(1), 63-78.
Kilic, O. M., Budak, M., Gunal, E., Acır, N., Halbac-Cotoara-Zamfir, R., Alfarraj, S., & Ansari, M. J. (2022). Soil salinity assessment of a natural pasture using remote sensing techniques in central Anatolia, Turkey. Plos one, 17(4), e0266915.
Kuhn, M., & Johnson, K. (2019). Feature engineering and selection: A practical approach for predictive models. Chapman and Hall/CRC.
Lv, Z. Z., Liu, G. M., Yang, J. S., Zhang, M. M., He, L. D., Shao, H. B., & Yu, S. P. (2013). Spatial variability of soil salinity in Bohai Sea coastal wetlands, China: Partition into four management zones. Plant Biosystems-An International Journal Dealing with all Aspects of Plant Biology, 147(4), 1201-1210.
McBratney, A. B., Santos, M. M., & Minasny, B. (2003). On digital soil mapping. Geoderma, 117(1-2), 3-52.
Metternicht, G. I., & Zinck, J. A. (2003). Remote sensing of soil salinity: potentials and constraints. Remote sensing of Environment, 85(1), 1-20.
Mohammadifar, A., Gholami, H., Golzari, S., & Collins, A. L. (2021). Spatial modelling of soil salinity: deep or shallow learning models. Environmental Science and Pollution Research, 28, 39432-39450.
Nielsen, D. R. (1985). Soil spatial variability. In Soil Spatial Variability. Proc. Workshop, 1985 (pp. 1-2). ISSS and SSSA.
Peng, J., Biswas, A., Jiang, Q., Zhao, R., Hu, J., Hu, B., & Shi, Z. (2019). Estimating soil salinity from remote sensing and terrain data in southern Xinjiang Province, China. Geoderma, 337, 1309-1319.
Saha, S. K. (2012). Microwave remote sensing in soil quality assessment. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 38, 34-39.
Scudiero, E., Corwin, D. L., & Skaggs, T. H. (2015, November). Regional-Scale Soil Salinity Assessment Using Landsat ETM+. In ASA, CSSA and SSSA International Annual Meetings (2015). ASA-CSSA-SSSA.
Singh, G., Bundela, D. S., Sethi, M., Lal, K., & Kamra, S. K. (2010). Remote Sensing and Geographic Information System for Appraisal of Salt‐Affected Soils in India. Journal of environmental quality, 39(1), 5-15.
Singh, A. (2022). Soil salinity: A global threat to sustainable development. Soil Use and Management, 38(1), 39-67.
Suleymanov, A., Abakumov, E., Suleymanov, R., Gabbasova, I., & Komissarov, M. (2021). The soil nutrient digital mapping for precision agriculture cases in the trans-ural steppe zone of Russia using topographic attributes. ISPRS International Journal of Geo-Information, 10(4), 243.
Staff, U. S. L. (1954). Diagnosis and improvement of saline and alkali soils. Agriculture handbook, 60, 83-100.
Vermeulen, D., & Van Niekerk, A. (2017). Machine learning performance for predicting soil salinity using different combinations of geomorphometric covariates. Geoderma, 299, 1-12.
Wang, Z., Zhao, G., Gao, M., & Chang, C. (2017). Spatial variability of soil salinity in coastal saline soil at different scales in the Yellow River Delta, China. Environmental Monitoring and Assessment, 189, 1-12.
Wang, J., Ding, J., Yu, D., Teng, D., He, B., Chen, X., ... & Su, F. (2020). Machine learning-based detection of soil salinity in an arid desert region, Northwest China: A comparison between Landsat-8 OLI and Sentinel-2 MSI. Science of the Total Environment, 707, 136092.
Wang, J., Peng, J., Li, H., Yin, C., Liu, W., Wang, T., & Zhang, H. (2021). Soil salinity mapping using machine learning algorithms with the Sentinel-2 MSI in arid areas, China. Remote Sensing, 13(2), 305.Yang, L., Huang, C., Liu, G., Liu, J., & Zhu, A. X. (2015). Mapping soil salinity using a similarity-based prediction approach: a case study in Huanghe River Delta, China. Chinese Geographical Science, 25, 283-294.
Willmott, C. J. (1982). Some comments on the evaluation of model performance. Bulletin of the American Meteorological Society, 63(11), 1309-1313.
ZHANG, W. T., Hong-Qi, W. U., Hai-Bin, G. U., Guang-Long, F. E. N. G., Ze, W. A. N. G., & SHENG, J. D. (2014). Variability of soil salinity at multiple spatio-temporal scales and the related driving factors in the oasis areas of Xinjiang, China. Pedosphere, 24(6), 753-762.