干旱区生态环境知识资源中心(Arid): Multi-task convolutional neural networks outperformed random forest for mapping soil particle size fractions in central Iran

Arid

DOI	10.1016/j.geoderma.2020.114552
	Multi-task convolutional neural networks outperformed random forest for mapping soil particle size fractions in central Iran
	Taghizadeh-Mehrjardi, R.; Mahdianpari, M.; Mohammadimanesh, F.; Behrens, T.; Toomanian, N.; Scholten, T.; Schmidt, K.
通讯作者	Taghizadeh-Mehrjardi, R
来源期刊	GEODERMA
ISSN	0016-7061
EISSN	1872-6259
出版年	2020
卷号	376
英文摘要	Knowledge about the spatial distribution of soil particle size fractions (PSF) is critical for sustainable management and resource assessment of the agricultural regions. Although conventional machine learning algorithms, such as random forest (RF) or support vector machine, have been extensively used in digital soil mapping to predict the PSF, less research examined the potential of state-of-the-art deep learning approaches for such processing. Importantly, deep learning approaches such as convolutional neural networks (CNNs) are able to incorporate contextual information about the landscape, which is of great use for DSM analysis. Accordingly, this study addresses this much-needed investigation by using a patch-based, multi-task CNN for predicting PSF of clay, sand, and silt contents at six standard layers given as soil depth increments as recommended by the GlobalSoilMap.net (i.e., 0-5, 5-15, 15-30, 30-60, 60-100, 100-200 cm). The depth functions were derived from equal-area smoothing splines in a region covering large parts (similar to 140,000 km(2)) of central Iran. The robustness of the proposed architecture is evaluated against RF. Additionally, to allow a fairer comparison between RF and CNN models, we used simple smoothing (mean) filters to effectively reproduce the auxiliary data which are then fed in the RF (RF). To evaluate the three models, we established a training (75%) and test set (25%). According to the test set, for all soil depths and all PSFs, the results demonstrate that CNN consistently outperforms RF and RF in terms of root mean square error (RMSE) and coefficient of determination (R-2). At the top layer, for example, CNN decreased the RMSE values for clay, sand, and silt contents compared to the RF (22.4%, 18.9%, and 10.7%) and RF* (18.0%, 7.4%, and 9.6%). These findings indicate that even the use of feature-engineered auxiliary data did not enable the RF* models to reach the performance of CNN. The resulting maps can be used as valuable baseline soil information for the effective management of agricultural and environmental resources in the study area and beyond.
英文关键词	Soil particle size fractions Digital soil mapping Convolutional neural network Patch-based Multi-task
类型	Article
语种	英语
收录类别	SCI-E
WOS记录号	WOS:000558138300003
WOS关键词	SPATIAL VARIABILITY ; DEPTH FUNCTIONS ; TEXTURE ; RESOLUTION ; PREDICTION ; DESERTIFICATION ; INTEGRATION ; TEMPERATE ; SELECTION ; VALLEY
WOS类目	Soil Science
WOS研究方向	Agriculture
资源类型	期刊论文
条目标识符	http://119.78.100.177/qdio/handle/2XILL650/325542
作者单位	[Taghizadeh-Mehrjardi, R.; Behrens, T.; Scholten, T.; Schmidt, K.] Univ Tubingen, Dept Geosci Soil Sci & Geomorphol, Rumelinstr 19-23, D-72070 Tubingen, Germany; [Taghizadeh-Mehrjardi, R.] Ardakan Univ, Fac Agr & Nat Resources, Ardakan, Iran; [Mahdianpari, M.; Mohammadimanesh, F.] Mem Univ Newfoundland, C Core, St John, NF A1B 3X5, Canada; [Mahdianpari, M.; Mohammadimanesh, F.] Mem Univ Newfoundland, Dept Elect Engn, St John, NF A1B 3X5, Canada; [Toomanian, N.] AREEO, Soil & Water Res Dept, Isfahan Agr & Nat Resources Res & Educ Ctr, Esfahan, Iran; [Scholten, T.] Univ Tubingen, CRC Ressource Culture 1070, Gartenstr 29, Tubingen, Germany; [Behrens, T.; Scholten, T.; Schmidt, K.] Machine Learning New Perspect Sci, Maria von Linden Str 6, Tubingen, Germany; [Behrens, T.; Scholten, T.; Schmidt, K.] Univ Tubingen, eSci Ctr, Keplerstr 2, Tubingen, Germany
推荐引用方式 GB/T 7714	Taghizadeh-Mehrjardi, R.,Mahdianpari, M.,Mohammadimanesh, F.,et al. Multi-task convolutional neural networks outperformed random forest for mapping soil particle size fractions in central Iran[J],2020,376.
APA	Taghizadeh-Mehrjardi, R..,Mahdianpari, M..,Mohammadimanesh, F..,Behrens, T..,Toomanian, N..,...&Schmidt, K..(2020).Multi-task convolutional neural networks outperformed random forest for mapping soil particle size fractions in central Iran.GEODERMA,376.
MLA	Taghizadeh-Mehrjardi, R.,et al."Multi-task convolutional neural networks outperformed random forest for mapping soil particle size fractions in central Iran".GEODERMA 376(2020).