Knowledge‐guided machine learning for improving daily soil temperature prediction across the United States

Abstract

Data-driven models used for predicting soil temperature usually have increasing errors with increasing depth. By exploring the integration of knowledge-based and machine learning approaches, this study used a novel transformation of meteorological variables to increase prediction accuracy of soil temperature with increasing soil depth. Using datasets for two soil textures (silty clay loam and loamy coarse sand) at two locations with different climates, predictive models were developed for five depths (5, 10, 20, 50, and 100 cm) as a function of meteorological features using an adaptive neuro-fuzzy inference system (ANFIS). For each depth, soil temperature was predicted with nontransformation (NT), autocorrelation (AC), moving average (MA), and a combination of transformations (NT-ACMA) of meteorological features. Across all depths, the predictive accuracy of NT-ACMA models was significantly higher than that of NT, AC, and MA models for both soil textures (R 2 = .99, RMSE = 1˚C for silty clay loam; R 2 = .99, RMSE = 1.2˚C for loamy coarse sand), with increasing prediction accuracy as soil depth increases. Results for different soil textures and climates in 11 locations across the contiguous United States show that, except for 100-cm depth, there seems to be significant positive linear relationships between the best moving average and the solar inclination. This makes our NT-ACMA technique transferable to any location in the contiguous United States irrespective of the location, climate, and soil texture. We conclude that integrating lag times and moving averages of meteorological features can lead to better prediction of soil temperature at different soil depths.

A Comprehensive Study of Artificial Intelligence Applications for Soil Temperature Prediction

“…Underfitting refers to a model that can neither model the training data nor the testing data. The sweet spot between underfitting and overfitting, which shows the good performance of a machine learning algorithm on both training and testing data, is a good fit [5,27].…”

“…Adaptive Neuro-Fuzzy Inference System (ANFIS) models combine fuzzy systems and the learning ability of neural networks. ANFIS is considered an ANN model doing the preprocessing step by converting numeric values into fuzzy values [5,26,[30][31][32].…”

“…So, it is essential to measure or estimate this parameter with a reasonable precision. Therefore, an accurate and cost-effective model which can predict soil temperature accurately is extensively needed [4][5][6].…”

Low-Level Groundwater Atrazine in High Atrazine Usage Nebraska Counties: Likely Effects of Excessive Groundwater Abstraction

“…Although soil temperature would be more critical to the kinetics of atrazine in groundwater than the air temperature, the lack of direct measurements of soil temperature resulted in the use of annual mean air temperature as a proxy for the soil temperature at depths where groundwater wells would be screened. This is valid because there is a strong relationship between mean air temperature and soil temperature due to the exchange processes between them [ 17 , 18 ]. Groundwater temperature is usually equal to the annual mean air temperature above the ground, and it generally fluctuates narrowly (based on depth) around this mean temperature year round.…”

A Novel Machine Learning Approach: Soil Temperature Ordinal Classification (STOC)

“…More specifically, this study brings together notions from the fields of agriculture and machine learning for information fusion. In the regression studies, DT (Sattari et al 2020;Sanikhani et al 2018), Support Vector Regression (SVR) (Li et al 2020a;Li et al 2020b;Shamshirband et al 2020;Delbari et al 2019;Mehdizadeh et al 2018;Xing et al 2018), RF (Alizamir et al 2020b;Tsai et al 2020;Feng et al 2019), NN (Abimbola et al 2021;Bayatvarkeshi et al 2021;Wang et al 2021;Hao et al 2020;Penghui et al 2020;Citakoglu 2017;Abyaneh et al 2016;Kisi et al 2015), ELM (Alizamir et al 2020a) algorithms have been preferred for predicting soil temperatures. In addition, some of the time-series studies have also applied NN (Li et al 2020c;Bonakdari et al 2019), ELM (Zeynoddin et al 2020;Mehdizadeh et al 2020), SVR (Nanda et al 2020) algorithms for the prediction performance comparison.…”