Researchers have developed a novel three-step hybrid variable selection strategy, SiPLS-RF(VIM)-IMIV, to enhance the accuracy and efficiency of soil potassium measurement using near-infrared spectroscopy, offering significant advancements for precision agriculture and real-time soil monitoring.
For the agriculture industry to thrive, the quality of the soil is of paramount importance. The nutrients present in the soil directly impact the quality of the crop being planted, as well as the crop yield when harvesting season comes around. Several nutrients in soil are integral for plant development, including potassium. Soil potassium influences rational fertilization plans and optimizing crop yields (1,2).
Recently, a team of researchers from China explored how farmers and scientists can measure soil potassium more effectively, with the aim being to help empower them to optimize their crop yields. Led by Huazhou Chen of Guilin University of Technology, the research team introduced a novel hybrid variable selection strategy to improve the effectiveness and accuracy of soil potassium content measurement. The research was published in Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy.
Lush green plant thriving in nutrient-rich soil, growth, vibrant, gardening, plant, healthy. Generated with AI. | Image Credit: © Andaman - stock.adobe.com
Potassium is a macronutrient, which means that plants consume a significant amount of it during their lifecycle (2). As a result, potassium plays an important role in plant growth. Potassium is associated with the movement and nutrients and water throughout the plant, making sure that plant tissues receive the nutrients and carbohydrates necessary to function, grow, and thrive (2). Enzyme activation also benefits from potassium, because it affects adenosine triphosphate production (ATP), which is important because ATP provides energy to the plant’s cells (2,3).
Although near-infrared (NIR) spectroscopy has emerged as a potent tool for real-time monitoring of soil potassium, it is often hindered by the "curse of dimensionality," where increased data complexity undermines prediction accuracy (1). To address this challenge, Chen's team developed a three-step progressive hybrid variable selection strategy that integrates multiple high-performance methods, termed synergy interval partial least squares–random forest variable importance measurement–improved mean impact value SiPLS-RF(VIM)-IMIV (1).
The researchers’ method involved the use of synergy interval partial least squares (SiPLS), random forest variable importance measurement (RF(VIM)), and the improved mean impact value (IMIV) algorithm. By leveraging these methods sequentially, the researchers crafted a fusion framework to identify essential variables while eliminating redundancy (1). Each of these played a critical role in the method. For example, SiPLS helped to identify key spectral intervals. This saved the researchers time, and it allowed them to focus on the relevant wavebands (1). RF(VIM) refined the selection process, identifying discrete variables within a threshold range of importance (1). And finally, the IMIV algorithm calculated the impact degree of variables, focusing on minimizing the root mean square error of cross-validation (RMSECV) (1). Using the IMIV algorithm helped determine which variables should be studied, which saved the research team valuable time. Instead of looking at the full spectrum, the researchers were able to discern through IMIV that only 296 variables needed to be studied, which was approximately 19.58% of the original full spectrum (1).
The hybrid strategy was embedded into a partial least squares (PLS) model to validate its effectiveness. Experimental results demonstrated dramatic improvements in prediction accuracy and model simplicity. The PLS model using SiPLS-RF(VIM)-IMIV achieved a root mean square error of testing (RMSET) of 0.01181% and a determination coefficient (RT) of 0.88246 (1). In comparison, the full spectrum PLS model had an RMSET of 0.01520% and an RT of 0.84652 (1).
By providing accurate, rapid, and non-destructive analysis of soil potassium content, the SiPLS-RF(VIM)-IMIV strategy offers actionable insights for sustainable farming practices. Beyond soil potassium analysis, the proposed method shows the value of applying advanced variable selection techniques. From monitoring soil quality to analyzing complex biochemical data, the SiPLS-RF(VIM)-IMIV strategy paves the way for innovative applications of NIR spectroscopy (1).
This study highlights the ongoing trend of applying novel analytical methods in the agriculture industry (1). With its potential for broader applicability, the hybrid strategy the research team proposed here can lead to better crop management and resource optimization.
New Spectroscopic Study Examines Hazelnut Authentication Methods
March 31st 2025A new study published in Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy demonstrates that near infrared (NIR) spectroscopy is a highly accurate and reliable method for authenticating hazelnut cultivars and geographical origins.
Exoplanet Discovery Using Spectroscopy
March 26th 2025Recent advancements in exoplanet detection, including high-resolution spectroscopy, adaptive optics, and artificial intelligence (AI)-driven data analysis, are significantly improving our ability to identify and study distant planets. These developments mark a turning point in the search for habitable worlds beyond our solar system.
Using Spectroscopy to Reveal the Secrets of Space
March 25th 2025Scientists are using advanced spectroscopic techniques to probe the universe, uncovering vital insights about celestial objects. A new study by Diriba Gonfa Tolasa of Assosa University, Ethiopia, highlights how atomic and molecular physics contribute to astrophysical discoveries, shaping our understanding of stars, galaxies, and even the possibility of extraterrestrial life.