MSc

Kerala, characterized by diverse physiography, landscapes, and soil compositions, confronts challenges related to soil quality, primarily due to unpredictable shifts in weather conditions. Given the dynamic nature of the soil, which adapts to evolving environmental factors, assessing soil quality becomes crucial. Traditional soil analysis methods are both time-consuming and expensive, necessitating the exploration of innovative alternatives to expedite the analysis process. Visible-NIR (Vis-NIR) reflectance spectroscopy has emerged as a rapid, cost-effective, environmentally friendly, non-destructive, reproducible, and repeatable analytical technique. Among various soil quality parameters, Soil Organic Carbon (SOC) plays a significant role in influencing the physicochemical and biological aspects of soil, thereby positively affecting crop growth. Determining SOC is particularly crucial in tropical regions due to its dynamic nature and its contribution to climate change adaptation through carbon sequestration. Against this backdrop, a research study was conducted between 2021-2023 with the aim of estimating and quantitatively predicting soil organic carbon in upland and wetland ecosystems of Kerala using hyperspectral reflectance spectroscopy in the Department of Soil Science & Agricultural Chemistry.
The study was conducted under the different land use systems of the three Agro Ecological Units (AEU) of Kerala namely Kole lands (AEU 6), North central Laterite (AEU 10) and Palakkad central plain (AEU 22). Georeferenced surface soil samples (0–15cm) were collected from 70 locations under each AEU and anlyzed for physico-chemical properties of soil including SOC. The results of physico-chemical analysis showed that, mean bulk density was lower in AEU 6 (1.12 Mg m-3) compared to AEU 10 and 22 (1.3 Mg m-3). SOC ranged from (0.6 to 4.26 %) in AEU 6, AEU 10 (0.53 to 6.42%) and AEU 22 (0.21 to 3.74 %). Total Carbon (Tot. C) in AEU 6 varied from (0.7 to 4.46 %), AEU 10 (0.53 to 6.42) and AEU 22(0.34 to 3.83 %). Available Nitrogen (Av. N) ranged from (125.4 to 363.2 kg ha-1) in AEU 6, AEU 10 (108.9 to 439.04 kg ha-1) and AEU 22 (75.2 to 489.2 kg ha-1). Total Nitrogen (Tot. N) values ranging from AEU 6 (0.04 to 0.36 %), AEU 10 (0.07 to 0.87 %) and AEU 22 (0.05 to 0.33 %).
Reflectance spectra were collected in controlled dark room condition for 210 samples in the wavelength region (350-2500 nm) using the ASD FieldSpec 4 spectroradiometer. The spectra were then pre-processed by using different techniques like First order derivative (FOD), second order derivative (SOD), log (1/R) transformation, and continuum removal (CR) for enhancing quality of the spectra for better prediction. Further, correlation analysis was carried for both raw and preprocessed spectra and correlated spectral bands were taken for developing PLSR (Partial least square Regression) model, the coefficient of determination (R2) was better for first and second order derivatives for SOC compared with raw and other preprocessed data. Hence, FOD was taken for developing PLSR model. The model was developed for the prediction of soil organic carbon (SOC), total carbon (Tot. C), available nitrogen (Av. N), and total nitrogen (Tot. N). For all the four parameters the datasets were divided in the ratio of 70:30, with 70% of the data (n=147) used for calibration and the remaining 30% used for validation (n=63).
The PLSR model developed in this study underwent validation based on R2 and RMSE, demonstrating its efficacy in predicting SOC, Tot. C, Av. N and Tot. N values. The model exhibited strong predictive capabilities, achieving R2 values of 0.88 for SOC, 0.86 for Tot. C, 0.85 for Av. N, and 0.87 for Tot. N, with corresponding RMSE values of 0.66 %, 0.73 %, 66.9 kg ha-1, and 0.06%, respectively. The high variability in carbon and nitrogen content within the soil samples underscored the model's effectiveness.
This study concludes that hyperspectral reflectance spectroscopy is a successful approach for predicting carbon and nitrogen levels in different ecosystems in Kerala. The research findings emphasize the significance of employing Chemometrics as an advanced tool for soil property prediction. Looking ahead, there is a pressing need to develop spectral libraries and prediction models tailored to regional and field-level variations in soil properties across diverse soil types in Kerala.