PG Thesis
Permanent URI for this collectionhttp://localhost:4000/handle/123456789/2
Browse
Item Soil carbon stock estimation and prediction of aggregate associated carbon under different land use systems in Palakkad central plain using reflectance spectroscopy(Department of Soil Science and Agricultural Chemistry, College of Agriculture , Vellanikkara, 2025-01-09) Abdul Hadi, K; Divya Vijayan, VThe distribution of soil organic carbon (SOC) across various aggregates and its changes over time is crucial for monitoring carbon dynamics and optimising nutrient management. Reflectance spectroscopy is a fast, non-destructive and economical solution for estimating SOC in the soil. In this context, the present study was proposed to identify the potential of different land-use systems for the sequestration of SOC and distribution of SOC among various aggregates of selected land-use systems (LUS), where the selected aggregates were macroaggregate (> 0.25 mm), micro aggregate (0.25-0.053 mm) and clay and silt fraction (< 0.053 mm). It also aimed to evaluate the predictive potential of reflectance spectra for estimating SOC associated with soil aggregates. The study was conducted in the Department of Soil Science and Agricultural Chemistry, College of Agriculture, Vellanikkara during 2023-24. The selected land use systems were natural forest, coconut plantation, rubber plantation, vegetable field, and paddy field, in the Palakkad Central Plain (AEU 22). Georeferenced surface (sample no. = 9), subsurface soil samples (sample no. = 9) and one profile sample were collected from each land use system. A total of 105 samples were collected. The soil samples were pre-processed and analysed for the different physico-chemical properties, wet aggregate analysis, soil carbon stock (SCS) and SOC in aggregate fraction. A part of the macroaggregates and microaggregates obtained from wet sieving was utilised to record the spectral signatures. Prediction models were then developed using the SOC content of these aggregates as the dependent variable, while the spectroscopic bands corresponding to each aggregate type served as the independent variables. The raw spectra obtained from the spectroradiometer underwent four preprocessing steps viz logarithmic transformation, Savitzky-Golay filtering (SG), first-order derivative (FOD), and second-order derivative (SOD) before model development. Correlation analyses were then performed to identify the most relevant wavebands associated with SOC. It reduced the number of bands needed for modelling. The prediction of aggregate-associated SOC was performed using a Partial Least Squares Regression (PLSR) model. The efficiency of developed models was analysed using the coefficient of determination (R2), root mean square error (RMSE), and the ratio of prediction deviation (RPD). The subset prediction models were also developed to identify which region is best for prediction. Finally, the variable importance in the projection score (VIP) was used to find which wavelengths are contributing the most of the variance in the PLSR model. The result showed that the perennial systems such as forest (47.95 Mg ha-1) and rubber (43.08 Mg ha-1) showed higher soil carbon stock. In contrast, annual cropping systems like paddy (SCS = 16.55 Mg ha-1) and vegetable (SCS = 19.14 Mg ha-1) exhibited lower soil carbon stock. The distribution of water-stable aggregates revealed that greater proportion of macroaggregates in rubber (96.05 %) and forest systems (95.18%), followed by coconut (83.12 %). Lower aggregate fractions, including microaggregates and clay-silt fractions, were more abundant in paddy (micro aggregate = 30.72 %; clay and silt fraction = 2.29 %) and vegetable fields (micro aggregate = 26.60 %; clay and silt fraction = 2.53 %). Mean weight diameter was highest in forest (2.65 mm) and rubber systems (2.44 mm) and lowest in paddy (0.73 mm) and vegetable systems (0.99 mm). These findings suggest that intensive tillage and lower SOC levels significantly degraded soil structure. The key spectral bands for predicting bulk SOC spanned the entire spectrum (400–2500 nm), whereas bands relevant to aggregate-associated SOC were primarily located in the shortwave infrared (SWIR) region (1000–2500 nm). These findings indicate that bulk SOC predictions were more influenced by the chromophore effect, while aggregate SOC predictions depended on chemical bonds in specific spectral regions. The preprocessing steps influenced the prediction ability of the model. second-order derivatives (SOD) produced the best models for the prediction of microaggregate SOC (MiSOC) and macroaggregate SOC (MaSOC), while Savitzky Golay-filtered full spectra were most effective for bulk soil SOC. Prediction accuracy was higher for MiSOC (R² = 0.84, RMSE = 0.25, RPD = 2.21) compared to MaSOC (R² = 0.92, RMSE = 0.25, RPD = 1.74). The R2 value of the test dataset further validated the prediction efficiency of the model. The R2 value of the test dataset were 0.79, 0.66, and 0.84 for MiSOC, MaSOC, and bulk soil SOC, respectively. The superior predictive performance for MiSOC can be attributed to the higher organic carbon content in microaggregates and the difference in the type of organic functional groups between the two fractions. The result shows that spectroscopy can be effectively exploited for the prediction of aggregate-associated carbon