Development of soil moisture content model using water cloud model and neural network techniques for PALSAR-2 in oil palm plantation

Water stress, the most severe environmental stress, negatively affects crop growth and development, reducing crop yield more than any other environmental component. When water stress is present, the soil moisture content is impacted by the interaction between the land surface and the atmosphere. Despite its importance, soil moisture content has yet to be widely used in modelling hydrological and biogeochemical processes because it is a complex parameter to monitor over a large area cost- effectively and routinely. Notably, return signals from synthetic aperture radar (SAR) are affected by surface properties, such as roughness, correlation length, and the dielectric constant of the soil. It is necessary to maintain soil balance and appropriate soil moisture content. Surface soil moisture estimation is commonly determined by volumetric soil moisture, where there is a linear relationship between the moisture and measured radar backscatter in decibels. However, standard soil moisture determination methods are complicated, time-consuming, and provide limited information concerning soil moisture content. The objective of this study is to employ PALSAR-2 observations in oil palm crops to evaluate the sensitivity of SAR signals by using a water cloud model (WCM) inversion to obtain soil moisture content. In this regard, fieldwork was carried out in Chuping District, Perlis, Malaysia. PALSAR-2 image acquisition was obtained over three days in 2019: January 17, April 16, and July 9. Study results revealed that using the Leaf Area Index (LAI), Leaf Water Area Index (LWAI), and Normalised Plant Water Content (NPWC) were able to quantify the impacts of vegetation on backscattering coefficients in oil palm crops. HV polarisation was more effective in reproducing backscatter coefficients than HH polarisation. The best fit was obtained by employing the LAI as a vegetation descriptor. Further, to understand PALSAR-2 capabilities in soil moisture content retrievals, SAR-derived vegetation descriptors were employed in the WCM approach. Among the indices considered was the Radar Vegetation Index (RVI), the ratio of the backscatter coefficients using polarisations of HH/HV (RHH/HV) and HV/HH (RHV/HH) to oil palm crops as vegetation descriptors in the WCM. These WCM-derived retrievals were compared to the physical soil sampling method for estimating soil moisture in the study site. The results showed that the WCM model using the LAI under HV polarisation obtained good accuracy, with the root mean square error (RMSE) measured as 0.036 m3 /m3 in the testing dataset and 0.037 m3 /m3 in the training data set. The results of field-based input, LAI, were equivalent to the RVI under HV polarisation, which exhibited accuracies of 0.035 and 0.032 m3 /m3 , respectively, in the training and testing dataset. In the HH polarisation, SAR-based input, RVI demonstrated RMSE = 0.037 m3 /m3 in the training and testing dataset. The outcome of this study suggests that the SAR-based indicator, RVI, utilised with PALSAR-2, can be applied to reduce field-related input in the retrieval of soil moisture content for WCM in oil palm crops. In addition, this research utilised the ANN approach to evaluate the potential use of machine learning in the application of PALSAR-2 images to exploit further soil moisture content determination in oil palm crops. The research findings from this study show that ANN contributes a relatively low RMSE of 0.012 m3 /m3 in the training data set and 0.010 m3 /m3 in the testing dataset. The ANN results were obtained by using HH, HV, and incident angle as input to the ANN model. The said results also contributed a considerably high R2 score for the training and testing data at 0.9551 and 0.9638, hence suggesting that the PALSAR- 2 sensor has good potential to be utilised in oil palm crops. Furthermore, soil moisture content retrieval using SAR data has significant potential as SAR can alleviate the cloud cover difficulties experienced while obtaining optical data in oil palm crops. It is possible to use SAR data to recover soil moisture since the backscatter coefficient is responsive to both soil and plants by penetrating through the vegetation layer.

Text FK 2024 81 - Full Text.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (4MB) Text FK 2024 81.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (5MB)

Actions (login required)

View Item