Monitoring vegetation dynamics using vegetation optical depth retrieved from L-band single-incidence angle backscatter observations: A field-based study over corn

Abstract

Vegetation water content (VWC) is an important parameter for sustainable land and water management. In agriculture, VWC can be used to monitor drought and assess crop productivity. Being able to monitor VWC is important to reduce agricultural vulnerability and scientifically manage agricultural water use. Due to the dense in-situ networks are expensive and have difficulties in capturing the large spatial variability of VWC. Therefore, remote sensing has great potential in VWC monitoring. With the application of remote sensing, worldwide data with coarse or fine spatial and temporal resolution can be extracted with relatively low cost. Vegetation Optical Depth (VOD), extracted from the radar remote sensing observations, is a dimensionless parameter that highly related to VWC. Thus, VOD can be used as an indicator for VWC. The present research aims to thoroughly analyze the relation between VOD and VWC. High temporal backscatter data and detailed field experiment data of soil moisture and vegetation water content during a full growing season of corn (between 18 April 2018 to 13 June 2018) were used. Correlations between VOD and VWC were analyzed. The result shows that VOD is highly related to $VWC_{bulk}$. However, the linear relation between VWC$_{bulk}$ and VOD is only valid before the heading stage at both co- and cross-polarization. Then, random forest machine learning was conducted to determine the sensitivity of VOD to the water content of different parts of the plant. This sensitivity analysis contains two parts: a) the sensitivity of VOD to the water content of different vegetation components and b) The sensitivity of VOD to stem and leaf water content at different heights. The results of a) show that VOD is more sensitive to stem and leaf water content in the vegetative stage whereas more sensitive to ear water content during the reproductive stage. Besides, stem, leaf and ear water content can better capture the VOD variation during the vegetative stage. The result from b) suggests that VOD can provide information about the vertical distribution of moisture inside the canopy. Finally, a cross-comparison was conducted between VOD and other commonly used vegetation indicators, which includes NDVI and Cross-Ratio. VOD is available regardless of cloud conditions and is, therefore, more reliable than NDVI. Compared with cross-ratio, VOD is better related to vegetation moisture dynamics.