Due to long term drought, engineered structures (e.g., dams and levees), and other stressors, river systems are at high risk of degradation. Riparian vegetation and river geomorphology are continuously changing. The change in river hydrology, geomorphology and riparian vegetation have cascading impacts on other ecological aspects of the river corridor system. In this study, spatiotemporal variations of the riparian vegetation and the river geomorphology have been characterized using machine learning techniques (in particular, random forest) over an evaluation period of three decades. The study area is the Middle Rio Grande, located in New Mexico, USA. For the study of vegetation, the normalized difference vegetation index (NDVI) was used. The land cover was classified, using Landsat images (1984 to 2020) collected from Landsat 5, 7 and 8, to determine the change in vegetation cover and river geomorphology. The trends of NDVI shows the increase in vegetation cover even during long term drought due to presence of groundwater dependent vegetation like cottonwoods. Similarly, the formation of new stable channel islands and narrowing of the channel are some major observations and changes in channel from this study. The availability of long-term datasets and machine learning algorithms in Google Earth Engine shows the potential in spatiotemporal analysis of riparian vegetation and river geomorphology. These long-term observations will help river managers to monitor the status of the riparian vegetation and the impacts on the river geomorphology.

Riparian vegetation provides many noteworthy functions in river and floodplain systems including its influence on hydrodynamic processes. Traditional methods for predicting hydrodynamic characteristics in the presence of vegetation involve the application of static roughness ( ns) values, which neglect changes in roughness due to local flow characteristics. The objectives of this study were to: (1) implement numerical routines for simulating dynamic hydraulic roughness ( nd) in a two-dimensional (2D) hydrodynamic model; (2) evaluate the performance of two dynamic roughness approaches; and (3) compare vegetation parameters and hydrodynamic model results based on field-based and remote sensing acquisition methods. A coupled vegetation-hydraulic solver was developed for a 2D hydraulics model using two dynamic approaches, which required vegetation parameters to calculate spatially distributed, dynamic roughness coefficients. Vegetation parameters were determined by field survey and using airborne LiDAR data. Water surface elevations modeled using conventional and the proposed dynamic approaches produced similar profiles. The method demonstrates the suitability in modeling the system where there is no calibration data. Substantial spatial variations in both n and hydraulic parameters were observed when comparing the static and dynamic approaches. Thus, the method proposed here is beneficial for describing the hydraulic conditions for the area having huge variation of vegetation. The proposed methods have the potential to improve our ability to simulate the spatial and temporal heterogeneity of vegetated floodplain surfaces with an approach that is more physically-based and reproducible than conventional “look up” approaches. However, additional research is needed to quantify model performance with respect to spatially distributed flow properties and parameterization of vegetation characteristics.