Landslides exhibit complex geomorphological features, which are difficult to monitor and map the changes. Photogrammetric methods have emerged as promising tools to overcome such problems due to 3D reconstruction from overlapping images without disturbing the surface. This study presents the structure from motion (SfM) technique for three-dimensional reconstruction using video and images of a large landslide which occurred on 13 August 2017 at Kotropi, Himachal Pradesh, India. In this study, we have used an Unmanned Aerial Vehicle (UAV) – DJI Phantom 3 Advanced to collect high-resolution images and video of landslide. A total of 98 images and 9 videos of 12:18 minutes duration were captured during the drone flight on 23 March 2019 covering the whole landslide and surrounding areas. The images were collected in unorganized manner for covering the whole landslide in the limited amount of time. We have used videos for coarse reconstruction and high-resolution images for fine reconstruction of landslide. The whole model was developed in two parts using MeshRoom and later merged using co-registration in CloudCompare. Based on feature detection technique, scale invariant feature transform (SIFT), image features were automatically detected, described, and matched between photos. A bundle block adjustment is performed on the matched features to identify the 3D position and orientation of the cameras, and the XYZ location of each feature in the photographs resulting in a sparse 3D point cloud. Finally, meshing is carried out using 3D Delaunay tetrahedralization. For visualization and analysis of final 3D model, open source software CloudCompare/MeshLab was used. The morphological parameters such as length, width, height, perimeter, area and volume were computed from the 3D model. From field observations and image analysis we will show that UAV-based image in combination with 3D scene reconstruction algorithms provide a flexible and effective tool to map and monitor large landslides such as Kotropi landslide of Himachal Pradesh.

Landslides cause billions of dollars in property damage and thousands of deaths every year worldwide. India has more than 15% of its land area prone to landslides, hence mapping of these areas for the presence of landslides is of utmost importance. Landslide susceptibility zonation maps give approximate information about the occurrence of landslides. There are various factors responsible for slope instability. In this work, 11 causative factors have been considered such as Aspect, Elevation, Geology, Distance from thrusts, Distance from streams, Plan curvature, Profile curvature, Slope, Stream power index, Tangential curvature, Topographic wetness index. Machine learning methods such as artificial neural network, support vector machine require a large amount of training data; however, the number of landslide occurrences are limited in a study area. The limited number of landslides leads to a small number of positive class pixels in the training data. On contrary, the number of non-landslide pixels (negative class pixels) are huge in numbers. This under-represented data and severe class distribution skew create a data imbalance for learning algorithms and sub-optimal models, which are biased towards the majority class (non-landslide pixels) and have low performance on the minority class (landslide pixels). Generally, the data is imbalanced when the class ratio is of the order of 100:1, 1000:1 and 10000:1 (i.e., one-class points are 100, 1000 or 10000 times more than that of another class points). In our work, class ratio is more than 300:1 (i.e. for each one landslide pixel, we have more than 300 non-landslide pixels). Thus, we can clearly say that our data is imbalanced. There are two major data balancing techniques, which are oversampling of a minority class and under-sampling of majority class. The minority oversampling cannot be applied, as it will create false landslide pixels. We have performed under-sampling of non-landslide pixels using various techniques. We will discuss landslide susceptibility zonation with and without using data imbalance technique and show major improvements in accuracy over imbalanced learning.

Regional crop production estimates are important in both public and private sectors to ensure the adequacy of a food supply and aid policymakers and farmers in managing harvest, storage, import/export, transportation, and anticipate market fluctuations. Food security will be progressively challenged by population growth and climate change. Thus, the prediction of accurate regional crop yield is essential for national food security and the sustainable development of the Indian agriculture sector. In this study, we have selected Punjab, the highest wheat yielding state in India. The district-wise wheat yield data were available for the year 2000 – 2019. We have used several covariates for crop health viz. normalized difference vegetation index (NDVI), leaf area index (LAI), fraction of absorbed photosynthetically active radiation (fAPAR); meteorological indicators viz. land surface temperature (LST), and evapotranspiration (ET); and surface characteristics viz. protrusion coefficient (PC). These indicators were generated at 250 m spatial resolution from the MODIS data using Google Earth Engine. The whole data was divided into two groups for training (2000 – 2009, 2011, 2013, 2014, 2016 - 2019) and testing (2010, 2012, 2015), which were randomly selected. This study uses the random forest (RF) regression method to create a wheat yield prediction model. We created several combinations of covariates and found that fAPAR and ET are highly correlated with NDVI and do not have much influence on the model’s prediction accuracy. Hence, only four out of six covariates were selected for final training. The coefficient of determination between district-level yield vs. (NDVI/LAI/PC/LST) was 0.37/0.31/0.15/0.13 respectively. We used randomized search cross-validation as well as grid search cross-validation for hyper-parameter tuning. Furthermore, we used mean absolute error (MAE) and accuracy as quality metrics. The MAE for training was 0.1870 t/Ha with 95.81% accuracy, whereas the MAE on test data was obtained as 0.4293 t/Ha with 90.02% accuracy. The results of this study are within acceptable error limits of the published research articles. Overall, this study demonstrates that covariates derived from coarse resolution satellite data can predict district-level crop yield with reasonable accuracy.