This paper proposes an improved localization that integrates LiDAR-SLAM, GNSS, and odometry in environments with sparse features such as rural areas with only a few surrounding objects and ample features such as urban street canyons. Localization is important for autonomous driving of automobiles, and SLAM is useful in environments without a map; however, it sometimes collapses in environments with few feature observations. Fusion of GNSS and odometry will complement the lack of observed features, but deciding when to switch between them can be difficult. This research proposes using a probabilistic data association filter to fuse the LiDAR-SLAM, GNSS, and odometry methods. This filter exploits probabilistic weighting for each measurement by calculating the uncertainty of observations. This study evaluated the performance of the method via numerical verifications in challenging environments, including urban and rural areas, modeled on a realistic simulator.