Preallocation. Mongo sets aside disk-space in empty containers, so when the time comes to write something to disk, it doesn't have to shuffle bits out of the way first. It does so by a doubling algorithm, always doubling the amount of disk space preallocated until it reaches 2GB; and then each prealloc file from thereon is 2GB. Once data is preallocated, it doesn't unallocate unless you specifically tell it to. So observable MongoDB space usage tends to go up automatically, but not down.
Some research on the Mongo preallocation...
reducing-mongodb-database-file-size
mongo-prealloc-files-taking-up-room
// compact the database from within the Mongo shell
db.runCommand( { compact : 'mycollectionname' } )
// repair the database from the command line
mongod --config /usr/local/etc/mongod.conf --repair --repairpath /Volumes/X/mongo_repair --nojournal
// or dump and re-import from the command line
mongodump -d databasename
echo 'db.dropDatabase()' | mongo databasename
mongorestore dump/databasename