A Merkle Tree is a data structure designed to build a hash of many separate pieces of data and allowing efficient recalculation of the hash when items are added, removed, or changed.
The transactions in a Block are built into a Merkle Tree.
The root of this tree, or Merkle Root, is a hash representing a compound, indirect hash of all of the items added to the Merkle Tree (i.e. transactions) and is included in the Block Header, ensuring the block hash is dependent (indirectly) on all of the transactions that are included in the block.
Using a Merkle Tree to manage hashing the transactions in a block enables efficient addition of newly submitted Transactions to a block being mined.
Rather than re-hashing the entirety of the data of the Transactions to be included in the block, the new Transaction can be hashed and then a small number of hashes can be recalculated, leading up to the Merkle Root.
For large blocks this can heavily reduce the amount of data to be (re-)hashed.
Using Merkle Trees also allows for transferring only a subset of a block while still providing confidence that the data provided in fact appeared in the block in question.
For more information, see Merkle Block.