Assessing recall bias and measurement error in high-frequency social data collection for human-environment research