Ridesourcing services are rapidly occupying a large share of the traditional taxi industry and understanding the zone-level mode choice difference between them is beneficial for forming zonal management policies. Studies on this topic are based on preference survey data using questionnaires, which may bias results from reality. This paper used ridesourcing and taxi real trip data, as well as attributes of taxi zone aggregated from census tract data to analyze spatio-temporal patterns of travel mode choice differences between ridesourcing and taxi in New York City. The results show that household income, employment income, and the number of mixed residential and commercial buildings are positively associated with ridesourcing share over taxi. The random-effects show customers tend to choose taxis in Manhattan but ridesourcing in Brooklyn. Periods comparison shows that taxi zones with more industrial and manufacturing buildings have more taxi demand only in the early morning. These findings will provide new insights to understand the zone-level mode choice difference between ridesourcing and taxi services.