Objective With the shift of urban development mode from “incremental expansion” to “stock renewal” in China, the renovation of urban historic districts has become the main work of urban planning and landscape architecture. Urban historic districts are primarily composed of high-density buildings. However, due to the hybridity of architectural styles and puzzling characters of historic districts, the past renovation usually uses a certain traditional architectural style to regularize all spaces, which destroys the rich historic connotation and potential of tourism development of urban historic districts. Accurately extracting the architectural characters and delineating the zoning of historic landscape characters of urban historic districts is an important basis for subsequent planning and design.

Methods With historical layering as a clue, this research proposes a technical system for the renovation of high-density urban historic districts, which combines landscape gene theory, historic landscape characterisation (HLC) and urban historic landscape (HUL). Based on this, this research optimizes the landscape gene theory for the buildings of urban historic districts by: 1) Strengthening the historic layering analysis of architectural gene; 2) adjusting the identification principles of architectural genes by not only adding a new principle of geographical representation, but also justifying the weight between principles; 3) implementing the binary identification of both architectural genes and recessive genes; 4) establishing a general classification system for architectural genes and recessive genes, with the new classification system involving three levels, of which the architectural genes and the recessive genes are respectively divided into six and seven subitems; 5) strengthening the correlation analysis of the binary genes. By this way, recessive genes, which do not have spatial attribution, can be mapped into space. Then the research utilizes Kmeans clustering algorithm to optimize the zoning method for historic landscape characters. Finally, the research takes the Pun Tong Historic District in Guangzhou as an example to test the effectiveness of the optimized method mentioned above.

Results The research shows that the the Pun Tong Historic District consists of four historic layering, namely the traditional rural period (from Tang Dynasty to 1840), the traditional township period (1840 − 1911), the modern township period (1912 − 1948), and the contemporary township period (from 1949 to the present). The Pun Tong Historic District is identified with a total of 38 architectural genes and 23 recessive genes. The research establishes four incidence relations for the Pun Tong Historic District: 1) The incidence relation between architectural genes and architectural styles; 2) the incidence relation between architectural genes and historic layering; 3) the incidence relation between recessive genes and historic layering; and 4) the incidence relation between architectural genes and recessive genes. Through Kmeans clustering algorithm and manual verification, the Pun Tong Historic District is divided into 5 historic landscape character sub-zones (Zones A − E) in combination with the types and historic layering tendency values of local buildings. Zone A, in which the Renwei Temple is located, has the most prominent characters of the traditional rural period and the township period; Zone B, which is located on the west side of the research area, has the most mixed historic landscape characters from all the aforesaid periods; Zone C, which is located on the northwest side to the Renwei Temple, has the most prominent characters from the period around 1980s; Area D, which is located on the north side of the research area, is dominated by the characters after 2000s (later contemporary township period); and Zone E, which is on the east side of the research area, mainly shows the characters before 1980s (early contemporary township period).

Conclusion The optimized method proposed in this research can effectively extract the architectural characters of urban historic districts, and deduce the distribution of historic layering tendency values and recessive gene tendency values according to the distribution of architectural genes, so as to realize the zoning of historic landscape characters of high-density urban historic districts. This method provides a new cognitive perspective and an improved technical method for renovation of urban historic districts, and can further be combined with HUL method to contribute to the preservation and development of urban historic districts. The practical application of this method in Pun Tong Historic District has tested the operability of the method. In the future, the method can be extended to the general urban historic districts by superimposing the distribution of natural elements and other faceted elements to achieve a more comprehensive landscape character assessment.