The manuscript addresses a relevant topic in the wind industry, providing a novel methodology to assess the impact of blade aerodynamic modifications on turbine performance, using SCADA and O&M data. The paper highlights the complexity of the study and the challenges encountered in completing the initial goal of the study.

The paper is well structured and in general clear - even though some details can be improved to facilitate the readability of the paper (see my comments attached). Otherwise, very good job!

The manuscript presents an analysis of the effect of software updates and various maintenance activities on turbine performance. An interesting and novel approach is proposed for measuring turbine performance in time in the form of a virtual sensor signal called TPI, and the effect of software updates and different maintenance activities is analysed in terms of the gradient of TPI in a given time range, indicating a positive or negative effect in turbine performance.

The long-term trend in TPI reveals that the software updates during the first ~5 years of the data lead to mostly positive effects in turbine performance, with mixed trends among turbines for the later section of data with no more software updates. However, the results regarding the maintenance activities are relatively inconclusive, with a large variance in the TPI gradient in most of the maintenance activities. The authors present an in-depth explanation of the uncertainties involved, which limit the practicality of the analysis given the available data.

Despite the limited success in achieving the defined goals, the manuscript presents a novel and interesting approach for monitoring turbine performance through time, which can prove valuable in similar future studies. Furthermore, it provides a good and detailed explanation of the limitations of the method and the sources of uncertainty involved, which can also prove valuable for fellow researchers.

In a nutshell, I think the manuscript is a decent contribution and is suitable for publication in the journal, with some minor revisions listed below:

• There are numerous typos and grammatical errors that need to be fixed.
• Citations are mixed with the body of text, leading to misunderstanding. I suggest wrapping them in brackets.
• Line 188: The range between 20% and 37% seems quite specific. How was it defined? Visually or according to a max allowed change in pitch?
• Section 2.8 General comment: Brief explanations of the methods implemented, the context, and the reasons for the analyses are missing.
• Line 248 - "Normality Assessment": Ambiguous. More context is needed in terms of "what" and "why".
• Line 389: How is the time window selected for the calculation of the gradient?