This paper presents an improved Detrended Fluctuation Analysis (DFA) method for detecting long-range correlations in time series. DFA is a well-established technique that helps identify long-range (auto-) correlations in data with non-stationarities. However, deviations from scaling at small time scales can affect the accuracy of DFA results, especially for short records. The authors propose a modified DFA method to remove these deviations, improving the reliability of correlation detection. The paper discusses the limitations of traditional DFA, particularly in distinguishing trends from intrinsic fluctuations. Trends, caused by external factors, can distort the detection of long-range correlations if not properly removed. The modified DFA method systematically eliminates different orders of trends, allowing for a more accurate analysis of the data's scaling behavior. The authors describe how crossovers in the correlation behavior can be reliably detected and quantitatively determined. They show that different types of trends affect the results of various DFA orders, and that the modified DFA method can handle these effects more effectively. The paper also demonstrates that the modified DFA method significantly reduces deviations from scaling at small time scales, making it more suitable for short records and data with non-stationarities. The method is particularly useful for identifying long-range correlations in data with broad distributions, where traditional DFA may fail. The study includes numerical examples showing the effectiveness of the modified DFA method in detecting long-range correlations and distinguishing them from trends. The results indicate that the modified DFA provides more accurate and reliable information about the scaling behavior of the data, especially for short records and data with complex trends. The method is also robust against oscillatory trends, which can distort the analysis of time series data.This paper presents an improved Detrended Fluctuation Analysis (DFA) method for detecting long-range correlations in time series. DFA is a well-established technique that helps identify long-range (auto-) correlations in data with non-stationarities. However, deviations from scaling at small time scales can affect the accuracy of DFA results, especially for short records. The authors propose a modified DFA method to remove these deviations, improving the reliability of correlation detection. The paper discusses the limitations of traditional DFA, particularly in distinguishing trends from intrinsic fluctuations. Trends, caused by external factors, can distort the detection of long-range correlations if not properly removed. The modified DFA method systematically eliminates different orders of trends, allowing for a more accurate analysis of the data's scaling behavior. The authors describe how crossovers in the correlation behavior can be reliably detected and quantitatively determined. They show that different types of trends affect the results of various DFA orders, and that the modified DFA method can handle these effects more effectively. The paper also demonstrates that the modified DFA method significantly reduces deviations from scaling at small time scales, making it more suitable for short records and data with non-stationarities. The method is particularly useful for identifying long-range correlations in data with broad distributions, where traditional DFA may fail. The study includes numerical examples showing the effectiveness of the modified DFA method in detecting long-range correlations and distinguishing them from trends. The results indicate that the modified DFA provides more accurate and reliable information about the scaling behavior of the data, especially for short records and data with complex trends. The method is also robust against oscillatory trends, which can distort the analysis of time series data.