Objectives The leaf chlorophyll content (LCC) is the most used index for timely, rapid and non-destructive estimation of crop growth and yield. The purpose of this study was to establish the best estimation model for LCC spectral monitoring, and to provide technical support for rapid and non-destructive monitoring of potato growth information, so as to achieve accurate management of potato production.

Methods From 2023 to 2024, field experiments on potato nitrogen application rates were conducted using five varieties across six locations. Hyperspectral reflectance data and SPAD values of potato leaves were measured during both pre-flowering and post-flowering stages, while corresponding chlorophyll content was determined using spectrophotometry. Based on two chlorophyll-sensitive spectral index algorithms—the selected double difference index (DD) and the triangular chlorophyll index (TCI)–optimized spectral indices were established using a band optimization algorithm to analyze the correlation between measured chlorophyll content and the spectral indices as well as SPAD values. Subsequently, 108 functions from 15 continuous wavelet families and 126 functions from 7 discrete wavelet families were calculated to screen the optimal wavelet function and wavelet features. The optimal wavelet function was applied to estimate leaf chlorophyll content (LCC) from extensive field experimental data, and the model was evaluated using independent experiments.

Results The DD index exhibited the strongest correlation with LCC (R²=0.62), followed by TCI (R²=0.58), while SPAD showed the lowest correlation (R²=0.50). This indicates that the spectral indices correlate better with LCC than SPAD does, and that DD outperforms TCI. The continuous wavelet transform (CWT) demonstrated the optimal performance at the 4th decomposition scale, with the LCC estimation model producing R² values ranging from 0.2 to 0.7, whereas the discrete wavelet transform (DWT) performed best at the 1st scale, yielding LCC models with R² values concentrated around 0.48, indicating that CWT provides superior accuracy in estimating chlorophyll content in potato leaves compared to DWT. Among the CWTs, the Symlet family demonstrated the best performance, with Sym15 achieving the highest R² value of 0.61. The spectrum of wavelets strongly correlated with LCC were predominantly concentrated within 450–900 nm, and the spectrum 629 nm, 630 nm, and 710 nm showed the most significant wavelet features at the fourth scale. The optimal wavelet is Sym15 (W₆₃₀, S₄), which was located in the 4th scale and the red light region, and the independent verification results of the model showed that compared with the model based on the double-difference index and SPAD, the model based on the wavelet feature sym15 (R²=0.73–0.80, RMSE = 0.22–0.32 mg/g) performed best in LCC prediction.

Conclusions This study identified sym15 (W₆₃₀, S₄) as the optimal wavelet function and feature for constructing a potato LCC estimation model, providing an effective method for the accurate estimation of potato leaf chlorophyll content.