With the development of spectrum observation technology, solar-induced chlorophyll fluorescence (SIF)—an effective substitute for photosynthesis—has been widely used to monitor crop stress, vegetation phenology and ecosystem productivity. The relationship between fluorescence and photosynthesis is complicated because they are sensitive to environmental changes. Understanding the response of SIF to environmental factors is of great significance for clarifying the variation dynamic and relationship between SIF and photosynthesis under different conditions. In this study, the canopy SIF and the environmental factors of a Quercus variabilis BI. plantation were observed simultaneously, and the response of SIF to environmental factors at a daily scale and at a half-hour scale was analyzed. The results showed that SIF had obvious seasonal and diurnal dynamics and was mainly driven by photosynthetically active radiation (PAR). The influence of PAR, air temperature (Ta), vapor pressure deficit (VPD), soil moisture (SM) and wind speed (Ws) on SIF varied with the lapse of the growing season. After eliminating the covariant effect of PAR on the Ta and VPD during the whole growing season, the relationship between VPD and SIF was found to be negative, and the effect of Ta on SIF disappeared. This study enriched the ground observation dataset and provided support for understanding the variations in the relationship between SIF and photosynthesis under different conditions.

Solar-induced chlorophyll fluorescence (SIF) is a powerful tool for the estimation of gross primary productivity (GPP), but the relationship between SIF and GPP under drought stress remains incompletely understood. Elucidating the response of the relationship between SIF and GPP to drought stress is essential in order to enhance the precision of GPP estimation in forests. In this study, we obtained SIF in the red (SIF₆₈₇) and far-red (SIF₇₆₀) bands and GPP data from tower flux observations in a Chinese cork oak plantation to explore the response of the diurnal GPP-SIF relationship to drought stress. The plant water stress index (PWSI) was used to quantify drought stress. The results show that drought reduced SIF and GPP, but GPP was more sensitive to drought stress than SIF. The diurnal non-linear relationship of GPP-SIF (R²) decreased with the increase in drought stress, but a significant non-linear correlation remained for GPP-SIF (R²_GPP-SIF₇₆₀ = 0.30, R²_GPP-SIF₆₈₇ = 0.23) under severe drought stress (PWSI_bin: 0.8–0.9). Physiological coupling strengthened the GPP-SIF relationship under drought, while canopy structure effects were negligible. Random forest and path analyses revealed that VPD was the key factor reducing the GPP-SIF correlation during drought. Incorporating VPD into the GPP-SIF relationship improved the GPP estimation accuracy by over 48% under severe drought stress. The red SIF allowed for more accurate GPP estimations than the far-red SIF under drought conditions. Our results offer important perspectives on the GPP-SIF relationship under drought conditions, potentially helping to improve GPP model predictions in the face of climate change.

Monitoring drought stress is crucial for estimating productivity and assessing the health status of forest ecosystems under global climate change. Solar-induced chlorophyll fluorescence (SIF) is mechanistically coupled to photosynthesis and has advantages over greenness-based vegetation indices in detecting drought. In recent years, SIF has commonly been used in monitoring drought stress in crop ecosystems. However, the response of tower-based SIF to drought stress in forest ecosystems remains unclear. In this study, we investigated the potential of tower-based SIF to monitor drought, which was quantified using the plant water stress index (PWSI) in a Chinese cork oak plantation. The results show the negative effect of drought on SIF, and afternoon depression of SIF emission under drought stress was observed. Canopy SIF (F) exhibited a nonlinear relationship with PWSI, while the quantum yield of SIF (ΦF) exhibited a significant linear relationship with PWSI at 687 nm and 760 nm (ΦF₆₈₇: R² = 0.90; ΦF₇₆₀: R² = 0.85). Incident radiation (PAR) and canopy structure affected the response of SIF to drought stress, with PAR as the main factor causing the nonlinear relationship between F and PWSI. Afternoon depression was described as the afternoon/morning ratio (AMR). AMR_F and AMR_ΦF exhibited a negative linear response to PWSI. AMR_F was less affected than F by PAR and canopy structures, and AMR_ΦF was more physiologically representative than ΦF. Moreover, AMR_ΦF was sensitive to VPD and REW, and it might be a good indicator of drought. Red SIF was more sensitive to drought than far-red SIF, as the R² of PWSI with AMR_ΦF687 (R² = 0.89) was higher than that with AMR_ΦF687 (R² = 0.84). These results highlight the potential of tower-based SIF, especially red SIF, for drought monitoring in a plantation, and consideration of the physiological diurnal variation in SIF under drought stress is crucial for improving the accuracy of drought stress monitoring.

Evapotranspiration is a key link in the water cycle of terrestrial ecosystems, and the partitioning of evapotranspiration is a prerequisite for diagnosing vegetation growth and water use strategies. In this study, we used double-layer eddy covariance (DLEC) measurements within and above the canopy of poplar plantations to divide evapotranspiration into transpiration and evaporation during the growing season. We diagnosed the coupling state of airflows in the canopy vertical layer and found that the daytime coupling state at the half-hourly scale can mask nighttime decoupling. Furthermore, we investigated the daytime and nighttime vertical layer airflow coupling states separately and quantified the effects of coupling states on the DLEC of resolved transpiration. The partitioning results of the DLEC method were taken as the standard after the airflow coupling test. Then, the performance and accuracy of evapotranspiration partitioning for the modified relaxed eddy accumulation (MREA), the conditional eddy covariance (CEC), and the flux variance similarity (FVS) with DLEC were compared. Transpiration calculated from MREA showed the best agreement with DLEC, and the other methods showed different degrees of underestimation (1:1 slope = 0.64–0.83). Evaporation calculated from FVS showed the best agreement with DLEC, while CEC and FVS made an overestimation of more than 26% (1:1 slope = 1.26–1.99), but MREA made an underestimation from 5% to 35% (1:1 slope = 0.65–0.95). The correlation coefficients between DLEC and MREA for transpiration were 0.95–0.97 with RMSEs of 15.52–17.04 W m⁻², and those between DLEC and FVS for transpiration were 0.73–0.78 with RMSEs of 10–21.26 W m⁻² at the daily half-hourly scale. A detailed comparison of the differences between DLEC and evapotranspiration partitioning methods from high-frequency eddy covariance data under the condition of canopy vertical layer airflow mixing provides knowledge about the consistency of results for evapotranspiration partitioning in poplar plantation forests.

Vegetation transpiration (Tr) is crucial for the water cycle, regional water balance, and plant growth but remains challenging to estimate at large scales. Sun-induced chlorophyll fluorescence (SIF) provides a novel method for estimating Tr, but its effectiveness is limited by species specificity, requiring continuous tower-based observations for comprehensive analysis across diverse ecosystems. In this study, SIF and Tr were simultaneously monitored in Chinese cork oak (ring-porous), poplar (diffuse-porous), and arborvitae (non-porous) plantations in northern China, and the SIF–Tr relationship was further analyzed. The results showed that SIF and Tr shared similar diurnal dynamics, although Tr exhibited midday saturation. SIF and Tr were closely correlated, and the correlation strengthened as the temporal scale aggregated. Environmental factors had nonlinear impacts on SIF and Tr. Therefore, the SIF–Tr relationship deteriorated to some extent at midday, with short-term stress reducing the correlation by 0.1–0.23. Compared to the linear empirical model, the inclusion of environmental factors improved the accuracy of SIF-based Tr estimation, increasing the R² value by 0.12 to 0.37. At the same level of accuracy, the number of environmental variables required was higher at the half-hour scale than at the daily scale. This study demonstrated the species-specific influence of environmental factors on SIF and Tr in different plantations, enhanced the understanding of the SIF–Tr relationship, and provided theoretical and data support for future large-scale Tr predictions using satellite-based SIF.

As an important reference for assessing plant water consumption and estimating plant transpiration, it is of great significance to achieve accurate prediction of plant sap flow. A number of deep learning models were established and compared using approximately 3 years of continuous eucalyptus flow time series data collected from the SAPFLUXNET open dataset and 6 environmental factors, including shortwave solar incident radiation, air temperature, air relative humidity, net radiation, vapor pressure deficit, and photosynthetic photon flux density. The experimental results show that the improved Transformer model, with the introduction of a two-step self-attention mechanism and simplified design, maintains significant predictive performance advantages compared to the original Transformer model, long short-term memory, gated recurrent unit, and temporal convolutional neural network models. In the shorter 1-h forecast, the mean squared error and coefficient of determination ( R ² ) of the improved Transformer model are 0.0191 and 0.965, respectively. Compared to the suboptimal typical Transformer model, the MSE is reduced by 22.9%, and R ² is increased by 1.0%. Additionally, the improved model maintains stable predictive performance advantages in long-term plant flow prediction. In the longest 8-h advance prediction, the MSE is reduced by 14.9% compared to the suboptimal Transformer model, and R ² increases by 3.0% compared to the Transformer model. The comprehensive experimental results show that the improved Transformer model makes more effective use of environmental information to achieve more accurate and long-term plant flow prediction. This study emphasizes the basic principle and validity of the two-step self-attention network structure and provides a valuable basis for developing more effective methods for predicting plant sap flow.