Photoperiod controls vegetation phenology across Africa

ARTICLE Photoperiod controls vegetation phenology across Africa Tracy Adole 1*, Jadunandan Dash1, Victor Rodriguez-Galiano2& Peter M. Atkinson1,3,4 Vegetation phenology is driven by environmental factors such as photoperiod, precipitation, temperature, insolation, and nutrient availability. However, across Africa, there’s ambiguity about these drivers, which can lead to uncertainty in the predictions of global warming impacts on terrestrial ecosystems and their representation in dynamic vegetation models. Using satellite data, we undertook a systematic analysis of the relationship between phenological parameters and these drivers. The analysis across different regions consistently revealed photoperiod as the dominant factor controlling the onset and end of vegetation growing season. Moreover, the results suggest that not one, but a combination of drivers control phenological events. Consequently, to enhance our predictions of climate change impacts, the role of photoperiod should be incorporated into vegetation-climate and ecosystem modelling. Furthermore, it is necessary to define clearly the responses of vegetation to interactions between a consistent photoperiod cue and inter-annual variation in other drivers, especially under a changing climate. https://doi.org/10.1038/s42003-019-0636-7 OPEN 1School of Geography and Environmental Science, University of Southampton, Southampton SO17 1BJ, UK. 2Physical Geography and Regional Geographic Analysis, University of Seville, Seville 41004, Spain. 3Faculty of Science and Technology, Lancaster University, Lancaster LA1 4YR, UK. 4School of Geography, Archaeology and Palaeoecology, Queen’s University Belfast, Belfast, BT7 1NN Northern Ireland, UK. *email: [email protected] COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 1 1234567890():,; Vegetation phenology, the seasonal growing cycle of plants, is known to be sensitive to seasonal variation in environmental drivers such as precipitation, temperature, solar insolation, photoperiodic signals, and nutrient availability1–3.At the same time, changes in vegetation phenology can themselves influence some of these varaibles, a phenomenon known as vegetation phenological feedback4. Thus, long-term observation of vegetation phenology can serve as a suitable means of monitoring ecosystem responses to climate variability and change5. Globally, numerous studies, both using ground-based observation and satellite remote sensing-based data have investigated the key drivers of vegetation phenological responses4,5. These studies have reported that in mid and high latitudes, vegetation phenology is controlled mostly by temperature and photoperiod1,6, while it is widely accepted that precipitation (water availability) is the dominant controlling effect on vegetation phenology in the tropics7,8. These factors can affect vegetation phenology together at the same time of year or separately at different times of year1,2, and in a changing climate they can alter the typical timing of phenological events. This has been observed across the northern hemisphere where warmer spring temperatures resulted in an earlier start of season and delayed end of season, leading to longer growing seasons9. While there have been many studies on the environmental drivers on vegetation phenology in the northern hemisphere, few detailed studies quantifying vegetation phenological responses to these environmental drivers have been undertaken in Africa10,11. Indeed, the triggers of onset of vegetation growth and the beginning of dormancy in Africa are poorly understood. There exists much ambiguity about the environmental drivers of vegetation seasonal growth and pattern, for example: the large prerain green up of plants in Africa12,13, irregularities in the relationship between onset of growing season and the beginning of rainy season in the Sahara desert14, and uncertainty in the drivers of changing vegetation greenness in the Sahel15. These uncertainties may have contributed to the observed large biases in atmospheric components (mean temperature/precipitation/solar radiation) of ecosystem models which consequently lead to imperfect predictions of global warming impacts on terrestrial ecosystems16. This is because as much as 38% of the global climate-carbon cycle feedback comes from Africa17. Therefore, it is important to properly characterise Africa’s vegetation phenological responses to environmental cues for accurate parameterisation of global land–atmosphere models. Apart from Guan et al.18, who found that insolation controls tropical evergreen vegetation development, studies investigating the relationship of Africa’s vegetation phenology with environmental factors other than precipitation are scarce. Moreover, most of these studies were undertaken at the regional or country levels. For example, Van Rooyen et al.19, discovered that rainfall is effective in the growth pattern of some woody species in southern Africa only when certain thresholds of temperature and photoperiod have been exceeded. Njoku20,21 also found that vegetation seasonality in some rainforest species in Nigeria is not marked by seasonality in rainfall, but by seasonal changes in day length. This multifaceted association of vegetation phenology with environmental drivers has not been investigated for the entire African continent. As a result, the sensitivity of the precipitationcontrolled African phenology to other environmental drivers is not well understood. This is particularly important in the context of climate change; for example, the effect of projected changes in precipitation on the pre-rain green-up of plants12, and the recently observed lengthening of the growing season in western Africa22. Also, large biases and failures of ecosystem models have highlighted the need to re-evaluate the drivers of vegetation growth and senescence16. Studies have also shown that more accurate representation of vegetation phenological processes in these models have improved the models’climate change predictions and sensitivities23. As a result this research aims to provide greater understanding of the climate-driven vegetation phenology of Africa by undertaking a systematic analysis of the relationship between remotely sensed vegetation phenological parameters (also known as land surface phenology, LSP) and a range of climatic drivers. Since several studies have shown that certain thresholds of temperature, insolation and water availability6,24 are needed to be attained before vegetation growth or dormancy is initiated, “preseason”climatic drivers were used for this research. The preseasons are well-defined periods before phenological dates. We, therefore, aim to quantify the LSP responses to these preseason climatic drivers spatially and, more specifically, determine the dominant climatic driver of the LSP of the major land cover types in Africa. Here, we demonstrate that photoperiod is a more important environmental cue than previously thought, while also acknowledging the significance of other factors. The results support the proposition that vegetation phenology is influenced by the combination of two or more factors rather than a single factor. It also supports previous ideas that LSP response is biomedependent. These findings enhance our understanding of how LSP will respond to variation in environmental drivers under a changing climate. They also highlight the importance of incorporating photoperiod and temperature settings when developing phenological models for tropical regions such as Africa. Results Correlation between Start of Season (SOS) and climatic factors. Differences were observed in the associations between the start of season (SOS, in Day of Year units) and climatic drivers which may be dependent on the vegetation type or the geographical region. The correlation analysis showed that photoperiod had the largest significant correlation with SOS in all vegetation types and geographical regions (Figs. 1–3, and Supplementary Tables 1–3), implying that SOS is mainly controlled by photoperiod in Africa. This large correlation was not as a result of latitudinal influence as mightbehypothesised(sincelatitudeisamajorfactorincomputing daylength; see Supplementary Figs. 1 and 2). Partial correlation results (Supplementary Tables 1–3) revealed that a combination of other climatic factors also exert substantial effects on SOS date. The SOS of all studied vegetation types in the northern region of Africa were significantly and positively correlated with photoperiod in both statistical models (Fig. 1, Supplementary Table 1). These results suggest that the longer the hours of daylight the later the SOS date. The partial correlation analysis revealed that a combination of these factors tends to increase the correlation for all climatic drivers, especially in the 30 and 40 days preseason periods (Supplementary Table 1). The results also showed that the amount of preseason rainfall was significantly and positively correlated with the SOS of croplands, grasslands and woodlands except for the first two preseason periods of shrublands, which had insignificant correlations. Correspondingly, preseason temperature was significantly and negatively correlated with SOS of croplands, grasslands and shrublands with the exception of woodlands which produced small positive correlations. For preseason solar radiation averages, all preseason periods for shrublands, and the first three periods for croplands were generally not significantly correlated with SOS. In contrast, preseason solar sums and SOS were negatively correlated in grasslands, and positively correlated in woodlands. Results in the southern hemisphere of Africa are similar to those in the northern hemisphere except for major differences in croplands and shrublands (Fig. 2, Supplementary Table 2). In ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 2COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio both models, SOS was positively correlated with photoperiod, implying that later vegetation onset dates in the southern hemisphere relate strongly with longer hours of daylight. However, exceptions were SOS dates for croplands in the South-western region, which had much earlier SOS dates, similar to those in western Africa but negatively correlated with photoperiod. In the same way, SOS dates in the extreme north of Africa for croplands and shrublands were significantly and negatively correlated with photoperiod. These outcomes suggest that for preseason periods with longer hours of daylight, decreases in daylength are associated with earlier SOS dates, while for preseason periods with shorter hours of daylight, increases in daylength are associated with earlier SOS dates (Fig. 4). For preseason temperature averages, significant negative correlation values were observed in preseason periods of 10–30 days for croplands and grasslands. For shrublands and woodlands, significant negative values were observed in preseason periods of 60–80 days. These significant negative correlations suggest that earlier SOS dates were associated with higher preseason temperatures. For the significant correlations between preseason solar averages and SOS dates, correlation values were small, suggesting that solar radiation may not have a substantial effect on onset of vegetation growth in these regions. Correlations between EOS and climatic factors. Contrary to the results for SOS, the statistical analysis for EOS revealed the influence of factors which, in addition to photoperiod, are major factors controlling EOS dates in Africa (Figs. 5–7, Supplementary Tables 4–6). For example, for shrublands, SOS had larger correlation values than any of the preseason climatic factors, suggesting that EOS dates are largely controlled by the date of onset of vegetation growing season. In the northern region, while preseason photoperiod was most important for croplands, grasslands and shrublands, in woodlands preseason rainfall had the largest correlation values (Supplementary Table 4). The preseason rainfall was significantly negatively correlated with EOS, suggesting that the greater the amount of preseason rainfall the earlier the EOS dates. Although significantly negative, the preseason photoperiod correlation values were small, suggesting that the duration of sunlight may not have a great effect on senescence of woodlands in northern Africa. However, for croplands and grasslands, the results showed preseason photoperiod as the most important controlling climatic factor. For croplands EOS was significantly and positively correlated with photoperiod. In contrast, for grasslands and shrublands preseason photoperiod was significantly and negatively correlated with EOS. These differences may be due to the length of growing season of croplands, 180 160 140 SOS (DOY) SOS (DOY) SOS (DOY) SOS (DOY) 120 100 80 60 180 160 140 SOS (DOY) 120 100 80 60 180 160 140 SOS (DOY) 120 100 80 60 180 160 140 SOS (DOY) 120 100 80 60 180 160 140 SOS (DOY) 120 100 80 60 SOS (DOY) 150 160 170 180 190 SOS (DOY) 60 80 100 120 140 SOS (DOY) 60 80 100 120 140 SOS (DOY) 60 80 100 120 140 SOS (DOY) 60 80 100 120 140 SOS (DOY) 150 160 170 180 190 SOS (DOY) 150 160 170 180 190 SOS (DOY) 150 160 170 180 190 0 0 0 0 20 40 60 80 100 2000 4000 6000 8000 10,000 800 850 900 950 1000 1050 1100 360 370 380 390 50 100 150 200 250 1000 1500 2000 2500 3000 3500 700 800 900 1000 370 375 380 385 390 20 40 Rain (mm) 60 80 50 100 Rain (mm) Solar (Wm –2 ) Croplands Grasslands Shrublands Woodlands Temp (°C) Photo (h) 150 200 R2 = 0.09 R2 = 0.11 R2 = 0.3 R2 = 0.02 R2 = 0.01 R2 = 0.89 R2 = 0.94 R2 = 0.15 R2 = 0.12 R2 = 0.37 R2 = 0.02 R2 = 0.4 R2 = 0.54 R2 = 0.09 R2 = 0.02 R2 = 0.93 4000 6000 8000 10,000 Solar (Wm –2 ) 4000 6000 8000 10,000 700 800 900 Temp (°C) 1000 1100 392 394 396 398 Photo (h) Rain (mm) Solar (Wm –2 )Temp (°C) Photo (h) Rain (mm) Solar (Wm –2 )Temp (°C) Photo (h) 400 402 404 1100 370 380 390 4001000950900850 180 160 140 120 100 80 60 180 160 140 120 100 80 60 180 160 140 120 100 80 60 Fig. 1 Scatterplots between SOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the Northern hemisphere of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 1). Shrublands were located in the horn of Africa COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 ARTICLE COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 3 and the photoperiod seasonality in the Sudano–Sahel region, which increases at the start of the year, peaks in the middle and declines towards the end of the year. As previously mentioned, preseason temperature, like photoperiod, plays an equally important role in EOS onset in the southern hemisphere (Supplementary Table 5), and is the primary factor in the extreme north of Africa (Supplementary Table 6). The exception to this is preseason temperature for croplands in southwestern Africa. Partial correlation analysis revealed that although preseason temperature was significantly and negatively correlated with EOS, correlation values in all preseason periods were less than that of photoperiod. For grasslands, shrublands and woodlands in southern Africa, preseason temperature was significant and positively correlated with EOS, suggesting that increases in preseason temperature may delay EOS dates. On the other hand, preseason temperature for croplands and shrublands in the extreme north of Africa was significantly and negatively correlated with EOS, suggesting that increases in preseason temperature may result in earlier EOS dates (Supplementary Table 6). These negative correlations at very high latitudes and positive correlations at lower latitudes suggest that preseason temperature increases at higher temperatures contribute to earlier EOS while temperature increases at lower temperatures contribute to later EOS. The results for preseason photoperiod were somewhat different from those of temperature. In the southern hemisphere, for grasslands, shrublands and woodlands, increasing preseason photoperiod was associated with earlier EOS dates, while in the extreme north, increasing preseason photoperiod was associated with later EOS dates. Changes in associated climatic factors and inter-annual variation in LSP. As expected, there was no change in inter-annual trends of the preseason photoperiod estimated from median SOS and EOS values over the time period, as photoperiod is consistent from year-to-year. Also, significant inter-annual trends in preseason precipitation and solar radiation were found to occur with no particular spatial pattern, and no association was found with significant changes in LSP dates. On the other hand, preseason temperature for SOS and EOS dates revealed significant increasing inter-annual trends with an average slope of 0.33 °C/ year, and mostly in the Sahel region overlapping with areas showing significant trends in LSP dates over the time period (Fig. 8). This suggests that changes in preseason temperatures may have influenced observed inter-annual variation in LSP dates. 140 120 SOS (DOY) 100 80 0 0 0 0 20 40 60 80 100 120 1000 2000 3000 4000 5000 6000 700 800 900 1000 1100 350 370 390 410 20 40 60 80 3000 4000 5000 6000 7000 500 600 700 800 900 1000 320 340 360 380 400 420 50 100 150 200 2000 3000 4000 5000 500 600 700 800 330 350 370 390 20 40 60 Rain (mm) TEMP (°C) Photo (h) Solar (Wm–2) Rain (mm) TEMP (°C) Photo (h) Solar (Wm–2) Rain (mm) TEMP (°C) Photo (h) Solar (Wm–2) Rain (mm) TEMP (°C) Photo (h) Solar (Wm–2) 80 100 R2 = 0.06 R2 = 0.03 R2 = 0 R2 = 0 R2 = 0.28 R2 = 0.99 R2 = 0.99 R2 = 0.97 R2 = 0.33 R2 = 0.06 R2 = 0.11 R2 = 0.17 R2 = 0.18 R2 = 0.83 R2 = 0.23 R2 = 0.26 120 2000 4000 6000 8000 400 500 600 700 340 360 380 400 300 290 280 270 260 250 240 SOS (DOY) 340 320 300 280 260 240 220 200 SOS (DOY) 360 340 320 300 280 260 240 SOS (DOY) 360 340 320 300 280 260 240 SOS (DOY) 360 340 320 300 280 260 240 SOS (DOY) 360 340 320 300 280 260 240 SOS (DOY) 340 320 300 280 260 240 220 200 SOS (DOY) 340 320 300 280 260 240 220 200 SOS (DOY) 340 320 300 280 260 240 220 200 SOS (DOY) 300 290 280 270 260 250 240 SOS (DOY) 300 290 280 270 260 250 240 SOS (DOY) 300 290 280 270 260 250 240 SOS (DOY) 140 120 SOS (DOY) 100 80 140 Croplands 120 SOS (DOY) 100 80 140 120 SOS (DOY) 100 80 Grasslands Shrublands Woodlands Fig. 2 Scatterplots between SOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the Southern hemisphere of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 2). Croplands were located in the southwestern region of Africa with a similar climate to the Sudano-Sahel region of western Africa ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 4COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 0.40 a b SOS SOS EOS EOS 0.35 0.30 EVI EVI 0.25 0.20 0.15 0.2 0.3 0.4 11.5 10.0 10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 12.0 12.5 Photo (h) Photo (h) 0.1 Aug Oct Dec Feb Apr June Aug Months Oct Dec Feb Apr Aug Oct Dec Feb Apr June Aug Months Oct Dec Feb Apr Fig. 4 Example of pixel profile for a complete EVI and Photoperiod daily time-series: aCroplands in the Sudano-Sahel region showing increase in (longer) growing season with increasing preseason photoperiod at the start of vegetation growing season (i.e., synchrony), bCroplands in the south western region showing increase in (shorter) growing season with decreasing preseason photoperiod at the start of vegetation growing season (i.e., asynchrony). EVI time-series is represented by blue curves while photoperiod is represented by red curves. Vertical dashed black lines show LSP parameters (SOS and End of Season; EOS) 360 340 320 300 SOS (DOY) 280 260 240 360 340 320 300 SOS (DOY) 280 260 240 0 20 40 60 80 100 120 4000 6000 8000 10,000 12,000 200 300 400 500 600 700 800 320 340 360 380 400 420 360 340 320 300 SOS (DOY) 280 260 240 360 340 320 300 SOS (DOY) 280 260 240 360 340 320 300 SOS (DOY) 280 260 240 0 50 100 Rain (mm) Rain (mm) Solar (Wm–2) Solar (Wm–2) Temp (°C) Photo (h) Temp (°C) Photo (h) 150 200 250 2000 10,000 300 400 500 600 700 800 900 300 320 340 360 380 400 420800060004000 360 340 320 300 SOS (DOY) 280 260 240 360 340 320 300 SOS (DOY) 280 260 240 360 340 320 300 SOS (DOY) 280 260 240 R2 = 0.01 R2 = 0.01 R2 = 0.64 R2 = 0.98 R2 = 0.98 R2 = 0.71 R2 = 0.13 R2 = 0.22 Croplands Shrublands Fig. 3 Scatterplots between SOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the extreme northern part of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 3) COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 ARTICLE COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 5 Discussion Across Africa, photoperiod was found to be the dominant factor controlling the onset and end of vegetation growing season. This highlights the high sensitivity of plants to photoperiod, a phenomenon that has been documented as early as 1950s20,21,25,26. This dominant control by photoperiod tends to corroborate earlier work which attributes photoperiod relative to some threshold to be the major determinant that allows other climaticdriven development to occur27. It also provides more evidence supporting the pre-rain green-up phenomenon observed in Africa12, further challenging the widely held belief that onset of vegetation growing season in Africa is water limited. The results are also in agreement with earlier studies indicating that a combination of climatic factors, either occurring simultaneously or preceding one another, controls LSP patterns, with their effects sometimes biome-dependent1,2,28. It further supports the idea of incorporating photoperiod into terrestrial biosphere models for increased accuracy of prediction29,30. In the northern latitudes of Africa, photoperiod and temperature were found to be the major climatic factors controlling the onset and end of vegetation growing season. This result for Africa is nevertheless consistent with other research which concluded that photoperiod is the major factor controlling phenological events in tropical ecosystems20,31. In the extreme north of Africa, the wet season is usually accompanied by a declining daylength duration and an increasing temperature. With this research revealing negative correlations between SOS and preseason photoperiod, and positive correlations between SOS and preseason temperature, it can be inferred that a combination of lower temperature limits and higher photoperiod limits are the cues required for the initiation of vegetation growth in the extreme north of Africa. These findings are similar to typical vegetation phenology drivers for northern hemisphere1,6. For onset of dormancy, the result suggests that the reverse (higher temperatures and lower photoperiod) may be the environmental cue, with temperature playing a more dominant role. In the Sahel region, onset of growing season for all studied vegetation types is predominantly controlled by photoperiod, suggesting its strong sensitivity to sunshine duration. This sensitivity has been reported for a wide range of vegetation types in the Sudanian region32 and is also known to be genetically based in some cereal crops, including major varieties grown in West Africa33,34. Nonetheless, in this region, the role of other factors and their combinations are still important as shown in the results. For example, the onset of vegetation growing season is characterised by increasing daylength duration and increasing temperatures. The photoperiod seasonality from the beginning of the 460 440 420 400 EOS (DOY) 380 360 330 320 310 300 EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) 290 280 330 320 310 300 290 280 330 320 310 300 290 280 330 320 310 300 290 280 450 400 350 300 EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) 250 450 400 350 300 250 450 400 350 300 250 450 400 350 300 250 0 50 100 150 200 250 300 1000 1500 2000 2500 3000 750 850 950 1050 360 365 370 375 380 420 410 400 390 EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) 380 370 360 0 20 40 60 80 2000 4000 6000 8000 10,000 900 1100 1200 355 360 365 3701000 420 410 400 390 380 370 360 420 410 400 390 380 370 360 420 410 400 390 380 370 360 0 010 20 30 40 4000 6000 8000 10,000 850 1050950 1150 355 360 365 370 375 5 Rain (mm) Rain (mm) Rain (mm) Rain (mm) Temp (°C) Temp (°C) Temp (°C) Temp (°C) Photo (h) Photo (h) Photo (h) Photo (h) Solar (Wm–2) Solar (Wm–2) Solar (Wm–2) Solar (Wm–2) 10 15 4000 6000 8000 10,000 900 950 1000 1050 1100 350 355 360 365 370 375 460 440 420 400 EOS (DOY) 380 360 460 440 420 400 EOS (DOY) 380 360 460 440 420 400 EOS (DOY) 380 360 Woodlands Shrublands Grasslands Croplands R2 = 0 R2 = 0.05 R2 = 0.07 R2 = 0.7 R2 = 0.92R2 = 0.15 R2 = 0.17 R2 = 0 R2 = 0 R2 = 0.22 R2 = 0.03 R2 = 0.03 R2 = 0.14 R2 = 0.1 R2 = 0.17 R2 = 0.19 Fig. 5 Scatterplots between EOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the Northern hemisphere of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 4). Shrublands were located in the horn of Africa ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 6COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com /commsbio year usually begins with longer day length of over 11 h, and rising temperatures of over 20 °C. With both factors having significant correlations (photoperiod with the largest), these observations suggest that lower photoperiod limits and warmer temperatures (negative correlation of preseason temperatures), coupled with the timing of the onset of rainy season could be responsible for the initiation of vegetation green-up in this region. These conditions are mostly favourable to tropical plants, known to grow well in warmer temperatures and with shorter photoperiods28, particularly millet and sorghum, two of the most important cereal crops grown in Sudano–Sahelian region35. Surprisingly, preseason rainfall had little or no significant effect on onsets dates, suggesting that the amount of precipitation plays a secondary or no role when compared to photoperiod and temperature in regulating the start of vegetation growing season in the Sahel. This is supported by earlier studies which reported a large percentage of pre-rain green-up has been observed in this region12. In addition, contrary to expectations, significant positive correlations were observed between preseason rainfall and SOS dates. A possible explanation for this might be that greater amounts of rainfall are usually accompanied by clouds, thus, reducing temperatures and sunshine intensity below growth initiation thresholds, hence resulting in a later onset dates36. For the onset date for grasslands, smaller positive correlations with preseason photoperiod and a slightly greater negative correlation of solar radiation and temperatures were observed in this region, compared to other land cover types. Studies have shown that for grasses (mainly of C4 type) found in many African ecosystems phenology is driven by high solar radiation and temperatures37. It has been established that C4 plant types are better adapted to warm climates because of their enzyme sensitivity to chilling temperatures38. They are also known to have greater photosynthetic capacity at higher sunlight and temperature levels39. These factors may explain the relatively significant negative correlation of solar radiation and temperature with the onset dates of grasslands. This further supports the recommendation that thermal scenarios should be considered when investigating grassland phenology40, since increased amount and duration of rainfall had no effect on its phenology events. These findings in general, raise the likelihood of a vegetation type dependency of LSP responses to climatic factors. Additionally, it also highlights the much reduced role of rainfall seasonality in the vegetation growth cycle. However, it is important to note that although most studies have shown a large correlation between rainfall and vegetation seasonality, this association is more related to timing (onset of raining reason and onset of vegetation 340 R 2 = 0.04 R 2 = 0.02 R 2 = 0.03 R 2 = 0.31 R 2 = 0.03 R 2 = 0.4 R 2 = 0.62 R 2 = 0.24 R 2 = 0.02 R 2 = 0.6 R 2 = 0 R 2 = 0 R 2 = 0.23 R 2 = 0.23 R 2 = 0.6 R 2 = 0.98 330 320 310 EOS (DOY)EOS (DOY)EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) EOS (DOY) 300 290 280 0 0 10 20 30 40 50 2000 3000 4000 5000 450 500 550 600 650 700 315 320 325 0 10 20 30 40 2000 4000 5000 6000 7000 500 600 700 800 900 310 320 330 340 340 360 380 400 350 360 3703000 0 50 100 150 200 250 1000 2000 3000 4000 5000 500 600 700 800 900 1100 190 180 170 160 180 140 120 200 150 100 50 0 200 150 100 50 0 200 150 100 50 0 200 150 100 50 0 160 180 140 120 160 180 140 120 160 180 140 120 160 190 180 170 160 190 180 170 160 190 180 170 160 20 Rain (mm) EOS (DOY) Rain (mm) Rain (mm) Rain (mm) 40 60 80 100 2000 Solar (Wm –2 ) Solar (Wm –2 ) Solar (Wm –2 ) Solar (Wm –2 ) Temp (°C) Temp (°C) Temp (°C) Temp (°C) Photo (h) Photo (h) Photo (h) Photo (h) 4000 6000 500 600 700 800 380 390 400 410 420 430 340 Croplands Grasslands Shrublands Woodlands 330 320 310 300 290 280 340 330 320 310 300 290 280 340 330 320 310 300 290 280 Fig. 6 Scatterplots between EOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the Southern hemisphere of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 5). Croplands were located in the southwestern region of Africa with a similar climate to the Sudano-Sahel region of western Africa COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 ARTICLE COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 7 growing season) than to pre-seasonal rainfall amounts. As shown in this research, the amount of rainfall has little or no significant influence on onset dates. Rather, the association is largely a timebased relationship as shown in previous studies12,41. Unlike the onset of vegetation growing season, for dormancy onsets, preseason photoperiod was not the only major determining factor. While preseason photoperiod was the predominant factor controlling dormancy onsets in croplands and Trend ab cd Negatively significant at 95% Positively significant at 95% Not significant Fig. 8 Spatial distribution of significant inter-annual trends of LSP parameters and the associated 30 days preseason cumulative temperature from 2001 to 2015. aSignificant SOS trends, bsignificant EOS trends. cPixels of significant trends in preseason temperature of SOS which overlap with significant trends in SOS dates. dPixels of significant trends in preseason temperature of EOS which overlap with significant trends in EOS dates EOS (DOY) 200 R 2 = 0.01 R 2 = 0.13 R 2 = 0 R 2 = 0 R 2 = 0.71 R 2 = 0.27 R 2 = 0.8 R 2 = 0.9 180 160 140 120 0 0 10203040506070 300 400 500 600 700 800 900 400 410 420 430 440 4504000 6000 8000 10,000 12,000 20 40 60 80 100 120 2000 4000 6000 8000 10,000 400 600 800 390 400 410 420 430 440 4501000 EOS (DOY) 200 180 160 140 120 EOS (DOY) 200 180 160 140 120 EOS (DOY) 200 180 160 140 120 EOS (DOY) 200 180 160 140 120 EOS (DOY) 200 Croplands Shrublands 180 160 140 120 EOS (DOY) 200 180 160 140 120 EOS (DOY) 200 180 160 140 120 Rain (mm) Rain (mm) Temp (°C) Temp (°C) Photo (h) Photo (h) Solar (Wm –2 ) Solar (Wm –2 ) Fig. 7 Scatterplots between EOS and climatic drivers cumulated over a 30 day preseason period across different vegetation types in the extreme northern part of Africa (All at P< 0.05). Plots for other preseason periods are not shown (see Supplementary Table 6) ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 8COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com /commsbio grasslands, other factors were shown to be more significantly associated with EOS dates. In shrublands, SOS and preseason rain were more dominant, and in woodlands, preseason rain and solar radiation showed more dominance, although their effect is dependent on the preseason period. This negative correlation of preseason rainfall could be caused by the accompanying reduced temperatures not favourable to vegetation growth as explained above. Apart from the human factor in agricultural lands (irrigation/ farmers’decision of sowing dates)2,33, the length of growing season as a function of the vegetation type could also be a factor that can contribute to the effect of preseason climatic factors on EOS dates. For example, in croplands preseason photoperiod was significantly positive, whereas that of grasslands was significantly negative. In croplands the length of growing season extends to periods at the beginning of the year where there is a small but increasing photoperiod. While, in grasslands with a short growing season of approximately six months, the photoperiod towards the end of the season is large but declining. Also, in shrublands, photoperiod had larger significantly negative correlation values in the preseason periods of 2–3 months before the onset of dormancy dates, suggesting that photoperiod 2–3 months before onset of dormancy plays a major role in regulating EOS in shrublands. Likewise important in determining EOS is the timing of the onset of growing season in shrublands (significant SOS values in shrublands). Similarly, in the south photoperiod was the major climatic factor controlling onset of vegetation growing season while other factors showed significant control of vegetation dormancy onset. These findings are consistent with Garonna et al.42 and corroborate the idea that photoperiod is the most reliable predictor of onset dates for southern African savanna trees43,44. Equally, the apparent positive effect of preseason photoperiod was as a result of higher and increasing preseason photoperiod. However, the observed negative correlation of preseason photoperiod on croplands in south-western Africa can be attributed to the declining duration of day length which is similar to the photoperiodicity of the extreme north of Africa. Also significant were preseason temperatures for grasslands and croplands which had a negative correlation, with warmer temperatures favouring earlier vegetation green up. These results confirm previous suggestions that a combination of photoperiod and temperature thresholds are environmental cues for vegetation growth in southern Africa19. Preseason rainfall amount had no effect on SOS, except for preseason periods in grasslands. This was expected as pre-rain green-up has been reported to be ubiquitous in southern African savanna, with as early as 60 days before the first rains12,13.In addition, these results seem to be consistent with other research which found that rainfall clearly had no effect on the development of leaves in some southern Africa savanna trees43. This further confirms that most of the associations between rainfall and vegetation seasonality are related mainly to time and productivity45. For example, the memory mechanism of miombo woodlands: greening-up in anticipation of onset of rains46 (timebased), and the intra-seasonal rainfall variability effect on sorghum yield47 (productivity-based). The onset of vegetation dormancy was influenced not only by photoperiod but also by other climatic factors. In croplands, a positive correlation of preseason photoperiod was dominant, while in other studied vegetation types, preseason photoperiod had a negative influence. However, this negative influence of photoperiod is secondary to the positive influence exerted by preseason temperature in grasslands. Also in shrublands and woodlands, the influence of preseason temperature was significantly high, suggesting that temperature increases postpone the onset date of vegetation dormancy. This observation is consistent with earlier studies which showed that increases in temperature may have extended the vegetation growing season in the Namaqualand, southern Africa45. In contrast, the effect of preseason rainfall suggests that increasing rainfall led to earlier onset date of vegetation dormancy. Again, the accompanied reduced temperature during rainfall could be responsible for the negative correlation48. In general, we observed an overall synchrony between photoperiod and LSP parameters across all of Africa, an observation supported by several studies highlighting photoperiod control of leaf flushing rather than rainfall31,49. A possible explanation for this may be due to the fact that photoperiod is the most consistent environmental signal from year-to-year1,49. As result of this consistency, plants may tend to rely more on specific day length signals to regulate their growth34. This can be seen in the results showing that increasing preseason photoperiods of above 12 h duration tend to be associated with later SOS and earlier EOS, while increases of above 10 h were associated with earlier SOS and later EOS. Similarly, decreasing preseason photoperiods of below 12 h were associated with earlier SOS and later EOS. Hence, it is possible to hypothesise that longer day length duration of above 12 h tends to delay the onset of vegetation growing season and initiate dormancy, while a duration of <12 h but above 10 h may initiate SOS and delay EOS. This suggest that a certain threshold of day length must be exceeded to initiate the onset of vegetation growing season, and in the same way initiate the end of vegetation growing season. This distinct change in the response of plants to small changes of 2 h or less in photoperiods has been reported previously31. Likewise, it has also been suggested that plants respond to specific critical daylength (varies from plant to plant) during which hormonal regulation of growth initiation or cessation hormones occurs50,51. This ability of plants to detect light and measure time very accurately has been attributed to an “endogenous time-keeping mechanism called the circadian clock”50, and the perception of light signals by photoreceptors3, and these clock and photoreceptors genes can be found in all living plant cells51. These responses to photoperiod have been shown to influence the population structure of major crops like millet and sorghum in Western and Central Africa than any other environmental factor, a sole factor for adaptation to environmental constraints35,52. However, there are still many unanswered questions about how this mechanism works with phenological parameters especially in different plant types, and more investigations are required as recommended by other studies4,53. A particular interesting observation which further supports the sensitivity of plants to small changes in photoperiods is the distinct response of croplands in the Sudano-Sahel region of western Africa and the croplands in south-western Africa. Although, estimated SOS beginning around May/June were similar for both croplands, however their responses to photoperiodic signals were very distinct. This can be attributed to the increasing preseason photoperiod at the start of the year observed in Sudano-Sahel region, and the decreasing preseason photoperiod observed at the same period in south-western Africa. This distinct response of crops to the direction of photoperiod also reflects the results from Nori et al.54, who found out that leafing rate was determined by the duration and direction of photoperiod at seed germination. Also, the length of growing season as a function of the crop type may also play a role in these responses. For example, maize crop mostly grown in the south-western region have shorter growing season and harvested much earlier than those (millet/sorghum/ cassava) in the Sudano–Sahel region55. Irrespective of the observed dominance of photoperiod, the partial correlation results also showed that LSP is influenced by a COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0636-7 ARTICLE COMMUNICATIONS BIOLOGY | (2019) 2:391 | https://doi.org/10.1038/s42003-019-0636-7 | www.nature.com/commsbio 9