Changes In Rainfall Seasonality And Their Impacts On Vegetation Across Tropical Savannas Based On Multi-Source Data

Global warming is expected to cause climate change with a higher frequency of extreme rainfall events.More extreme events may affect the water balance and primary productivity,ultimately threatening food security and people's general wellbeing.Therefore,knowledge on changes in rainfall seasonality and how such changes affect vegetation productivity and ecosystem functioning are important for understanding vegetation phenology and guiding agricultural production.Moreover,it also contributes to the understanding of global warming effect on rainfall regimes and our biosphere.Here,our study applied daily satellite based rainfall estimates(TRMM,CHIRPS,GPCC and MSWEP),monthly rainfall data sets(CRU TS4)and Normalized difference vegetation indices(GIMMS 3g-vl and MODIS NDVI),leaf area index(GIMMS 3g LAI)and vegetation depth optical(VOD)to investigate:1,changing characteristics of rainfall seasonality;2,impacts of changed rainfall regimes(e.g.onset of wet season,rainfall intensity)on vegetation productivity and season NDVI;3,increased heavy rainfall frequency impact on ecosystem structure(by altering the herbaceous/woody vegetation composition).The main results of this study were shown as follows:(1)In this study,the combined use of a generalized additive model(GAM)(to fit a model describing changes of cumulative rainfall anomalies on a daily scale)and the Akaike information criterion(AIC)(generally used to assess the goodness of fit)was found to succeed in capturing global wet season domains.Areas of distinct rainfall seasonality are observed to be mainly located in the Amazon rainforest,northern Australia,southern Asia,South Africa and Sahel.Seasonal rainfall amount accounted for 82 ± 10(%)of the annual rainfall over wet season domains.The seasonality of rainfall and the length of wet season showed decreasing tendencies moving from the equator towards higher/lower latitudes,and the onset and cessation of wet season showed opposite patterns in the northern and southern hemispheres.We then found that largely consistent results of inter-annual variability in seasonal rainfall metrics(seasonal rainfall,onset,cessation and length of wet season)can be derived for different rainfall products for different monsoon regions(Brazil,Sahel,South Africa and India).Only exception is for GPCC rainfall of which moderate differences in seasonal rainfall metrics were observed in Sahel and southern Africa when compared to the other three rainfall products(TRMM,CHIRPS and MSWEP).The significant correlation(p<0.05)between the onset,cessation and length of wet season and seasonal rainfall implied that the changed seasonal rainfall metrics can contribute to understanding the change of rainfall.Specifically,the correlation between seasonal rainfall and the length of wet season is the strongest with r =0.58,while the onset of wet season is negatively related to seasonal rainfall with r =-0.45.(2)This study tested the importance of rainfall metrics in the wet season(onset and cessation of the wet season,number of rainy days,rainfall intensity,number of consecutive dry days and heavy rainfall events)on growing season ANPP.We focused on the Sahel and north-Sudanian region(100-800 mm yr-1)and applied daily satellite based rainfall estimates(CHIRPS v2.0)and growing season integrated NDVI(MODIS)as a proxy for ANPP over the study period 2001—2015.Growing season ANPP in the arid zone(100—300 mm yr-1)was found to be rather insensitive to variations in the seasonal rainfall metrics,whereas vegetation in the semi-arid zone(300—700 mm yr-1)was significantly impacted by most metrics,especially by the number of rainy days and timing(onset and cessation)of the wet season.We analyzed critical breakpoints for all metrics to test if vegetation response to changes in a given rainfall metric surpasses a threshold beyond which vegetation functioning is significantly altered.It was shown that growing season ANPP was particularly negatively impacted after>14 consecutive dry days and that a rainfall intensity of?13 mm day-1 was detected for optimum growing season ANPP.(3)This study analyze impacts of seasonal rainfall metrics(onset,cessation,length of wet season and seasonal rainfall)on season NDVI across tropical savannas for areas of rainfall showing a well-defined annual cycle.The study makes use of the daily of rainfall estimates(CHIRPS v2)and the average NDVI(GIMMS 3g-vl)of the growing season covering the period 1982 to 2015.We first used the timing of wet season to determine the growing season and the period between onset and cessation was used to calculate the growing season NDVI at the per-pixel level.Seasonal rainfall amount was found to be significantly positive correlated with NDVI for largest areas across tropical drylands,closely followed by sizable areas characterized by a significant correlation between NDVI and length and onset of wet season.The spatial distribution of season NDVI most significantly affected by the different seasonal rainfall metrics were mainly found in dryland areas with seasonal rainfall less than-900 mm yr1 and at the same time characterized by a distinct seasonality of the rainfall distribution.In addition,significant changes in the sensitivity of season NDVI to seasonal rainfall metrics were observed during the past three decades in most drylands,particularly with a significant decreased trend in sensitivity of vegetation to seasonal rainfall amount shown in Sahel.(4)Changes in the temporal distribution of rainfall are projected to intensify in the future and may impact on the structure of tropical savannas ecosystems by favoring woody plants,relative to herbaceous vegetation.Our studies found that the frequency of heavy rainfall increased on average by 2.6%for tropical savannas during this period.This increase in heavy rainfall frequency was linked with a decoupling sensitivity of herbaceous vegetation(expressed as the maximum LAI during growing season)from annual rainfall showing that the functioning of global tropical savannas ecosystems has been substantially changed in recent decades.A significant decoupling(p<0.05)was observed over 20%of the tropical savannas area(189 750 km2),of which 35%also showed a simultaneous increase in woody cover and heavy rainfall frequency.In total,23%of the tropical savannas area(218 700 km2)had shown a spatio-temporally consistent increases in both woody cover(by 5.5%)and heavy rainfall frequency.Overall,the methods presented in this study collectively provide a new objective approach for identification of monsoon domains and our findings cast new recognition on the role of global intensified rainfall under global warming.We concluded that number of rainy days and the timing of the wet season are seasonal rainfall metrics being decisive for favorable vegetation growth in drylands which needs to be considered when modelling primary productivity from rainfall in the drylands of Sahel and elsewhere.The changes in frequency of heavy rainfall and woody cover and relationship between rainfall and LAI observed in our study suggest a climate-induced shift in the coexistence of woody and herbaceous vegetation in savannas ecosystems,possibly caused by altered hydrological conditions with global significance for land cover and associated biophysical effects such as surface albedo and evapotranspiration.Thus,improved knowledge on the vegetation productivity response to the seasonal variability of rainfall and impacts of increased frequency of extreme events on ecosystem functioning is crucial to better interpret the consequences of climate predictions of an altered global hydrological cycle.This is of particular importance for implementation of appropriate adaptation measures to climate change and food security in tropical drylands.