Using a robust recent dataset, our analyses show that cocoa production is significantly affected by the maximum magnitude of ENSO phase during the current and previous purchase years (Fig. 2). The instantaneous effect is negative, followed by delayed positive effects in the two following years and negative in the third following year, combining to give a picture of multi-year fluctuations in cocoa production as a result of El Niño/La Niña events. Using a 70-year dataset, we show significant changes in these instantaneous and delayed ENSO-production relationships between recent and past time periods (Fig. 3). Using ERA5 data for the cocoa production area of Ghana, summarised at the same temporal resolution as the production data, we demonstrate significant relationships between ENSO phase and climate, with significant changes in mean climate and in ENSO-climate relationships (Fig. 4) between recent and past time periods. This agrees with prior work suggesting that ENSO may impact West Africa5,15, despite no current evidence of teleconnections between ENSO phase and West African climate17.
Our 70-year production dataset represents a temporal extent unmatched by other research, however was aggregated to fewer replicates than the 21-year analysis (6 regions vs 68 districts). While this may represent reduced power, results from the overlapping time period of the two datasets strongly agree. The computation of yield, a more comparable metric between different-sized areas than total production, was not possible because data on area under production (AUP) were not available. However, the detrending process employed successfully eliminated variation between districts or regions (of which AUP is likely a substantial component) and long-term technological trends that would otherwise confound our ability to isolate the ENSO signal (Supplementary results).
Perennial crops have multi-year growing patterns, with allocation of resources to growth, development and reproduction driven by climate in ways that are not fully understood29. ENSO generally peaks between October and December, also the busiest cocoa purchase period: thus we observe a relatively instantaneous apparent effect of ENSO phase on cocoa production. This reduction in cocoa production under El Niño inis consistent with results from farm monitoring8 and large-scale farm surveys30 evidencing production declines in from other regions (where teleconnections are understood), and with analyses of production data from West Africa31. During the main cocoa purchase period, coinciding with the minor wet and major dry seasons, we observe increases in water deficit during El Niño, leading to drought stress conditions. In small-scale cocoa studies, drought stress is correlated with reduction in pod production and increased tree mortality8,32, and in similar studies of other tree crops drought is directly linked to reduction in fruit or nut production33, although in all cases the mechanisms are unclear. Drought may generally create unfavourable conditions for growth and reproduction through reduced availability of water for vital processes, or more specifically by promoting disease incidence and pod rot8, increasing the chance of fire, increasing competition for soil moisture32, and/or reducing pollinator populations34. Alternatively, cocoa may respond to reduced water availability by reallocation of resources away from energetically expensive reproduction: rainfall exclusion experiments suggest that in the medium term, while bean production drops, vegetative growth is not significantly reduced during drought32.
The significant increases in mean temperature and average drought stress we observed in some seasons over time is such that the climate experienced during El Niño events in recent decades represent novel extreme conditions for Ghana’s cocoa agriculture. This causes significant changes in the responses of cocoa production to ENSO phase over the same time period. One explanation for this may be that the warm, dry El Niño conditions in Ghana in the past were within the environmental tolerance of cocoa, leading to allocation of resources to reproduction in response to drought, increasing cocoa bean production and resulting in less severe instantaneous and delayed responses to ENSO phase (Fig. 3a–d) However, in recent decades this level or greater drought stress has become the norm (Fig. 4i–l), with El Niño conditions apparently triggering a different response mode, allocating resources away from reproduction in the short term and creating oscillating resource allocation over the following years.
However, understanding the delayed responses of cocoa is challenging, especially as these represent a novel finding. There is little research that explores multi-annual physiological or ecological responses of cocoa to drought, and the explanation is likely to be a combination of both residual/delayed climatic responses to ENSO phase, and of life history strategies. The observed increase in production during the two years following El Niño may be explained by post-drought reallocation of resources to reproduction as remediation for lost reproductive output in the instantaneous response, or a shift to a ‘faster’ strategy by allocating resources to reproduction over the longer term, becoming evident in the data in subsequent years. Alternatively, this may be explained by favourable climatic conditions occurring during an El Niño event that impact the following years’ crop. March and April is a crucial time for cocoa pod development in Ghana and in recent years El Niño appears to bring greater rainfall during these months. Given the 6–9 months development of cocoa beans, the effects of this increased rainfall and reduced water deficit on cocoa production will be seen in the delayed response. We see evidence of this in the climate-change driven reversal of March–April rainfall patterns: while in the past El Niño has consistently resulted in drought stress, this reversal provides a respite from drought, buffering trees from reduced rainfall during the major wet season and giving sufficient resources for improved production in the following year.
The robustness of our results provide evidence that may aid development of resilience strategies for ENSO-driven cocoa production variation in Ghana, but we may also consider whether these results can be generalised to the production of cocoa and/or perennial tree crops globally. The climatic impact of ENSO observed in Ghana is broadly consistent with many regions of the tropics2, the instantaneous cocoa production responses to El Niño are consistent with findings in these regions, and so we may expect these regions to see a similar pattern of multi-annual cocoa production variation in response to ENSO phase. However, there is considerable variation in ENSO responses among and within other perennial tree crops in regions where climatic responses to ENSO are similar to Ghana. Oil palm yields have been negatively associated with ENSO phase in Malaysia9, as have olive yields in Morocco (delayed by a year)33. Conversely, apple yields have been positively associated with ENSO phase in China10, as have coffee yields in Brazil35; however, no effect at all is seen in coffee in India over a 35-year time series7. Most of these analyses considered only a single ENSO phase (usually El Niño), and most considered only instantaneous impacts. However, it is clear that most of these crops do respond to ENSO, and given the shared biology it is reasonable to assume that delayed effects of ENSO phase are likely and should be considered to understand the full picture of ENSO impacts on perennial tree crops.
The larger body of research into ENSO impacts on annual crops includes many studies using long time series, reporting high heterogeneity in space and among crops11,36,37. However, there appears to be little examination of changes in the direction and magnitude of ENSO responses over time; thus our findings are timely and signal that further research is needed to examine how changing climates may force novel extreme climatic conditions and shift response patterns to ENSO phase. Given that perennial tree crops are generally cash crops, and the utility of these crops to farmers are to a greater or lesser extent mediated by market forces, there is a need for improved forecasting of yield in response to changing climate and ENSO patterns to withstand production fluctuations. The low perishability of many perennial tree crops means that with accurate forecasting, supply may be managed or even exploited to ensure consistency of income both for farmers and those whose livelihoods depend on related food manufacturing industries.
Our approach to understanding the responses of a perennial tree crop to ENSO phase and anthropogenic climate change exploited existing global, national and subnational datasets for climate and production with appropriate spatial and temporal resolution. We use freely available geographic and climate data, and employ highly replicable methods: a simple pipeline of climate data aggregation and summary computation, coupled with standard detrending and straightforward analytical methods with a relatively small computational requirement. This “big data” approach to agriculture-climate research demonstrates a relatively straightforward framework for understanding responses of agricultural productivity to climate and identifying temporal changes in these relationships. While small-scale studies examine the mechanisms of climate impacts through the interacting effects of agricultural practices, abiotic conditions, disease incidence and multi-trophic interactions, large-scale studies across regions and over time scales encompassing many ENSO oscillations are required to understand the global picture of perennial tree crop production security. Combined with local context-specific studies on governance arrangements16, such approaches could be crucial for reducing future vulnerability of these industries to increasing volatility under anthropogenic climate change. The main barrier to this research is the availability of production data from state or commercial entities.
Source: Ecology - nature.com
