Our remote sensing analysis demonstrates that banana production in the DR is highly exposed to flood risks, with export banana production being concentrated in the regions of Valverde, Monte Cristi and, to a lesser extent, in Santiago and accounting for 21,561 ha of production area (Fig. 2). The region is dominated by the drainage basin of the Yaque del Norte River that has suffered from severe deforestation over the past decades35, affecting the hydrological regime, and exacerbating the scale of floods at time of heavy rain.
GFVC-actors use diverse strategies to reduce hurricane impacts
From analysis of our workshops and interviews, 18 months after the flood events, we found that DR-UK Banana value chain stakeholders, including smallholder farmers, importers, exporters and the DR government, take a variety of actions in preparation for, response to and recovery from the hurricane induced flooding (Fig. 1). In terms of preparation, smallholder banana farmers reported having a limited range of actions available to reduce the direct impact of the flooding on their farms. For example, actions such as reinforcement of containing walls, to prevent inundation, were perceived to be of low efficacy. Importers and exporters took less direct action related to their activities and performed more co-ordinating actions leading up to the hurricanes. This included avoiding purchasing from high-risk farmers, but in contrast, also working with farmers to reduce flood risk by supporting the establishment of buffer zones near water sources. The Ministries of the Environment and Natural Resources reported taking two key actions in preparation for a hurricane in relation to the banana value chain: preparation of a disaster response plan, and consequently, damage limitation activities involving relocating people from vulnerable areas and dam venting.
Responses following the start of flooding in September 2017 were enacted by these GFVC actors at multiple scales, including farm, watershed, nationally and internationally. Farmers reported taking key damage limitation actions, such as rescue operations for people and livestock, as well as communicating loss of production to buyers. Importers and exporters took two key types of action “switching sourcing location” and “communication” to inform buyers. Government responses focussed on saving lives through rescue operations and provision of shelters.
Analysis of household survey data revealed that the adoption of resilience enhancing strategies was relatively uniform across farmers that directly experienced flooding in 2017 versus those that did not (Suppl. Mat. 1). These included; crop diversification (mean number of crops farmed = 2), intercropping (40% practising), income diversification (41%), training in flood damage prevention (56%) and insurance (23%). Insurance was the only strategy for which there was a significant difference between farmers that were flooded in 2017 and non-flooded farmers, with 36% flooded farmers adopting versus 9% non-flooded (Chi-squared 21.217, p < 0.01). However, insurance adoption of 36% amongst flooded farmers is still relatively low, with farmers citing cost and trust in the scheme as the main concerns. This compares, for example, with 63% of export banana farmers adopting weather insurance in the Windward Islands34.
Unpreventable “all-or-nothing” damage makes recovery key to farmer resilience
Analyses of Synthetic Aperture Radar (SAR) imagery revealed that 2,447 ha, or 11.4% of banana production area in the three regions were affected by hurricane related damage, and largely concentrated around the Yaque del Norte river (Fig. 2). This estimate includes damage caused by open-water flooding in the immediate aftermath of the hurricanes, as well as more protracted storm damage over a period of three months since the hurricanes. In our survey of 158 banana farmers, 80 (51%) reported being directly impacted by the hurricanes. Of these flooded farmers, 75% (60 farmers) reported 90% of their production area flooded (Fig. 3c). This suggests an ‘all-or-nothing’ nature of storm damage, i.e. when farms are affected, damage is complete and catastrophic. No farm scale strategies were reported that were able to prevent this damage.
The flood waters (Photographs in Suppl. Mat. 2) caused the destruction of fruit that was already growing on the plants. For flooded farmers, on average 83% of ongoing production was destroyed. The majority (77%) of banana plants in flooded areas were destroyed during the inundation with water and subsequent submersion period. Observations from surveys are also reflected in our regional-scale remote sensing analyses, where the canopy signature of flooded banana plantation pixels showed a sharp deviation away from values for non-flooded banana plantation pixels immediately after the hurricanes (Fig. 3a), indicating a rapid change in the canopy structure of flooded plantations. Based on the productive farm index (PFI) – a surrogate for the proportion of productive banana plantation pixels – we observed losses of production area continuing until mid-December 2017 (Fig. 3b) before signs of production capacity recovery were detectable. This suggests that the true extent of production area loss is not immediately apparent after the initial hurricane shock, but, accumulated up to three months after the event. Beyond the damage to banana plants there was also significant infrastructure damage, with 18% of farmers experiencing cable ways being destroyed, 15% with packhouse damage, 68% with drainage canals destroyed and 71% having roads on their farms destroyed (Suppl. Mat. 2).
Recovery time is highly variable between farmers
After flooding, there are a key set of activities that farmers reported performing to return banana production to full capacity (Suppl. Mat. 3). These are moderated by natural processes after the flood event, such as the drainage of flood waters and soil aeration. Following this, the farmers and labourers cultivated the field using traction, prepared drains and paths and planted seed material. After the replanting phase, there was a nine-month growth phase before fruits were harvestable and saleable. Concurrently, importers’ and exporters’ recovery process involved assessing losses from existing contracted farmers and then switching their sourcing to other locations not hit by the hurricane within the DR, as well as abroad. Coordination with other suppliers to fill gaps in order fulfilment was also performed by importers and exporters. The government response involved repairing damage to major infrastructure, as well as the provision of financial support to farmers and the purchasing of fruit from farmers that had lost market access.
At the farm-scale, we found large variation in recovery times between farmers (Fig. 4). Recovery times, reported by farmers and considered from an agricultural perspective (marketing aspects are covered in section 2.4, below), covering the time between fields draining and completion of replanting, ranged from two weeks to more than 11 months (min. = 14 days, max. = 343 days, mean = 99 days), with the difference between the slowest quartile and fastest quartile of recovery times being 91 days. The dynamics of the recovery process were significantly affected by delays (Fig. 4), the time between when farmers judged fields were ready to cultivate and when they were effectively able to start. Fifty-three percent of surveyed-farmers that were flooded (42 farmers) experienced delays in replanting. These delays vary between 7 and 352 days, with a mean delay of 35 days. For farmers that experience delays, it on average increased the overall recovery time by 96% and therefore significantly inhibited the recovery process.
Our analyses of farmer recovery data (Suppl. Mat. 4-8) found four factors that influence smallholder recovery times (Fig. 5, Suppl. Mat. 6): (1) Scale of damage (based on: Flooded area, Replanted area, Total banana farm size; b = 0.262, t = 2.12, p < 0.05), (2) farm and livelihood diversity (Agricultural crop diversity, Non-agricultural income diversity, Banana income dependency; b = 0.240, t = 2.21, p < 0.05) and (3) drainage time (b = 0.298, t = 2.99, p < 0.01) all increased recovery time. In contrast, (4) farmer flood training (Flood protection training, Flood recovery training; b = -0.256, t = -2.65, p < 0.01) made recovery quicker. These findings were supported and augmented by qualitative evidence from interviews, with cooperatives and farmers citing flood water drainage, availability of finance to purchase materials and labour for cultivation and replanting, as well as the availability of planting material as major constraints to recovery. We found farmers were supported in several ways, including by the cooperatives, who reported having to take out bank loans using their office buildings as collateral.
While farmer surveys capture recovery in terms of the time required to prepare and then replant farms, remote sensing analyses gave us a clearer picture with respect to trajectory of recovery of regional productive capacity. Based on banana canopy backscatter values from SAR data, and using lenient criteria to define recovery, we found that production recovery (replanting and regrowth) was completed, at the earliest, by June 2018 (Fig. 3a). However, canopy signatures of flooded plantations began tracking that of non-flooded plantations more closely only by late September 2018 – approximately 380 days after the first hurricane (Fig. 3b). Applying more stringent criteria using the PFI, we estimate that the region’s productive capacity returned to pre-hurricane levels by the beginning of December 2018, approximately 450 days after the first hurricane (Fig. 3b). Hence, DR’s banana production system is likely to have seen below capacity production for a period of 15 months due to hurricanes Irma and Maria. Remote sensing analyses also revealed that by 2019, production area had exceeded pre-hurricane area under cultivation by 10.8% (loss of area = 5,048 ha; gain of area = 7,385 ha; net gain = 2,337 ha). However, 26.9% of new plantation area co-occurred at locations that had experienced damage from hurricanes Irma and Maria. This represents a net increase of 2.9% of plantation area at risk from the reoccurrence of a similar extreme weather event (Suppl. Mat 9).
Exposure to global markets leads to spillover of flooding impacts beyond those experiencing direct damage
Recovery is determined by both farmers replanting their crop, and the ability to sell their produce and thus generate income to replenish household and farm assets. This in turn is influenced by responses of the downstream value chain. Interviewed multi-national importers reported switching sourcing to other countries, e.g. Mexico – an emerging region for organic banana production – in the aftermath of the hurricanes. Consequently, 43% of surveyed farmers in the DR reported market inaccessibility in the following year, which impacted both flooded (40% of surveyed farmers) and non-flooded (45%) farmers. However, on average, flooded farmers saw greater reductions in the proportion of production sold to the export market (30% of harvested production) compared to non-flooded farmers (21%). These farm scale results are supported by downstream UK import data which show a rapid decrease in DR banana imports following the hurricane events in 2017 and a limited recovery in the subsequent months (Fig. 6).