How does water-reliant industry affect groundwater systems in coastal Kenya?

• New water-reliant industries are becoming established on the East African

2 in an area supported by a water resources monitoring network. Geochemical models to understand the 22 geochemical processes occurring in the area affected by seawater intrusion are developed. The results

23
show that the present level of groundwater abstraction does not significantly affect aquifer water levels.

24
Recharge under the current scenario has the greatest effect on groundwater level variation. Groundwater 25 abstraction by agriculture and mining exploit a deep aquifer, which is resilient to the drought periods.

26
However, community handpumps are prone to drying-up during drought, as they exploit a shallow aquifer 27 that is less resilient to drought. Future increase in groundwater abstraction during long drought periods is 28 analysed by means of simple calculation methods to understand the limits and interactions between water-29 reliant industry and groundwater sustainability, particularly in predictable future drought events. In 30 conclusion, we argue that in Kenya, as in much of Africa, groundwater risks are not systematically 31 quantified prior to major abstraction. Government and the private sector need to share the risks and 32 responsibilities for improving policy and practice, which currently neither protects the resources 33 underpinning growth nor promotes sustainable development opportunities.

37
The exploitation of groundwater generates different types of negative externalities (Giannoccaro et al.,

54
In the last couple of decades many African countries have seen unprecedented economic growth rates, 55 and this has drawn the region into the global limelight (World Bank, 2013). This industrialization process 56 has led to an overall increase in groundwater abstraction in most African countries (Adelana and 57 MacDonald, 2008). The drilling of new deep boreholes with higher abstraction rates than traditional dug 58 wells or shallow borehole handpumps has increased in many areas to meet the water demands of these 59 new economic activities (Comte et al., 2016).

60
The high socio-economic and ecological importance of groundwater and the fact that groundwater is an 61 important strategic resource are recognised throughout Sub-Saharan Africa (SSA). However, data on 62 groundwater systems are sparse and the current state of knowledge is poor (Pavelic et al., 2012). Most of 63 the time the data that are available are often spatially and temporally inconsistent (Candela et al., 2014) 64 These are serious limitations for the sustainable development of groundwater resources (Gaye and 4 Tindimugaya, 2018). Key aquifers need urgent characterization to change the current situation, in which 66 development proceeds with insufficient aquifer knowledge (Olago, 2018). One of the main challenges 67 when studying these aspects is the lack of information, especially with respect to abstraction data and the 68 location of abstraction well fields, in order to determine the possible future impacts at the local and/or 69 regional scale on groundwater systems.

70
A study by Pavelic et al. (2012) emphasizes that data on groundwater systems throughout SSA is sparse, 71 so the current state of knowledge creates a barrier to sustainable groundwater development. In order to 72 define realistic local management policy it is essential to understand groundwater use and users. One of 73 the major challenges to proper governance is lack of scientific and technical knowledge about aquifers.

74
Without adequate technical understanding of aquifers, actors may not properly identify the source of 75 aquifer pollution or depletion and may be prone to blaming each other for mismanagement (IGRAC, 76 2019). Thus, in the absence of coordinated efforts to manage aquifers, it is unlikely that any advanced 77 technical understanding will be achieved. This paradox is the crux of the groundwater governance 78 challenge and perhaps explains why effective groundwater governance regimes are still elusive today.

79
Therefore, the aim of this study is to assess how increased competition for water may be affecting

99
Protected wells and boreholes are accessed by more than a fifth of the county's populace and 43% of 100 households use an improved drinking water source (Foster and Hope, 2016). There are around 300 101 handpumps providing drinking water to local communities, schools and healthcare centres scattered 102 across the study area. These handpumps are used daily by the population to fill buckets for different 103 purposes, such as drinking and domestic water uses. The coastal strip has a long established coastal 104 tourism industry at Ukunda/Diani. Most of the hotels are located in the coastal area in the north of the 105 study area. Furthermore, the Ukunda area has many private homes that have their own shallow well or 106 borehole. In the last two decades, the acquisition of small parcels of land has increased in this area to 107 build bungalows/maisonettes for which the source of water for construction and supply is often 108 groundwater. The highest tourism season is from October to March and the lowest from April to July.   Currently KISCOL's sugarcane fields occupy a total area of 5500 ha, of which 4100 ha has been put to 121 cultivation of sugarcane since 2008; 800 ha are currently under sub-surface drip irrigation. The fields are 122 located in the Kinondo, Milalani/Shirazi and Nikaphu areas, the last one being located south of the study 123 6 area (Fig. 1). The factory has the capacity to crush 3000 tonnes of cane per day and it is projected to 124 produce 3500 tonnes/day of sugar at full capacity, self-generating 18 MW of electricity in a bagasse-fired 125 power plant, and producing around 50,000 L/day of ethanol (http://www.kwale-group.com). The planned 126 area for irrigated (not rain-fed) sugar at KISCOL is 3000 ha, to be achieved when all dams and the bulk 127 water system (BWS) is completed in the coming years.

128
The study area is divided into 4 zones ( Fig.1) that represent the areas where each economic activity takes 129 place. Zone 1 covers the area where the sugar fields irrigated with groundwater from in-situ boreholes are

134
The area experiences a bimodal rainfall pattern: 1) "long rains" generally fall from April to June (AMJ),

167
The shallow aquifer is directly recharged by local rain through the ground surface and the deep aquifer is    In order to assess the sustainability of the aquifer system during the 2016 La Niña drought, the total 181 recharge to the aquifers of the study area has to be known, which is the main input of the system (Ferrer

184
In that study, groundwater recharge was calculated for each land cover type and each soil type, following

196
In order to study the aquifer recovery after the La Niña event, hitherto not studied, the groundwater level

204
To represent EC evolution in the study area, this information was mapped for each of the seven field

212
In order to understand the geochemical processes occurring in the area affected by seawater intrusion 213 (SWI), a geochemical modelling exercise was carried out to understand the long-term evolution in this 214 geological context and the potential impacts of SWI dynamics. Given the composition of the Pleistocene 215 corals, different geochemical models considering several conceptual hydrogeological models were 216 generated to understand which reactions are taking place, to what extent, under which conditions, and 217 how water quality and aquifer mineralogy could change due to SWI.

218
PHREEQC software was used to simulate the mixing between fresh groundwater located inland in the 219 Pleistocene formation with EC < 1000 µS/cm and one sample from the saline water upwelling on the 220 beach (Diani), which is 83% seawater according to chloride concentration (Table 1

253
Despite hotel abstraction data not having the same degree of accuracy as KISCOL or Base Titanium data, 254 using this methodology it is possible to estimate the order of magnitude of hotel abstraction.

255
The average abstraction of the community handpumps was obtained from Water point Data Transmitters

261
The abstraction of the water-reliant industries will increase in the near future; Base Titanium planned to 262 drill more boreholes within the same wellfield, thus increasing the total groundwater abstracted. In the 263 absence of any better estimates, we arbitrarily assumed a 20% increase in groundwater abstraction for 264 irrigated sugar. The Draft Kwale Water Master Plan has assumed a 1% growth per year in water demand 265 for the tourism sector over the next 20 years. In order to supply more water to the population, the Water

281
Total recharge volume was calculated for an area of 660 km 2 . This area is bigger than the four study 282 Zones (Fig. 1), covering the recharge area of the shallow and deep aquifers from the sea to the Shimba 283 Hills (Ferrer et al., 2019). While the shallow aquifer is recharged directly from the surface, the underlying 284 deep aquifer is recharged from the Shimba Hills. To estimate recharge across the study area, 123 soil 285 water balances were calculated (Fig. 3).

286
The spatial distribution of recharge follows the rainfall spatial pattern. Higher recharge occurs near the   (Table 2) shows that there is 295 minimal correlation between total annual rainfall and total annual recharge. This is because the rainfall 296 intensity and distribution through the year influences net recharge, rather than the total annual volume of 297 rainfall. High rainfall peaks produced by intense but short storms are more effective in driving recharge 298 than lower, more continuous rainfall. An intense rainfall event (>100mm) on a saturated catchment leads 299 to intense and significant recharge. This is consistent with other studies on the phenomenon (Taylor et al.,      (Table 3). After the La Niña event, the daily average 319 abstraction fell by around 26% (3370 m 3 /d) in 2017, compared to 2016 (Fig. 4). It should be pointed out 320 that Base Titanium recycles a considerable proportion of process water: in 2016, it recycled >70% of the 321 total daily water use. It improved in 2017, recycling around 78%.

322
The mine site is located on the Pliocene formation but the Base wellfield is on the Kilindini sands 323 (Pleistocene) east of the mine (Fig. 2). These production wells are screened in the deep aquifer, to ensure

363
together with water coming from the dams, which is the other water source for sugar irrigation.
Groundwater acts as a strategic water reserve; volumes used are small compared with water from dams.

365
According to current WRA rules, the maximum volume that may be pumped is 60% of the well test 366 discharge rate over a ten-hour pumping day. It means that the mean estimated abstraction rate for the 367 eight KISCOL boreholes is 531 m 3 /day (Table 3). This value is in accordance with the KISCOL test 368 yields for these eight boreholes and it is in the same range as other unpublished data from KISCOL.

369
KISCOL wells are multi-screened, taking water from multiple water-bearing zones in the shallow and

389
From the data obtained from Google Earth, 85% of the hotels located on the coast are located in Zones 3 390 and 4 on the Diani coast, with only a few situated on the Msambweni coast in Zones 1 and 2 (Fig. 1).

391
The highest tourism season is from October to March and the lowest from April to July. Hotel water use

403
We estimated hotel groundwater abstraction according to the different type of data source (WRA 404 allocations and hotel interviews). We also estimated hotel groundwater abstraction using the number of 405 rooms and the hotel class type.

446
During the La Niña event, there was a groundwater level decline in 86% of the measured shallow wells.

447
In the remaining wells, the groundwater levels were nearly constant. However, levels in 95% of the wells 448 affected by the La Niña drawdown recovered after the first rainy season (AMJ) in 2017 (Table. 2S 449 Supplementary material). In this regard, the first rainy season (AMJ) is more effective in the recovery of 450 the groundwater system than the short rains (OND). correspond to a saline geological formation, the Maji ya Chumvi beds (Caswell, 1953) 474 We compared hydrogeochemical modelling results with the samples from wells/boreholes affected by 475 SWI in the shallow aquifer to understand the importance of the SWI change. Field samples contain 476 between 0% and 30% of seawater (Fig. 8a), except for the sample taken from a beach upwelling, which 477 had 83% seawater. Of the conceptual models tested (data not shown), the one that gives results closest to 478 24 the observed field samples is the mixing of fresh and saline water, both in equilibrium with calcite (i.e. 479 Fig. 8a).

480
Looking at the delta ion evolution for calcite (total quantity of precipitated/dissolved calcite mineral) in 481 this conceptual model, during mixing between fresh groundwater and saline water (Fig. 8b)

497
Not all water users exploit the same aquifers (formations). The community wells, handpumps and hotels 498 mainly abstract groundwater from the shallow aquifer. The recharge areas of this aquifer unit are those 499 exhibiting more volume variation between drought and a normal climatic year (i.e. 2017) (Fig. 3). The

500
shallow aquifer unit is less resilient to climate variation than the deep one. This explains why some wells 501 located in the Kilindini and Magarini sands became dry during the La Niña drought. The aquifer system 502 exhibited swift recovery after the first normal rainy season in 2017. This allowed the system to return to 503 the average groundwater budget and to face the next drought period.

504
One consequence of wells becoming dry is the increase in walking distance to collect water. As reported 505 during fieldwork in June 2016, during the La Niña event, some communities stated that they had to walk 506 longer distances to collect water because the nearest borehole or well was dry. Amongst other impacts,

518
Unlike communities and hotels, Base Titanium exploits the deep aquifer and KISCOL both aquifer units.

519
The fact that the recharge variation is less in the Shimba Hills than in the lowland means that the deep 520 aquifer is more resilient to drought events. This favours groundwater abstraction by these users, since 521 they can continue to exploit the deep aquifer during periods of drought lasting at least as long as the last 522 La Niña event.

523
Focusing on mining, the abstraction rate depends on rainfall patterns, increasing during the dry period in

555
In the coastal areas with major tourism concentrations, recent years' data seem to show a local 556 salinization effect in the shallow aquifer due to the higher abstraction induced by tourism and associated 557 activities. However, a longer period of observation is needed to determine the saline water intrusion 558 dynamics in order to consider rainfall fluctuations and to differentiate between seasonal effects (still 559 unclear but possible) and long-term trends. As the zone with most of the hotels is also the area with the  Total groundwater abstraction is expected to increase by a factor of four over the current rate (see Table   570 3). This level of abstraction would represent 22% of the total recharge occurring during the La Niña new areas. The significance of SWI is that only 2 to 3% of seawater mixed with fresh water is enough to 603 make the resulting water useless for most purposes.

604
The calculation of the freshwater-saltwater mixing zone is a complex task, but an approximation can be

613
This advance of the saline wedge could affect hotel groundwater supply and community handpumps 614 located near the coast.

615
We also calculated the saline wedge depth for different distances from the coast and for the different 616 geologies near the coast, i.e. the Pleistocene corals and Kilindini sands (data in Supplementary material).

617
The results show that even during the future abstraction scenario during a drought year like the 2016 La

624
The future consequences of borehole salinization would be an increasingly salty taste and at some point

662
This study uses simple calculations to illustrate the possible future risks of increased abstraction under 663 climate stress to an aquifer system in Kwale County. At present, under 'normal' climatic conditions, we 664 have observed no adverse consequences in the aquifer system since major abstraction started in 2012.

665
However, the study underlines the importance of evaluating all risks to any aquifer system prior to major 666 groundwater abstraction.

689
Future risk should therefore be predicted under different abstraction future scenarios, before major 690 abstraction takes place.

691
While gambling with groundwater may be common in Africa and globally, this study shows that 692 groundwater resources can be significant and resilient to unpredictable but recurrent drought events.

693
Given over half a billion dollars in capital investment in the two water-reliant industries in Kwale, in 694 addition to tourism and related investment, understanding investor risk and liability from groundwater 695 sustainability would seem prudent, if not a legal obligation, before major abstraction starts. Government 696 leadership is essential to manage the aquifer as a system for all, including environmental services, rather 697 than for the powerful few. Without technical, material and political support, water resource management 698 agencies face stark choices in Africa as limited staff and capacity are unable to ensure that adequate 699 monitoring systems exist to guide regulations that manage water resources in the public interest.

700
Governance failure can promote market failure, by mismanaging groundwater, by design or by accident.

701
However, this is not inevitable and we see evidence of good corporate water management as a catalyst for 702 providing critical deep aquifer data to inform a credible model for future groundwater management and 703 resource allocation in Kwale. Furthermore, using often simple information sources (interviews, Google

704
Earth, Trip Advisor, basic analytical methods, etc.), enables groundwater abstraction to be estimated,

705
allowing potential future risks to be assessed as has been shown by this study.

33
We strongly recommend compliance with existing regulations and codes of practice and strengthening of 707 the capacity of the Water Resources Authority to monitor such compliance. Further, in line with the 708 current updating of the national water policy, we advocate for improvement of the legal framework and 709 regulatory reach so as to encompass the notion of shared risks and responsibilities among the government 710 and the private sector, and creation of a business climate that espouses resource protection for growth and 711 promotion of sustainable development.