Petrology and geochemistry of selected nepheline syenites from Malawi and their potential as alternative potash sources

The Sub Saharan Africa agricultural sector is one of the most disadvantaged regions, partly due to high fertiliser import costs from the northern hemisphere. Malawi is one such country which faces these fertiliser challenges for the agricultural sector growth and food crop production. However, Malawi has numerous intrusive alkaline rocks, nepheline syenites, especially within the Chilwa alkaline province. This study was therefore conducted to assess these nepheline syenites for their potential as potassium sources. We used Malawi's new airborne geo- physical gamma ray data acquired in 2013, coupled with satellite remote sensing, to identify nepheline syenites suitable as possible sources for alternative silicate K-fertiliser, and carried out geochemical analysis of whole rock samples. Results show that the K 2 O content for the nepheline syenites varies from 3.17 wt % to 9.14 wt % with an average of 5.22 wt %. The K 2 O/Na 2 O ratio for Malawi's nepheline syenites ranges from 0.41 to 1.28 with an average of 0.65 showing that the nepheline syenites are mostly sodic but with variable composition. In addition to nepheline, the calcium feldspathoid davidsmithite ((Na,Ca)AlSiO 4 ) was identified in the syenites using scanning electron microscopy with energy dispersive analysis. Although the different intrusive complexes are not homogenous, our results show that, generally, the nepheline syenites from Malawi have similar geo- chemistry and mineralogical composition to those which have been used as crushed-rock fertilisers in other parts of the world.


Introduction
The global community faces several major challenges concerning food security. One of these is the cost of conventional fertiliser, particularly potassium (commonly called potash; K 2 O), which is currently so expensive that it is inaccessible to many farmers, especially in Africa, on the grounds of purchase price and transport costs (Fuentes, 2013). By 2020, without any increase in fertiliser use and amidst increased crop production, leading to soil nutrient deficiency, annual depletion rates of potassium in Africa will likely increase to 36 kg ha −1 K (Sheldrick and Lingard, 2004). However, alternatives are available, which may help farmers to replenish the potassium removed by crops (Manning, 2015(Manning, , 2017, including the use of nepheline syenites (Goldschmidt, 1922;Jena et al., 2014).

Situational context
By 2004, all but only four African countries (Botswana, Namibia, Somalia and Niger) were K deficient (Sheldrick et al., 2002). Because most potassium fertiliser is imported from the northern hemisphere, costs are largely determined by the import and transport costs. Other additional costs are associated with fertiliser distribution within the country as well as the trader and agro-dealer margins (Chirwa and Dorward, 2013;Fuentes, 2013). The situation has worsened in the last decade due to the global economic recession, fertiliser price adjustments and increasing poverty levels in Africa. An average African farmer pays for potassium fertiliser nearly twice as much as their counterparts in Europe and America. This is partly because the First World dominates fertiliser production (Roberts and Vilakazi, 2014). Malawi, in particular, faces severe fertiliser challenges for her agricultural sector growth. The Government introduced the farm input fertiliser subsidy program (FISP) in early 2000s to help vulnerable https://doi.org/10.1016/j.jafrearsci.2020.103769 Received 20 July 2019; Received in revised form 13 January 2020; Accepted 21 January 2020 farmers. The scheme faces challenges including corruption, dependency on donor funding (Vinet and Zhedanov, 2010) and a long supply chain (Fig. 1). Two key state-owned local fertiliser suppliers are the Smallholder Farmers Fertiliser Revolving Fund of Malawi (SFFRM) and the Agricultural Development and Marketing Corporation (ADMARC).
Previous work has shown that crushed rocks offer a viable alternative to chemical fertilisers, especially in highly weathered and leached environments (Harley and Gilkes, 2000;Theodoro and Leonardos, 2006; van Straaten, 2007;Priyono and Gilkes, 2008;Mohammed et al., 2015;Ciceri et al., 2015;Gupta, 2015;Manning, 2010Manning, , 2017. Use of locally available crushed rocks, termed 'remineralizers', for agriculture in tropical soils is established in Brazil. In Brazil, it is particularly attractive to small farmers who produce food (especially horticultural crops) for their own consumption and for domestic markets (Manning and Theodoro, 2018). Crushed rocks release nutrients more slowly than chemical fertilisers (Harley and Gilkes, 2000), which are vulnerable to fast removal due to leaching and erosion. A number of studies have shown that rocks containing biotite and nepheline demonstrably provide potassium for plant uptake (Gautneb and Bakken, 1995;Bakken et al., 2000;Mohammed et al., 2015). Although other silicates containing potassium are common rock-forming minerals, the nutrient release from these minerals, such as feldspars, is slow compared to chemical fertilizers such as KCl (Priyono and Gilkes, 2008;Manning, 2010;Mohammed et al., 2015). This deters their use except where particular soil requirements are favorable and where rock powder properties have been improved to required standards (Harley and Gilkes, 2000).
The suitability of silicate rocks as alternative source for potassium does not depend on their absolute content but rather the dissolution rates of their potassium-bearing minerals (Manning, 2010). Although potassium feldspars have greater absolute contents of potassium than nepheline, their suitability as a source of K is limited (Priyono and Gilkes, 2008). This can be attributed to their slow dissolution rates. Other experimental studies suggest that the dissolution of silicate minerals, particularly nepheline, is largely influenced by the acidic conditions in the soil (Priyono and Gilkes, 2008;Jena et al., 2014). Dissolution rates also depend on the minerals' surface area, and can therefore be enhanced by grinding or milling, which means adding extra capital and production costs.
Few plant growth trials have been conducted using nepheline syenite compared to potassium feldspar (Manning, 2010;Manning et al., 2017). Bakken et al. (1997) carried out field trials with crushed rock containing orthoclase, nepheline and biotite, and mine tailings from nepheline syenite production at North Cape (Norway), to assess their potential to release potassium to support Italian ryegrass growth over a six months period. The plant growth was highest from nepheline followed by biotite, then orthoclase and mine tailings, suggesting fastest potassium release from nepheline. This also suggests that there was slow breakdown and dissolution from the tailings, hence the inability of the plants to extract sufficient nutrients from them. While agreeing that biotite release is higher than most soluble potassium sources, Bakken et al. (1997) reported that there was a slower release for biotite sources compared to nepheline. These findings agree with experimentally determined dissolution rates for silicate minerals as reported by Palandri and Kharaka (2004), which found the highest dissolution rates in nepheline (Table 1). Pessoa et al. (2015) also assessed the solubility of nepheline syenite as a function of organic matrices, using coffee husk in Brazil. Results showed that regardless of potassium content, when a nepheline syeniteorganic mixture was incubated, potassium solubility was high, especially when extracted with 2% citric acid compared with water. The physical properties, notably particle size and surface area of the potassium-bearing silicate minerals (Priyono and Gilkes, 2008;Mohammed et al., 2015), play an added role in accelerating their dissolution and suitability as potassium sources. The use of nepheline syenites as an alternative to K-feldspars is also supported by studies on Colombian Savanna native grasses (Brachiaria dyctioneura) and the legume Pueraria phaseoloides (Gautneb and Bakken, 1995). In their studies of these grasses' dry matter yield, Gautneb and Bakken (1995) and Sanz Scovino and Rowell (1988) found that only about 10% of the feldspar's K was absorbed by plants in 14 months, compared to 25-68% of KC1 uptake from the crops in the same period. Since the dissolution rate is also dependent on temperature (Harley and Gilkes, 2000), nepheline syenites would be suited in tropical areas. The high tropical temperatures, unlike Norway and other cold areas where the rocks have A. G. Chiwona, et al. Journal of African Earth Sciences 164 (2020) 103769 been tested (Bakken et al., 1997), would support more rapid dissolution of the nepheline. This study compares, for the first time, the geochemical composition of nepheline syenites from Malawi with examples used elsewhere in studies of the potential of nepheline-bearing rocks as a source of K. Given the wider occurrence of nepheline syenites within the East African Rift System, the results have broader application. At present, most nepheline syenite intrusions, especially in Africa, have not been exploited for this role. More widely, our aim is to demonstrate that airborne geophysical gamma ray data coupled with petrological and geochemical analysis of rock samples collected in the field can be used to identify and map nepheline syenites suitable for consideration as sources for potassium as fertiliser. Fig. 2 shows the areas considered in this study. Fig. 2 also shows that some intrusions which were identified as possible nepheline syenites or syenites that had not been documented as alkaline rocks prior to this study. The nepheline syenites and alkaline rocks of Malawi are distributed all over the country (Woolley, 2001) although most of them occur largely in the Chilwa Alkaline Province (CAP) of Early Jurassic to Late Cretaceous age. The CAP largely comprises plutonic rocks in the form of syenites, nepheline syenites, carbonatites, and some volcanic vents associated with carbonatite bodies, feldspathic rocks, breccias and agglomerates outcrop in a number of parts of the country (Mshali, 2009).

Regional geological setting of the study area
Malawi lies within the Mozambican Mobile Orogenic Belt, which is associated with reworked meta-volcanic and meta-sedimentary rocks of Late Precambrian to Early Palaeozoic age (Carter and Bennet, 1973;Mshali, 2009), locally known as the Malawi Basement Complex (MBC). The MBC has been affected by three orogenic episodes, namely the "Ubendian" cycle (2100( -1950, the "Irumide" cycle (1600-900 Ma) and the "Mozambican" cycle (700-400 Ma). The Malawi Rift is part of the Miocene-Quaternary East African Rift System (EARS). This is associated with faulting and the linear graben that covers most of east of the country (Thatcher and Walter, 1968;Mshali, 2009). Fig. 3 shows a simplified regional geological map of the study area.
There is little published information about the petrology and geochemistry of these nepheline syenites. Therefore, this paper is important because it provides the much-needed information on the distribution of Malawi's nepheline syenites, permitting assessment of their potential as a K silicate fertiliser. The sampled areas in Kasungu (Fig. 3 A) occur in the central region. The other areas fall within the Chilwa Alkaline Province, S. Malawi.

Data used in the study
In order to label these rocks as possible nepheline syenite targets, airborne gamma-ray spectrometry data and partly ASTER digital terrain models were used. Malawi's most recent countrywide airborne gammaray spectrometry data were acquired by Sanders Geophysics Limited as part of Malawi's countrywide geophysical mapping programme, using the Exploranium GR-820 gamma-ray spectrometers (Bates and Mechenneff, 2013). The pixel size on the ground depends on the number of samples collected per second by the sensors. Most airborne geophysical surveys are conducted at a sampling rate of 1 count per second (1 Hz; Bates and Mechenneff, 2013) which is equivalent to 50-80 m pixel size on the ground (Horsfall, 1997;Beamish, 2014). Malawi's airborne geophysical gamma-ray data was acquired at a line spacing of 250 m and 60 m flying height (Bates and Mechenneff, 2013) and was gridded at 50 m grid cell size (Bates and Mechenneff, 2013).

Identification of potential areas for sampling
Data for petrological and geochemical analyses were acquired from the fieldwork conducted by the authors in key locations in Malawi. The fieldwork survey areas were selected largely based on results from gamma-ray airborne geophysics, remote sensing data (especially digital terrain models), and review of previous geological information of Malawi. The potential nepheline syenite areas were selected for fieldwork based on results from the gamma-ray spectrometry processing and analysis. This was done using individual total count maps of three radiometric channels, namely uranium (U), thorium (Th) and potassium (K) and ternary composite maps. Geochemically, nepheline syenites show high-K content relative to Th and U (Tye et al., 2017). On ternary plots, the potential nepheline syenites and related rocks were those which showed high potassium content. The ideal areas for fieldwork, were therefore, those with high K, relative to U and Th. This was best reflected on the gamma-ray ternary plots because they combine the three radiometric channels and highlight areas where each of the channels (K, U, Th) is highest relative to the other two. Nepheline syenites, carbonatites and other alkaline rocks are usually associated with ring structures, lineaments and clusters (Jaireth et al., 2014;Woolley and Kjarsgaard, 2008). Potential areas were, therefore, those whose results were those with high K content (between 3.43 wt% to 7.52 wt%) on the K gamma ray total count maps. The high potential areas were selected for ground follow-up, if they had higher K content relative to Th and U on the gamma ray ternary maps and showed ring structures and/or clustering on digital terrain models (DTMs).

Field sampling
Thirteen areas were surveyed during the fieldwork. Two areas were from central Malawi (9 and 10), three from the carbonatite associated areas (32,33 and 35), four from southern Malawi (25-28) and three from south-east Malawi (undocumented) in Fig. 2. During the survey, different rock types were sampled. Emphasis was on nepheline syenites and related alkaline rocks to assess their potential as potassium silicate fertiliser sources.

Ground gamma ray geophysics
Field gamma-ray data were collected using an RS 125 gamma-ray field spectrometer manufactured by Terraplus Inc. This instrument was Table 1 Silicate minerals and their dissolution rates (after Palandri and Kharaka, 2004 used to obtain K, U and Th concentrations from different rocks in the areas surveyed. The spectrometer has assay mode readout for K in %, and U and Th in ppm (Madi et al., 2014). We used the Total Count readout at 1×/sec rate in the Survey Mode (www.terraplus.com/ products/pdf/rs-125-portalbe-super-gamma-ray-spectrometer.pdf) and the sample time was set at 60 s.

Petrography
A detailed petrographic study of the samples was carried out on thin sections prepared by the Sample Preparation Facility of the School of Geosciences, The University of Edinburgh. Samples were studied using a Nikon Eclipse e200 conventional petrographic microscope and the Zeiss SIGMA HD VP Field Emission scanning electron microscope (SEM) hosted at Edinburgh University (https://www.ed.ac.uk/geosciences/ facilities/sem/specification). Semi-quantitative mineral chemistry in the samples was measured using the Oxford AZtec energy dispersive spectrometer (EDS) fitted to the system, calibrated on a cobalt standard.

X-ray ray fluorescence (XRF) analysis
Major and trace elements were analysed using the Panalytical PW2404 X-Ray fluorescence (XRF) instrument hosted at the School of Geosciences, The University of Edinburgh (https://www.ed.ac.uk/  (2001)). The intrusions indicated with asterisk (*) are known nepheline syenites while those indicated with double stars (**) comprise both carbonatite and nepheline syenite.
geosciences/facilities/xrayfluorescence). The rock samples were firstly finely ground using a tungsten carbide grinding mill and an agate mill. While both mills were used to produce a fine powder from the rock samples, the tungsten carbide mill was needed for grinding very hard rock chips to a fine powder of around 120 μm. Major-element concentrations were measured on 40 mm-diameter fused glass discs; about 0.9 g of sample powder was mixed with a borate flux using a 5:1 (flux: sample) dilution procedure Gill (2014). Thereafter, the samples were fused and heated in Pt-5% Au crucibles at 1100°C. The trace element concentrations were measured on pressed pellets with~8 g of powder used to make a 40 mm-diameter pellets.

Weathering intensity of the rocks
The Parker chemical alteration index (CIA) and the plagioclase alteration index (PIA) (Nordt and Driese, 2010;Meunier et al., 2013;Mohanty et al., 2016) were used to determine the alteration states of the rocks and whether K 2 O and nepheline abundance could be related to this. The CIA degree of alteration ranges from 0 to 100. The optimum index value for fresh/less altered rock is < 50 whereas 100 is the maximum index value for 'complete' alteration (Price and Velbel, 2003). The PIA focusses more on plagioclase alteration. Using molecular proportions of elemental oxides the two indices are calculated (Price and Velbel, 2003) as: (3)

Geology and field observations of the surveyed areas
The samples of nepheline syenites, syenites and other rock types were collected from the different areas that were surveyed. As shown in Fig. 2, the areas were grouped in clusters based on the locations where they occur. Fig. 4(A-D) shows the general geology of four of clusters which were surveyed. The maps in Fig. 4 are only for clusters which included at least a nepheline syenite or syenite intrusion.

Central Malawi nepheline syenites (Fig. 4 (A))
The central Malawi unit comprises the Kasungu and Kasungu-Chipala nepheline syenites (areas 1 and 2 respectively in Fig. 4A), which were intruded in a suite of medium to high-grade metamorphic rocks of the Mozambique Orogenic Belt, dated~500 Ma (Eby et al., 1998). In hand specimens, nepheline syenites from both intrusions are medium to coarse-grained and contain xenolithic inclusions of gneisses and diorites. Mesocratic, very coarse-grained nepheline syenites, which grade into syeno-granites, occur sparsely. The rocks are variably weathered; lichens and small herbaceous plants are abundant.
The Kasungu-Chipala nepheline syenites (2 in Fig. 4A), are in some localities bounded by diorites, although the contact zone between these rock types is not clear (Peters, 1969). Localized faulting with some heavily folded biotite schist and fenites along the nepheline syenitediorite contact zone occur in few locations. This may suggest a contact metasomatism event, which might have preceded micro-faulting and later quartzitic vein development (Peters, 1969).

South East Malawi quartz syenites and Mangochi hill syenite (Fig. 4 (B))
The south-eastern quartz syenites and syenite include the Nkhuzi Bay and Mauni intrusions ( Fig. 4B; 1 and 2 respectively). These are part of a chain of some NNE trending oval-shaped undulating hills located on the western side of Mangochi town. The Nkhuzi Bay area is characterized by coarse-grained-mesocratic-weathered quartz syenite. The weathered rocks are more altered and show a pale colour while the fresher specimens are more pinkish. King and Dawson (1976) reported that the quartz syenites in this area occur together with syenites but during this fieldwork only quartz syenites were observed. In hand specimens, the quartz syenites show more K-feldspars (≥40%), plagioclase (≤25%), quartz (≥15%), muscovite, biotite (10%), minor occurrences of hornblende plus other unidentified minerals.
The Mangochi Hill syenite ( Fig. 4 B (4)), which is located further north-east of the Junguni intrusion, shows similar mineralogy to the Zomba syenite although the Mangochi Hill syenite has more K-feldspar. The Mangochi Hill syenites are coarse to medium-grained and mesocratic. Hand specimens of the rocks mainly show K-feldspars (≥50%), plagioclase and micas.
The Chinduzi and Chikala intrusions (area 2 and 5 respectively, in Fig. 4C) were not surveyed during the present fieldwork. The Chaone ring structure ("4" in Fig. 4C), comprises coarse grained, leucocraticmesocratic, nepheline syenite bounded by orthogneisses. The nepheline syenites are largely weathered, with lichen and moss growth evident on some of the outcrops. In some localities, the nepheline syenites have inclusions of diorite xenoliths, which shows that the nepheline syenites are younger. The alkali granites/syenogranites probably occur within contact zones of the gneisses and nepheline syenites. This shows the possible interaction between the gneisses and nepheline syenites. The Mongolowe intrusion sits in the middle-western part of the Chinduzi-Mongolowe-Chaone-Chikala structural chain of igneous intrusions. The intrusion mainly comprises medium-coarse-grained, mesocratic weathered nepheline syenite rocks. Some outcrops are heavily weathered and show coarse-grained biotite and muscovite mica.
The Junguni nepheline syenite ("6" in Fig. 4C) is a horseshoe-shaped 2.5 km diameter intrusion, situated about five km north of the Chikala-Mongolowe hills (Woolley, 2015). It comprises coarse-medium grained mesocratic nepheline syenites with K-feldspar, nepheline, biotite and pyroxene. The grain size increases uphill and field gamma-ray measurements for K 2 O values also tend to be higher in the southward direction and toward the summit of the intrusion. The Zomba Mountain ("1" in Fig. 4C) is predominantly a syenite intrusion, which also has other rocks including quartz syenites and charnockitic gneisses.
The Songwe-Mauze complex ("3" in Fig. 4D), contains fine-grained, mesocratic to light-reddish, highly weathered carbonatite rocks, which are localized on the Songwe-Mauze area. The nepheline syenites occur east of the Songwe-Mauze Hill and on Mauze Hill. The complex is also characterized by fenites, which occur mostly along the carbonatite and nepheline syenite contact zones. The fenites are fine-medium grained, mesocratic to light reddish weathered rocks consisting of calcite and quartz, with some mafic minerals banded with orthoclase and

Table 2
Description of rocks' hand specimens based on field observations.    plagioclase. The fenites which occur close to the Mauze nepheline syenite intrusion are dark-coloured suggesting nepheline syenite metasomatism while those closer to or on the edges of Songwe-Mauze Hill carbonatite are light-reddish coloured suggesting carbonatite metasomatism with the country rocks or carbonatite-nepheline syenite interaction (Swinden and Hall, 2012). These authors have also argued that mineralisation in this complex is associated with potassic fenitisation alteration and low temperature hydrothermal/carbohydrothermal secondary alteration. The occurrence of these carbonatites and nepheline syenites may suggest carbonatite-nepheline syenite magma liquid immiscibility (Robb, 2005). Field spectrometry found that in these three complexes fenites in the carbonatite complexes are more potassic than the nepheline syenites.

General description of the rocks in hand-specimens
The nepheline syenite samples collected from the field show differences in colour, grain size, texture and mineral compositions. In hand specimen, nepheline crystals are not easily identifiable in most of the nepheline syenites. Weathering varied in intensity and samples were taken from outcrops for which visible weathering was limited to a thin (< 1 cm) zone. Fig. 5(A-D) shows typical photographs of the outcrops from which rock samples were collected. Table 2 provides a summary of the rocks based on the observations in the field. For the purpose of grain size classification, the hand specimens were classified as fine-grained (< 1 mm), medium-grained (< 3 mm) and coarse-grained (> 3 mm) according to Gill (2010). The mineral proportions are classified and given in their order of abundance.

Petrography
Twenty-three thin sections were studied for their minerals. In most nepheline syenite thin sections, nepheline is etched around the grain boundaries and shows alteration. Orthoclase and plagioclase also show alteration to clays. Fig. 6 shows examples of thin sections for selected rocks, and Fig. 7 shows back-scatter electron images. Nepheline is clearly identifiable in the Tundulu nepheline syenite (Figs. 6B, 7A and 7B). Some opaque minerals also are also present in some highly altered rocks such as Junguni (Fig. 6C). In terms of composition, modal Table 3 Composition, mode of minerals and brief description of selected rock samples in thin sections.  Fig. 4B). Abundance of microcline suggests exsolution of albite (Na-rich) out of a K-feldspar host.
Nepheline syenite nepheline ranges from 10 to 35% in the nepheline syenite thin sections as shown in Fig. 6B, Table 3. In some foid-bearing syenites and nepheline syenites, a mineral first taken to be quartz appears yellower than the normal first-order interference colour quartz (Fig. 6D). If this were due to the thin section thickness, adjacent feldspars would also show similar interference colours. SEM analysis showed that this high birefringence mineral might be davidsmithite, a Ca-bearing nepheline (see later). Composition of most rocks in thin sections (as shown in Table 3) confirm the field observations of the rocks' mineralogy in hand specimens (shown in Table 2). Confirming the petrographic observations using EDS, the key minerals in the nepheline syenites include anorthoclase, albite, actinolite, nepheline, orthoclase, sanidine, sphene, fluorapatite, calcite, biotite and iron oxide. Fig. 8 shows the K/Na bivariate x-y graph for atomic proportions of Na and K, calculated from the EDS output, for the Kfeldspars and nepheline. While alkali feldspars show a continuous range, nepheline clusters around (K 0.25 Na 0.75 ) AlSiO 4 .
As shown in Fig. 8, and Table 4, some of the nepheline syenites (from the EDS analyses) also appear to contain a calcium-rich nepheline (Na,Ca)AlSiO 4 or (KNa) 8 CaAl 8 Si 8 O 32 , whose occurrence has not been previously reported in Malawi. We suspect this mineral to be davidsmithite, an uncommon silicate mineral of the nepheline group. Davidsmithite is associated with the heterovalent replacement of Ca 2+ for K + (and is therefore K-deficient; Kechid et al., 2017;Rossi et al., 1989), whose occurrence has been reported in few areas.

Major elements
The geochemical analytical results (Table 5) show that Malawi's nepheline syenites and syenites are heterogeneous. There are some marked differences between and among the rocks from the different localities. The whole rock geochemistry shows that the rocks are generally silica-undersaturated and their SiO 2 content ranges from 44.6 wt % to 69.57 wt% with an average of 57.69 wt%. They also have low CaO content (0.56 wt% to 6.86 wt%, average 2.79 wt %), MgO (0.07 wt % to 2.78 wt %, average 1.23 wt %), TiO 2 (0.25 wt % to 2.05 wt % average 1.18 wt %), very low MnO (0.16-0.27 wt %, average 0.17 wt %) and P 2 O 5 (0.03 wt% to 0.80 wt%, average 0.40).
The 16 samples which showed normative nepheline plot as nepheline syenites on the TAS plot for intrusive rocks (Wilson, 1989 , Fig. 9). Based on Shand's (1943) alkalinity index, the rocks analysed in this study are mostly metaluminous with Al-(KNaCa) values ranges of between −205 up to 0. This is consistent with the presence of biotite and augite among the modal minerals.
Oxides versus SiO 2 variation diagrams are presented in Fig. 10. From this figure, some of the intrusions show similar geochemical composition. The central Malawi nepheline syenites are the least potassic of all the clusters but they have high TiO 2 and FeO contents. Geochemical analyses reported by Eby et al. (1998) and this study for the central Malawi nepheline syenites show similar results.
The R 1 -R 2 classification scheme for intrusive rocks (De la Roche, Leterrier, Granclaude and Marchal, 1980) shows that most of the rocks are nepheline syenites (Fig. 11 and Table 7). Based on this classification, two major groups of nepheline syenites can be identified, classified as group A and group B (Fig. 11). Group A, which includes the Tundulu, Songwe-Mauze, and Junguni and Mongolowe intrusions, has a higher nepheline and potassium content.
However, the nepheline syenites are heterogeneous and may be classified into three main categories: (I) the central Malawi nepheline Table 4 Analyses for selected nepheline and number of cations for nepheline based on 4 oxygens for Kasungu (KU) and Songwe-Mauze (SONG) intrusions.       Group B is slightly silica-rich (Fig. 11) and comprises mostly the Central Malawi nepheline syenites. The carbonatite-associated nepheline syenites plot in group A on the R1-R2 but some plot into B on the TiO 2 vs P 2 O 5 plot (Alle, 2007). The K 2 O/Na 2 O ratio for the nepheline syenites varies from 0.41 to 1.28 with an average of 0.65, which shows the nepheline syenites are more sodic than potassic, with variable K 2 O and Na 2 O content. The K 2 O content for all the rock units varies from 3.17 wt % to 9.14 wt % with an average of 5.22 wt %. The K 2 O content for nepheline syenites only ranges from 3.17 wt% to 11.32 wt % with an average of 5.22 wt %. The total alkali content for the nepheline syenites ranges from 9.62 wt% to 17.77 wt% with an average of 13.26 wt% while the average content for all the rock units is 11.92 wt%.

Trace elements
The nepheline syenite and related rocks from Malawi contain varying amounts of trace elements (Table 5). The South Malawi and carbonatite associated intrusions tend to be more enriched in LREEs in contrast with the rocks from the north and central areas of Malawi. The trace element concentration is typical of alkaline rocks with an abundance of large ion lithophile (LIL) elements such as Rb, Sr, Ba, Nb, Ta, Th compared to the high field strength REEs (Fig. 12). Fig. 12 shows an abundance of Ba and Rb which suggests replacement of K in K-feldspar, micas or hornblende (Deer et al., 1982). The abundance of Rb in some rocks, notably the carbonatite-associated Northern Malawi and the Central Malawi nepheline syenites, could be attributed to the chemical similarities of K and Rb. This enables cations of elements like Rb to be substituted for K in the alkali feldspar structures during crystallisation of the magmatic rocks, and Ba is further enriched by weathering of these rocks (Ollila et al., 2014). Some elements may be toxic above certain concentrations, which makes nepheline syenites with low concentrations, if used as fertilizers, much safer for humans and the environment. Fig. 13 shows variations in terms of selected trace elements, which could be important for agriculture, versus relation to K 2 O content in the rocks. All decrease with increasing K content. The rocks have low amounts of the elements except the central Malawi nepheline syenites, which have comparatively high U and Th values.

Weathering intensity of the rocks
Based on the Parker's chemical alteration index (CIA) scale, almost all the rocks were relatively fresh with their CIA < 50 ( Fig. 14 and Table 7). The Tundulu nepheline syenite was the least altered sample whereas the Nkhuzi Bay, Chaone and Kasungu nepheline syenites had some of the most altered rocks. The results show almost no association between the chemical index of alteration and nepheline abundance in the rocks. This suggests that nepheline concentration is not related to the intensity of weathering. However, a weak positive association is noted between the chemical index of alteration and normative orthoclase. Fig. 9. Total silica alkali (TAS) for the nepheline syenites and other alkaline rocks from Malawi (after Cox et al. (1979) and modified by Wilson (1989).

Comparison of XRF results with airborne and ground gamma ray spectrometry
Some similarities and differences were noted in the data acquired from the airborne and ground spectrometry compared to laboratory XRF analyses (Table 8). The XRF K 2 O values vary from 3.17 to 11.32 wt% (average 5.22 wt %). The airborne gamma K 2 O values range from 0 to 5.62 wt% with an average of 2.78 wt%, while the ground gamma ray spectrometry gave K 2 O values which range from 3.40 to 8.98 wt% and an average of 4.89 wt%. The U values from X-ray fluorescence (XRF) analyses range from 0.40 to 14.80 ppm and an average of 3.49 ppm. Airborne gamma U values range from 0 to 7.46 ppm and an average of 2.95 ppm, while the ground gamma ray spectrometry show U values which range from 0.73 to 9.36 ppm and an average of 3.06 ppm. The XRF Th values range from 3.20 to 42.10 ppm and an average of 13.60 ppm. The airborne gamma Th values range from 0 to 22.81 ppm and an average of 10.27 ppm while the ground gamma ray spectrometry shows Th values which range from 0 to 22.81 ppm and an average of 18.42 ppm.
The field gamma-ray spectrometry has also shown that the nepheline syenites and alkaline rocks vary considerably in their K content. Table 8 shows summary descriptive statistics of the data for the ground field spectrometry survey and the XRF analyses of the same samples. For the rock samples which were acquired by field spectrometry, the South Malawi nepheline syenites from Junguni area (Table 8) showed the highest K 2 O content. For the XRF analyses, the highest K 2 O content was found in the carbonatite-associated nepheline syenites (from the Songwe-Mauze Complex) as shown in Fig. 15 and Table 8. Fig. 15 summarises the K, U and Th content in the areas surveyed during this study's fieldwork. As further shown in Table 9, the results also show that there is a weak correlation between data analysed by XRF instrument and the gamma-ray field spectrometer for potassium. There is however no significant correlation for U-Th pairs of data acquired using the field gamma ray spectrometer and the XRF laboratory instrument.

Discussion
Our results show that, in general, there is a moderate positive agreement for most rocks between the K content (recalculated as K 2 O wt.%) reported by the field gamma ray spectrometer and the results obtained using XRF (Table 8). The results show that the direct measurement of radioactive elements using the field ground geophysical survey was effective in detecting the radioactive elements. However, we also observed that carbonatites showed poor agreement, and this may be due to sample heterogeneity or weathering. The results suggest that hand-held field spectrometry can be used to identify rocks with high K 2 O content when laboratory XRF analysis is not available.
While some of the carbonatite-associated and the South Malawi nepheline syenites, (particularly the Junguni, Tundulu and Songwe-Mauze nepheline syenites) are similar (Fig. 11), the rocks from the Junguni intrusion have the highest amount of normative nepheline (40 wt%) as seen in Table 6. These results agree with petrographic studies done by Woolley (2015), who found up to > 60 wt% normative nepheline in some rocks of the Junguni intrusion, with strong peralkalinity characterized by normative acmite values of up to 28 wt%.
The Songwe-Mauze and Tundulu intrusions also show high normative nepheline coupled with the presence of normative leucite. The positive correlation of normative nepheline with the normative Kfeldspar (especially orthoclase), as reflected in the results, suggests that in mapping nepheline syenites it is likely that the other alkaline rocks or those closely related to the nepheline syenite could be mapped together with nepheline syenites. Alternatively, this suggests that field spectroscopy is not fully effective in distinguish clearly members of the silicate family from each other, as found in some previous studies (Hecker,der Fig. 11. Classification of alkaline rocks from Malawi after De la Roche et al. (1980). *W stands for data from Eby et al. (1998). Red-dashed area shows the region for alkaline rocks from other parts of the world. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Fig. 10. Harker diagrams for major elements diagrams. *W denotes data from Eby et al. (1998).
The weathering indices have further shown that the rocks are weakly altered, with three clusters. One cluster comprises the rocks from Junguni and Mongolowe (the South Malawi nepheline syenites) and those from Tundulu and Songwe-Mauze (carbonatite-associated nepheline syenites) which are relatively less altered while another cluster has some of the Kasungu, Kasungu-Chipala (the Central Malawi nepheline syenites) and some of the Junguni nepheline syenites and these are more altered. The third group is for the most altered nepheline syenites and includes the Chaone (the South Malawi nepheline syenites) and some of the Kasungu-Chipala nepheline syenites. The carbonatiteassociated nepheline syenite at Tundulu differs characteristically from all the other intrusions. Whether this is due to the level of alteration or geochemical composition is not clear. Both factors may apply because it is also the only sample which does not show normative orthoclase as already shown in Table 5.
In terms of chemical composition, the nepheline syenites from Malawi, especially those the Illomba, the Junguni, Chaone, Songwe-Mauze and Tundulu complexes, show similar geochemical composition to the nepheline syenites from Cape Dome in British Columbia and  Pearce (1983) according to Sun and Mc Donough (1989) reconstructed using the PINGU tool by Cortes and Palma (2014).
North Cape in Norway (Fig. 11) that have been used in previous studies relating to soil fertility. Based on this geochemical similarity, our results therefore offer strong confidence that Malawi's nepheline syenites would equally be suited for this project goal. This has further demonstrated that Malawi with her numerous alkaline rocks could be an important source of alternative potassium fertiliser using both the nepheline syenites and carbonatite resources as candidates.
The results have also shown that the Tundulu complex is quite different from the other studied nepheline syenites. The distinctive characteristics of the Tundulu nepheline syenite may be interpreted as possible evidence of the carbonatite-nepheline melt mixing/interaction, perhaps like the nepheline syenites that surround the apatite-rich carbonatite on Nathace Hill (Malawi; Broom-Fendley et al., 2016). In addition, we note that some of the areas had previously been wrongly mapped on geological maps of Malawi. For example, the Nkhuzi Bay and Mauni quartz syenites (from the South-east Malawi quartz syenites) are mapped as granites on the existing Geological Survey's geological maps. This shows that there is a need to re-map the country's geology, as proposed by the Geological Survey of Malawi.
The results further show relationships between geographically dispersed rocks based on the R 1 -R 2 (De la Roche et al., 1980) classification of igneous rocks. As shown in Fig. 11, the X-ray fluorescence analyses for the Junguni, Chaone, Songwe-Mauze and Tundulu complexes, are closely related and form one cluster, which is geochemically different from the other intrusions. Similar results were obtained by Eby et al. (1998) who also mapped the |Central Malawi nepheline syenite (Kasungu and Chipala intrusions) and the Northern Malawi intrusions (Illomba and Ulindi nepheline syenites). Based on these secondary data, we conclude that the two Northern Malawi intrusions also have potential as potassium fertiliser sources. If fully exploited as soil remineralizers, the potassium silicate resources in Malawi and other rift tectonic settings of the world could greatly contribute to achievement of Sustainable Developmental Goal (SDG) 2 of ending hunger and achieving global food security.
In addition, the results of the geochemical analyses of the rocks from  Malawi have shown similarities in chemistry between geographically dispersed rocks based on the R 1 -R 2 (De la Roche et al., 1980) classification of igneous rocks (Fig. 11). In addition, the previous geochemical data by Eby et al. (1998) for some of northern Malawi nepheline syenites, such as the Illomba, the Ullindi show similar plots to some samples analysed in this study. The geochemical analyses further show that some of the South Malawi and carbonatite-associated nepheline syenites are similar in geochemical composition to the nepheline syenites from other parts of the world. For example, the geochemical compositions of Chaone, Songwe-Mauze, Junguni and Tundulu nepheline syenites are similar other countries such as the North Cape nepheline syenites of Norway, and others from British Columbia. These nephelines syenites have been tested and found to be viable potash sources. The results, therefore, suggest that Malawi's nepheline syenites have potential for use just like those from other areas in other parts of the world. Such information opens doors for further exploration and exploitation of these agro-minerals by potential investors.

Conclusions
The results have shown that Malawi's nepheline syenites have potential as sources of potassium for crop nutrition and may be classified into two major groups. Group A nepheline syenites show higher potential as an alternative potassium source due to their high nepheline content than Group B nepheline syenites. In addition, Group A nepheline syenites have similar geochemistry to the rocks used as potassium sources in other areas, such as the North Cape and Cape Dome nepheline syenites. This suggests that Malawi has potential for alternative potassium reserves derived from nepheline syenites. Novel alternative potassium sources in Malawi and other parts of Africa will greatly benefit millions of farmers in the developing world, particularly in Sub-Saharan Africa (SSA), a region affected by high fertiliser costs. Further work is needed, such as detailed sampling and field gamma ray spectrometry to assess the available nepheline syenite resources in Malawi, and plant growth trials to fully ascertain the suitability of these rocks as potash sources. The identification of davidsmithite in Malawi is also another significant finding, which needs further investigation. Finally, this study has shown that some areas were previously wrongly or inconsistently mapped as shown on the documented geological maps of Malawi. This implies that there is a need to re-map the country's geology, as also proposed by the Geological Survey of Malawi.  Fig. 15. Discrimination plots of gamma-ray data measured by ground (field) spectrometry vs XRF data for (a) potassium; (b) thorium. (c) uranium (d) thorium/uranium.

Table 9
Correlation coefficients for pairs of gamma-ray data measured by ground (field) spectrometry vs XRF data for (a) potassium; (b) thorium. (c) uranium (d) thorium/ uranium. In addition, Fig. 15 and Table 10 show that, generally, higher values were acquired using the gamma ray field spectrometer for all the three elements.  Note: Nsy = nepheline syenite; Qsy = quartz syenite.