The study was conducted in Nebbi and Mukono districts in the West-Nile and L. Victoria Crescent agro-eco- logical zones (AEZs) of Uganda, respectively (Figure 1).

Nebbi and Mukono districts were selected because they are the main hotspots for J. curcas L. growing in Uganda. Besides, the two districts have differences in agro-ecological conditions (Table 1) which could cause variations in J. curcas L. oil production.

J. curcas L. seeds were collected and bulked from randomly selected plants from farmers’ plantations in Nebbi and Mukono districts. This was done in three phases (June 2014, November 2014 and March 2015) according to the region in order to get a composite sample. The collected seeds were cleaned thoroughly, sun dried, kept in muslin bags and stored at room temperature until analysis for oil quality and quantity. Prior to oil extraction and analysis, three samples per region, each weighing 5 kg, were then dehulled to remove seed coats by using a small stick and later the hulls and the seeds were separated by winnowing.

The oil contained in 3 kg of J. curcas L. seeds was extracted in a Soxtec apparatus for 4 hours, using petroleum ether (boiling point of 40˚C - 60˚C) as an extraction solvent. The extracted oil was recovered by solvent evaporation using a Soxhlet apparatus to remove the majority of the solvent; followed by rotatory evaporation at 40˚C under reduced pressure. The extracted seed oil was weighed and stored in amber glass flasks, at 18˚C, for subsequent analysis. The amount of oil in seeds was calculated by dividing the mass of extracted oil by the mass of crushed Jatropha seed kernels. For each accession, the oil content of three samples was determined in triplicate tests in order to enhance accurate statistical inference.

The extracted oil was analysed for physicochemical properties (kinematic viscosity, cetane number, acid value, iodine value and free fatty acids (FFA)) using American Standards and Testing Methods (ASTM D6751). The calorific value and Density were determined by bomb calorimeter and digital density analyser, according to ASTM D240 and ASTM D5002-13, respectively; whereas fatty acids and ester contents were determined using high performance liquid chromatography (HPLC; Model: RF-20AXS) basing on EN 14103:2011 standards. The element content (P, Na and K) of the oil was determined by flame spectrometry using Spectro Arcos ICP-OES analyser according to procedures described in European standards (EN 14107, EN 14108 and EN 14109, respectively). Each parameter was determined thrice for each sample per geographical location.

Means and standard deviations for the individual parameters in triplicate samples were determined to compare the quality and percent (w/w) quantity of J. curcas L. oil from Nebbi and Mukono districts. Data were statistically analyzed using paired-samples T-Test to study the difference between J. curcas L. oil quality and quantity from the study areas. All significance tests were set at P ≤ 0.05, and the statistics were analyzed using the statistical software package (SPSS version 16).

The percent (w/w) quantity of J. curcas L. oil recovered from seeds varied significantly (p < 0.05) between Nebbi and Mukono districts as shown in Figure 2.

Nebbi depicted high oil recovery (51.33% ± 0.58% w/w) from J. curcas seed kernel than Mukono (48.00 ± 1.00% w/w), though both values fall in a range (40% - 60% w/w) reported elsewhere [9] [10] [14] [24] - [28] . The difference in oil content could be attributed to variations in altitude, genetics, climatic and soil conditions between the two regions [25] [29] - [31] . Nebbi and Mukono districts have different soils, mean annual rainfall, temperature and altitudes, which could have contributed to the variance in seed oil content. Ovando-Medina, et al. [30] also found a negative correlation of the oil content with the altitude at which J. curcas L. plant is grown. In the same way, this conforms to the study by Pant, et al. [29] who determined seed oil content variation in J. curcas L. plant grown in two different soil conditions and three altitudinal ranges in India. They found higher oil content (43.19%) of seeds harvested from the plant grown in non-arable soils with low altitude (400 - 600 m) than arable soils with high altitude (38.66% and 800 - 1000 m). Temperature has also been found to have a significant effect towards oil content [16] and this has been supported by several authors [1] [8] [32] who found that J curcas L. performs very well in regions of high temperatures ranging from 20˚C to 40˚C mean annual.

Nevertheless, improper processing methods such as prolonged exposure of harvested seeds to direct sunlight can deteriorate the oil yield significantly [26] .

Analysis of J. curcas L. seed samples from Mukono and Nebbi revealed significant differences (p ≤ 0.05) in Free Fatty Acids (FFAs) and acid value (Table 2). However, the mean values of ester content (triglycerides), kinematic viscosity, calorific values, density, cetane number (CN) and iodine value of J. curcas L. oil were not significantly different (p > 0.05). The ester content of the extracted oil was in a range of 92.8% to 95.1% for the two districts, close to the values (96% to 98%) reported by other authors [13] [33] . Yet the acidity of the oils was found significantly higher (36.42 ± 0.24 mg KOH/g for Mukono and 35.43 ± 0.15 mg KOH/g for Nebbi) and deviated from biodiesel standard value (2 mg KOH/g [34] ) for diesel engine. The difference in FFAs and acid value in both regions could be due to variations in climatic conditions [8] , as well as the presences of high unsaturated oleic and linoleic acids which make J. curcus L. oil prone to oxidation in storage [8] .

The high acid values recorded in J. curcas L. oils from Mukono and Nebbi was generally due to the presence of high FFAs (1.93% ± 0.15% and 1.52% ± 0.10%, respectively, compared to 0.02% standard value for biodiesel engine), primarily resulting from hydrolysis of triglycerides [36] . This degradation reaction is supported by temperature and it mainly happens in damaged seeds by the action of lipases in the existence of water [36] . FFAs or triglycerides are usually transformed into alkyl-esters (biodiesel) by reaction with short-chain alcohols (e.g. methanol or ethanol) in the presence of a catalyst (Equation (1) and (2))―a process known as esterification and transesterification, respectively [13] . However, high amounts of FFA (mostly above 3%) would lead to formation of soap and emulsions, owing to excessive alkaline catalyst consumption, hence significantly reducing the biodiesel yield [32] [37] .

Despite the suitability of J. curcas L. oil for biodiesel production from Mukono and Nebbi in terms of ester content (93.30% ± 0.50% and 94.70% ± 0.40%, respectively; biodiesel standards: >96.5% [34] ), it has been

^*Significant at p ≤ 0.05; otherwise not significant.

found that kinematic viscosity (32.27 ± 0.81 and 32.90 ± 0.35 mm²/s, respectively) deviated greatly from biodiesel standard value (3.5 - 5 mm²/s) for diesel engine [34] . In line with Ferrari, et al. [38] , high viscosity of J. curcas L. oil may affect its flow and ability to spray in the engine and at low temperatures may cause mechanical failure of injection pump drive systems. This may be reduced significantly through etherification and transesterification [38] [39] of the oil, by converting its high-molecular ester compounds (Triglycerides) into simpler Fatty Acid Methyl Esters (FAMEs). Biodiesel, a mixture of various kinds of esters such as FAMEs (esters of fatty acids), has been documented as having physical characteristics closer to those of petroleum diesel than pure vegetable oil, depending on the type of vegetable oil [16] [34] [37] [38] . Moreover, Drapcho, et al. [16] found that biodiesel contains about 13% lower energy density than petroleum diesel, though has greater lubricity and undergoes further complete combustion.

In this study, results presented in Table 2 show that calorific value of J. curcas L. oil from Mukono and Nebbi (37.10 ± 2.87 and 41.47 ± 0.76 MJ/kg, respectively) fall within what is recommended for biodiesel (>36 MJ/kg [34] ) and compares with petroleum diesel (43.35 MJ/kg [40] ). Notably [41] , the high calorific value could be attributed to the presence of high quantity of polyphenol and hydrocarbon in the oil. Yet according to Xiu, et al. [39] , addition of polar solvents such as methanol, ethanol, and furfural to the oil could further increase its calorific value, although it may have a substantial effect on cetane number (CN) [38] . As said by Ferrari, et al. [38] , cetane number is the tendency of a fuel to combust under certain conditions of pressure and temperature. Literature [38] suggest that high CN is associated with fast engine starting and smooth combustion, whereas low cetane deteriorates engine performance and leads to high exhaust emissions of hydrocarbon gases and particulate. In this work, results show that CN of J. curcas L. oil from Mukono and Nebbi was found to be 36.00 ± 0.00 and 37.00 ± 1.00, respectively―lower than fossil diesel (47, [40] ) and the norm (>51, according to EN ISO 5165 standard) for biodiesel in diesel engines (Table 2). However, it is important to note that CN value of biodiesel can be increased by increasing the length of hydrocarbons of fatty acid and ester groups, but can also be reduced significantly by increased number of double bonds [38] . It is therefore necessary to mention that the lower CN in the present work could be attributed to higher fractions of unsaturated fatty acids and esters (C18:1 and C18:2) in J. curcas L. oil than saturated acids and esters (C16:0 and C18:0), as depicted in Table 2.

There was no significant difference between Iodine values of J. curcas L. oil from Mukono (108.97 ± 1.17 mg/g) and Nebbi (108.53 ± 1.05 mg/g) (Table 2). The Iodine value of J. curcas L. oil in both study sites was found within biodiesel standard value (<120) for diesel engine in reference to EN 14111 standards. This value, usually expressed in grams of iodine reacted with 100 grams of biodiesel, is a measure of the total unsaturation within a mixture of fatty acids [38] . A study by Prankl and Wörgetter [42] on an engine has found that fuels with higher iodine number have a tendency to polymerize and form deposits on piston rings and piston ring grooves when heated. Such high-molecular compounds, which to a large extent appear only in fatty acid esters with three or more double bonds, have a negative effect on the lubricating quality, thus resulting into engine damage [42] [43] . To avoid this difficulty, it has been recommended by some biodiesel experts [38] to limit the content of linolenic acid methyl esters and polyunsaturated biodiesel instead of the total degree of unsaturation as it is expressed by the iodine value.

Results on chemical characteristics (fatty acid, ester and element compositions) of J. curcas L. oil are presented in Table 3. The oil extract from Mukono and Nebbi consists of fractions of saturated (palmitic acid: 14.19% ± 0.14% and 14.60% ± 0.27%, respectively; and stearic acid: 3.91% ± 0.19% and 3.83% ± 0.12%, respectively) and unsaturated (oleic acid: 33.90% ± 0.40% and 34.90% ± 0.50 %, respectively) fatty acids, comparable with those reported by Augustus, et al. [41] . The difference in the mean values of oleic acid component between the two regions was significant; while for Palmitic and Stearic acids, the difference was insignificant. Linder [44] revealed that saturated fatty acids store additional energy per carbon than unsaturated fatty acids; though, unsaturated fatty acids have melting points much lower than saturated fatty acids. Furthermore, saturated fatty acids including Palmitic and Stearic acids influence a higher pour point temperature compared to unsaturated fatty acids in biodiesel [27] . According to Tremolieres, et al. [45] , there is also a strong correlation between oleic acid formation in the seed and temperature; whereby oleic acid increases under high temperature and decreases under lower temperature conditions during seed maturation. This probably explains the significant differences in oleic acid content between Mukono and Nebbi―where mean annual temperature is higher.

There were no significant differences in the content of the elements of Na, K and P in J. curcas L. oil between Nebbi and Mukono districts (Table 3). However, K and P values exceeded the standard value for biodiesel in the diesel engine (Table 3). According to literature [32] [34] , high levels of Na, K and P in biodiesel fuel may lead to end-use problems; Na and K forms ash within the engine and may lead to clogging of filters and fuel injection pumps, while P (in form of phospholipids) may lead to the impairment of the catalyst and hindrance of phase separation, during production process. Na and K (alkaline metals) may be introduced into biodiesel during production process, as a result of catalytic degradation, since they form a major constituent of the catalyst. Nevertheless, P in biodiesel may come from phospholipids and inorganic salts contained in the feedstock oil. As reported by Ferrari, et al. [38] P has a strong negative impact on the long term activity of exhaust emission catalytic systems and basing on this, its existence in biodiesel is limited by specification. On this note, oils with higher P content should first go through degumming before being processed into biodiesel [32] .

^*Significant at p ≤ 0.05; otherwise not significant.