Seasonal fuel consumption, stoves, and end-uses in rural households of the far-western development region of Nepal

Seasonal fuel assessments were conducted in households in the High Hills region of the Far-Western Development Region of Nepal (Province 7). Households were enrolled from villages in four village development committees (VDCs) within the Dadeldhura and Baitadi districts. Two VDCs were in a low elevation subtropical climate zone ('lower elevation,' 1000–2000 m), and two in a high elevation (above 2000 m) temperate climate zone ('high elevation'). Households in the low elevation were closer to protected forest areas ('community forests'), so fuel collection for these homes was more regulated than in the high elevation.

Fuel assessments were conducted over four two-week periods at the midpoint of four seasons: winter (January), spring (April), monsoon (July), and post-monsoon (October). Ambient air temperatures taken outside a subsample of 12–20 kitchens over a three-week period each season were slightly warmer than historic station data but showed similar seasonality. The high region was approximately 2 °C colder than the low region, except during the post-monsoon season (figure 1).

A total of 110 households, encompassing 630–700 individuals, were enrolled at the start of the study. Sixty (55%) were in the subtropical low region, and 50 (45%) in the temperate high region. If a household was unable to undergo monitoring during any season due to temporary absence or activities resulting in atypical fuel use, it was excluded from that season but kept in the cohort in later campaigns. No houses requested to be dropped from the study. The dataset contains 786 days of fuel-use data, with 96 households having data for all four seasons with 51 (53%) and 45 (47%) being in the high and low elevations, respectively.

Questionnaires, described in the next section, were translated and administered in Nepali by local surveyors employed and trained by The Centre for Rural Technology-Nepal (Kathmandu, Nepal). All measurement instruments and protocols were approved by the Office for the Protection of Research Subjects at the University of Illinois at Urbana-Champaign.

A Seasonal Kitchen Performance Test (SKPT) was developed to measure seasonal changes in household fuel consumption (fuel use) characteristics. The SKPT is an expansion of the Kitchen Performance Test (KPT), which is a common protocol used to evaluate changes in household fuel consumption resulting from introduction of new cooking devices [41]. The KPT uses fuel weight to quantify consumption, as opposed to participant-reported fuel weights or proxies, and measurements are performed in houses under typical use conditions (uncontrolled).

In the SKPT, and most KPT variants, fuel consumption is quantified by weighing fuels at the beginning and end of a 24 hour period, and repeating this procedure over sequential days. The SKPT expands upon the KPT in several respects: it apportions fuels to different stoves; it collects more information on the end-uses of fuel and stove use events associated with each stove; and it is structured and worded for repeated use in the same home in order to examine changes over time. Household demographic information and agricultural practices were also collected each season from participant self-reporting.

Fuel consumption was measured over two sequential 24 hour periods in each household and season. Resuts from a single-season KPT assessment by Berrueta et al showed that most of the variability in population fuel estimates is attributable to between-home differences and that the greatest precision gain occurs from the second day of measurement [42]. It was thus deemed more beneficial in this study to maximize the number of households with a smaller number of repeat measurements in each house.

At the first visit of each season, participants were asked to identify stoves that were used, regardless of frequency. During each visit, the fuel allotted to each stove was weighed with a resolution of 10 g. Fuel allotted to a specific stove was placed beside it, and instructions to use only these fuel piles given to participants. Wood moisture content (%, wet basis) was measured in triplicate at each visit for fuel dedicated to each stove using a Delmhorst moisture meter (J-2000, Delmhorst Instrument Co., New Jersey, USA). Average stove- and day-specific moisture values were used to adjust fuel weights for moisture content. All fuel masses reported are dry mass (e.g. kg dry fuel) even if not explicitly stated. After each 24 hour sample period, participants were asked to recall the stove events of the previous day for each stove, including a description of the task and its approximate duration. The procedure for using reported stove tasks and durations to distribute fuels to end-uses is described in section 2.4. Reported duration of stove use was also used to estimate the average burn rate of stoves used by the family by dividing fuel consumption by the corresponding reported stove use duration (kg fuel hour⁻¹). The use of fuels other than wood was low in the sample population, so this study focuses primarily on wood, and the term 'fuel' in the remainder of this paper implies wood unless otherwise stated.

We separate stoves into two categories: main cooking stoves and supplemental stoves. The main cooking stove was identified by participants, and was typically where most meals were prepared. Identification of the 'main' or 'primary' cooking device is common within household energy surveys, despite its subjective interpretation. Supplemental stoves are any other devices where fuel is used for cooking or non-cooking needs. In many household surveys, supplemental stoves are not typically identified, although the fuel used in them would be quantified in a household consumption survey.

We also use several terms to describe seasonal changes in stove use characteristics. Stove 'prevalence of use' refers to the proportion of homes using stoves at a given point in time. We intentionally use this term as distinct from the term 'stove usage,' which has become synonymous with direct stove use monitoring measurements (SUMs). We use the term 'activity level' to describe the rate of fuel consumption for an end-use or stove in a season (e.g. kg wood per day).

Wood stoves were of two types: traditional 'chula' (singular 'chulo') which were stationary mud stoves (no chimneys) built by members of the home and found in all kitchens, and three-stone fires, which were typically used outdoors but occasionally found inside the kitchen along with the chulo. Thirty-four households—26 (43%) in the low and 7 (14%) in high elevations - had a rocket-type stationary mud-brick stove with metallic combustion chamber (rocket stoves) distributed as part of Nepal's National Rural Renewable Energy Program. Too few rocket stoves were present in the sample to have sufficient statistical power for an assessment of these stoves' performance with a cross-sectional design. The main cooking stove was typically located in an enclosed kitchen separated from the living and sleeping areas of the house. As many as 65% of the houses reported some use of metal heating pans, or 'makal,' for space heating. These makals did not burn fuel directly, but carried hot charcoal made in the stoves. Photos of stove designs in the study population are available in the supplementary information (SI) available at stacks.iop.org/ERL/12/125011/mmedia.

Daily fuel consumption was distributed across end-uses using participant-reported information on stove tasks and their durations. For each stove, we assumed that the amount of fuel used for an end-use was proportional to the fraction of time spent performing tasks associated with an end-use category. All time corresponding to meal events on a stove were combined into a single 'all meals' end-use category. Other end-uses distinguished were drinks, space heating, animal food, and water heating. Any remaining tasks were grouped into an 'other' end-use. These fuel end-use estimates were generated for each stove in a house on each measurement day. Other studies have disaggregated hours of stove operation into individual meal tasks (e.g. rice, beans, soup, bread) [9, 21, 40]. We use a similar approach to disaggregate fuel consumption to end-use categories selected to distinguish between cooking and non-cooking needs.

We allocated fuel to end-uses within each home from relative time spent on tasks in order to minimize bias associated with recall-based information in the process of apportioning fuel to end-uses. While the perception of time may vary between people, the relative allocation may be more reliable (e.g. the evening meal took twice as long as drink preparation). The amount of fuel (Fuel) used for end-use, i, on stove, j, in house, k, is calculated from the reported duration of total stove use (t) on a measurement day as:

$$\begin{equation} {\rm Fue}{{\rm l}_{i,j,k}} = {\rm Fue}{{\rm l}_{j,k}} \times {f_{i,j,k}} = {\rm Fue}{{\rm l}_{j,k}} \times \frac{{{t_{i,j,k}}}}{{\mathop \sum \limits_{i = 1}^n {t_{i,j,k}}}}. \end{equation}$$

This time-based allocation approach inherently assumes that stove burn rates, or rate of fuel consumption when the stove is operating, are the same for all end-uses in a home on a given day.

Table 1 summarizes characteristics of the 110 households. Household size averaged 6.2 ± 1.7 (arithmetic mean ± one standard deviation), varying 10% across seasons but never differing significantly between elevation groups (elevation). Homes in the high elevation group had a head of household that was on average 7 years older, and had 0.9 more animals requiring cooked food.

All households reported use of wood and over 97% reported wood as a primary fuel each season. Wood moisture content varied seasonally from an average low of 9.1% ± 2.7% in spring to 24.5% ± 7.2% in monsoon. No families reported use of dung, charcoal (excluding charcoal generated from stove use), or kerosene for any end-use in any season. LPG stoves were present in 10% of houses, but were used sparingly based on the duration between cylinder (14.2 kg) purchases, approximately one year on average. Use of supplemental stoves typically constituted one stove in addition to the main cooking stove. Less than 5% of homes used three supplemental stoves, and none reported more than three. Ninety-four percent of houses had connections to grid electricity, and almost all of those (96%) reported it as their main source of lighting. Houses without electricity connections reported wood or chargeable lanterns as main lighting sources. Grid-powered lighting was supplemented with charged lanterns (e.g. wind-up, charged from grid electricity) or candles during load shedding events (brown-outs).

Zoom In Zoom Out Reset image size

Average household and per-capita consumption ranged seasonally from 5.1−9.1 kg day⁻¹ and 1.0–1.7 kg capita-day⁻¹ (figure 2). Households using supplemental stoves consumed 30%–50% more fuel per capita than those using a single stove (p-values < 0.05), except in the monsoon months when the prevalence of supplemental stove use was lowest (table S2). Fuel use peaked in winter, increasing by 3.6 kg day⁻¹ (95% confidence interval: 2.9, 4.2) or 0.6 kg capita-day⁻¹ (0.4, 0.7) relative to the average of all other seasons, corresponding to a 53% (44%, 58%) increase in per-capita fuel use. Pairwise comparisons among non-winter seasons showed a monsoon to post-monsoon decline of 15% (p-value = 0.023), but no other seasonal differences exceeded 5%. Changes in per-capita fuel use from winter, compared with the average across other seasons, will be referred to as 'winter changes.' The average winter change for a household using a supplemental stove in winter was an increase of 66% (52%, 79%), while the change for a household using a single stove was 30% (11%, 48%), a difference of 36% (p-value = 0.0013).

Winter changes could not be attributed to the main cooking stove. Average fuel use in the main cooking stove in winter and monsoon was not significantly different at 1.10 ± 0.8 and 1.07 ± 0.7 kg capita-day⁻¹, respectively. Post-monsoon and spring fuel use were also not significantly different at 0.89 ± 0.5 and 0.86 ± 0.6 kg capita-day⁻¹. The largest difference in main cooking stove activity levels in any two seasons was 0.2 kg capita-day⁻¹, or a 27% increase.

The seasonal change in fuel use was, instead, attributable mainly to fuel usage in supplemental stoves. The prevalence of use of supplemental stoves changed by a factor of four across seasons, reaching a low of 17% in monsoon, and a 68% high in winter. Average fuel consumption (activity levels) in supplemental stoves changed by a factor of two across seasons, reaching a low of 0.4 ± 0.4 kg capita-day⁻¹ in monsoon, and a high of 0.9 ± 0.5 kg capita-day⁻¹ in winter. The monsoon season differed the most from winter in both prevalence of use and activity levels of supplemental stoves. In the full cohort, the supplemental stove accounted for 19% of annual household fuel consumption, and 6% (monsoon) to 32% (winter) of fuel use in any season. Among only homes that used supplemental stoves, those uses accounted for 32% ± 27% (monsoon) to 47% ± 14% (winter) of household fuel consumption.

Although household fuel use was influenced by both the main cooking and supplemental stoves, neither device category was solely responsible for seasonal trends in household fuel use. Supplemental stoves drove most of the observed winter change, but the seasonal pattern in household fuel use was affected by changes in both stove categories. In monsoon, activity levels on main cooking stoves was high relative to other non-winter seasons (figure 2, middle). This fuel increase was offset, however, by a low prevalence and activity level of supplemental stoves (figure 2, right). The total household fuel use in monsoon was then similar to other non-winter seasons (figure 2, left). In winter, when household fuel use was highest, activity on main cooking stoves was also elevated, but also accompanied by high supplemental stove prevalence and activity.

Figure 3 is a visualization of the annual fuel consumption for the average home in each elevation, distributed among seasons, stove categories, and end-uses. The flow diagram reflects the temporal dimension of fuel use, and the relationship between fuel, stoves and energy needs. The finding that household fuel use is highest in winter and similar among other seasons is shown by the relative sizes (widths) of the seasonal nodes. The relative sizes of the stove nodes demonstrate the division of fuel use between main and supplemental stoves. Colors identify different seasons, showing that houses at high elevations use more fuel relative to those at lower elevations at most times of the year.

Zoom In Zoom Out Reset image size

Figure 3 also encapsulates the following findings: (1) by end-uses, meals account for the largest single share of annual fuel use, and total non-meal end-uses account for approximately one third of annual fuel requirements. (2) Water heating and preparation of animal food account for over half of the fuel used for non-meal end-uses. (3) Meal-related fuel needs are relatively constant across seasons in both regions, and are met predominantly by fuel used in the main cooking stoves. (4) Non-meal end-uses fluctuate by season, peaking in winter from fuel used predominantly in the supplemental stoves. Annual fuel flows emphasizing individual seasons and other alternative flow visualizations are presented in the supplementary information.

Figure 4 (left) presents the proportion of reported stove events by season, stove, and end-use. A difference in the distribution of end-use events provided by each stove type is apparent in every season. The main cooking stove was dominated by meal and drink preparation, while the supplemental stoves were used for tasks involving the heating of large pots, either for animal food or heating water. During non-winter seasons, main cooking stove events dominate household event profiles.

Total hours of stove operation increased in winter, as did hours spent on most end-uses (figure 4, right). Additional time was spent on meal preparation (+1.3 hours), supporting the hypothesis that meal-making stoves might be used to provide warmth. Formal space heating also increased in winter, but even at its peak accounted for less than 5% of total events and hours of stove use. Two of the largest increases in stove operation—animal food (+0.7 hours), and water heating (+1.3 hours)—were predominantly performed on supplemental stoves. These additional stoves significantly altered the household stove event profiles, although they did not alter uses of the main cooking stove.

Zoom In Zoom Out Reset image size

Per-capita fuel use among homes at high elevations was greater than those at low elevations by at least 30% (p-values < 0.03) in all seasons but the monsoon. Fuel end-use distributions in the average high elevation home showed a larger fraction of fuel going to water heating and animal food in non-winter seasons; the latter is consistent with the greater number of animals requiring cooked food (figure 5, right).

The increase in per-capita fuel use in winter relative to other times of the year was greater among homes at higher elevations. Compared to the average per-capita consumption during spring, monsoon, and post-monsoon, the average home used 77% (59%, 94%) more fuel in winter in the high elevation and 35% (22%, 49%) more in the low elevation (figure 5, left), a difference of 42% (p-value = 0.0001). Fuel used on the main cooking stove in winter increased modestly at high elevations (15%) but remained unchanged in the low-elevation homes, further demonstrating the important role of supplemental stoves in driving seasonal household fuel use trends. End-use profiles in winter were very similar between elevation groups, suggesting that no individual end-use was responsible for the large difference in relative winter change between elevations.

Zoom In Zoom Out Reset image size

Participant-reported stove use characteristics and fuel consumption are often assumed to be unreliable in many research and evaluation contexts. For this reason, physical measurements are often recommended, such as stove surface temperature that indicates usage or weighed fuel to indicate consumption. Seasonal trends of self-reported stove use collected in this study, however, closely reflected those of measured fuel weights (figure S5), suggesting that survey-based measures may be useful for estimating seasonality in fuel consumption. Spearman correlation coefficients of stove use duration from questionnaires and direct measures of fuel consumption were moderate: winter (ρ = 0.75), spring (ρ = 0.52), monsoon (ρ = 0.41), and post-monsoon (ρ = 0.62). An estimate of the average winter fuel consumption based on reported stove use durations (relative to winter) and a single non-winter season direct fuel measurement campaign would have been within 20% of actual fuel use. Visual comparisons between reported stove use and measured fuel consumption are presented in the supplementary information.

Estimated burn rates of stoves varied little across season, implying that seasonal changes in fuel use were driven predominantly by shifts in the duration of stove operation. The annual average fuel burn rate across all homes was 1.1 ± 0.5 kg hour⁻¹. Season-specific burn rates were within 20% of each other, ranging from 1.0 (winter) to 1.2 kg hour⁻¹ (post-monsoon). With the exception of the post-monsoon (1.2 ± 0.6) and winter (1.0 ± 0.4), no statistically significant seasonal differences in burn rates were measured (p-values > 0.05).