Program Report | Irrigation scheduling chameleon sensor | Page 1 of 14 Do Chameleon Moisture Sensors adopted by farmers increase their irrigation intensity and yields? A proof-of-concept field experiment November, 2025 Working paper Program Report | Irrigation scheduling chameleon sensor | Page 2 of 14 Contents Contents Introduction 3 Objectives 4 Methods 4 Results and discussions 8 Conclusion 11 Reference 12 Authors: Koirala, P., Khadka, L., Acharya, H.P., Adhikari, S., Lamichhane, H.P., Adhikari, S., Adhikari, S., Shrestha, N., KC, J., Krupnik, T.J. Program Report | Irrigation scheduling chameleon sensor | Page 3 of 14 Abstract The adoption of advanced irrigation technologies, such as Chameleon Moisture Sensors, has the potential to enhance water efficiencies and increase crop yields. However, we know little about whether Chameleon Moisture Sensors have potential to improve farmers’ water use efficiencies and subsequent crop yields in the context of Nepal. To this end, we designed a proof-of-concept quasi-experiments and conducted with five treatments: (1) Farmers using Chameleon Sensors (T1), (2) Neighboring farmers of T1 (T2), (3) non-neighboring farmers of T1 (T3), (4) Farmers using conventional moisture technology located at distance from T1 (T4), and (5) Neighboring farmers of T4 (T5) with a total of 138 wheat growing farmers of Kailali district of Nepal. We employed with and without subject design and collected periodic datasets for each five days, including baseline and endline survey, to evaluate the effects on irrigation practices and crop yields. Descriptive and regression analyses revealed that farmers using Chameleon Sensors (T1) increased their irrigation intensity by +1 event compared to their counterparts using conventional technology (T4). Furthermore, farmers with the sensors experienced a positive increase in yield by 0.43 MT/ha, equivalent to NPR 12,000 Mt/ha compared to conventional methods. It also appeared that indirect yield increment of 0.36 Mt/ha compared to neighbors (T2) of the sensors adopting farmers, indicating sensors’ potential spillover effects on yield. Overall, the results demonstrated that the adoption of Chameleon Moisture Sensors leads to increased irrigation intensities and yields among wheat farmers, implying their effectiveness over conventional technologies. These findings from the proof-of-concept highlight the potential of Chameleon Moisture Sensors to improve water and wheat yield, suggesting an emerged scope for promoting such technologies along with a full-phased experiment to draw insights to scale up. Keywords: Chameleon Moisture Sensor; Agriculture; Irrigation Intensity; Yield; Farmers; Experiment; Nepal Introduction Agriculture in Nepal faces increasing stress from climate variability, uneven rainfall, and rising temperatures (Koirala et al., 2022). These changes pose serious consequences on the stability of agricultural productivity, especially for smallholder farmers who depend on rainfall and traditional irrigation systems (Porter et al., 2017). In recent decades, shifts in precipitation patterns have disrupted soil moisture dynamics and shortened the growing period of key cereals such as wheat (Kaini et al., 2022; Risal et al., 2022). In many areas, farmers have been facing both drought and flooding risks within the same season, making efficient water management more critical than ever (Porter et al., 2017; Risal et al., 2022b). Historically, Nepalese farmers have managed water through indigenous and community-based irrigation systems such as canals, ponds, and kulos (traditional water channels) (Joshi, 2015). These systems, grounded in local norms and collective governance, have fostered cooperation and equitable resource sharing (Dixit et al., 2009). However, population growth, land fragmentation, and changing hydrological patterns have weakened these traditional systems (Risal et al., 2022a). Farmers often rely on subjective judgment or social cues to decide when to irrigate, leading to inefficient use of water resources. Over-irrigation or delayed irrigation not only wastes scarce water but also reduces soil fertility and crop productivity. Timely and adequate irrigation planning, and execution are two critical factors for increasing the yields of agricultural crops. While a significant number of farmers in Nepal rely on rainfall for water supply, those in certain areas, such as Kailali, already have sufficient water supplies for year-around irrigation in the absence of ample rainfall (Risal et al., 2022). However, farmers in these regions are observed to overlook needed water management practices, specifically irrigation scheduling together with some exogenous governance issues, thereby failing to achieve potential crop yield (Parajuli et al., 2021). The ignorance basically lies when the lack of adequate information provision to guide farmers on when and how much to irrigate (Kapri & Ghimire, 2020). Farmers in Africa and other regions are increasingly utilizing modern technologies, such as the “Chamelian Moisture Sensor,” which can accurately monitor and signal farmers about the optimal irrigation scheduling (Kaveney et al., 2025). However, the extent to which farmers can benefit from such technology remains unknown. This experimental study aims to investigate how the adoption of the "Chamelian Moisture Sensor" by farmers supports their irrigation scheduling and, consequently, the yields of their major crops. To address these challenges, policy makers have been promoting data-driven, climate-smart irrigation practices that complement community systems. Technologies providing real-time soil information can help farmers synchronize irrigation with actual crop water needs (Nepal et al., 2024; Thapa & Scott, 2019). One such innovation is the Chameleon Moisture Sensor (CMS), a low-cost, color-coded device that allows farmers to visually assess soil moisture levels. It is considered a simple tool that informs farmers soil moisture status on the real-time basis (Kaveney et al., 2025). The sensor translates complex soil data into simple visual signals, enabling timely irrigation decisions even for farmers with limited literacy or technical training. Although similar tools have shown potential to enhance water productivity and resilience across South Asia (Ladha & Singh, 2009; Shrestha et al., 2015), there is lack of sufficient evidence that assesses their effectiveness along with farmers’ behavioral and yield impacts. In this context, this is proof-of-concept experiment study aims to address this gap by investigating the impact of the Chameleon Moisture Sensor through a pilot field experiment that explicitly measures both direct and spillover effects on irrigation behavior and yield among smallholder farmers. The field experiment designed not for definitive Program Report | Irrigation scheduling chameleon sensor | Page 4 of 14 hypothesis testing, but to assess feasibility, gather preliminary effect sizes, and refine protocols for a future large- scale randomized controlled trial (RCT). The primary objectives were to: (1) evaluate the practical implementation of the Chameleon sensor and treatment allocation in a real-world setting, (2) estimate initial trends in treatment effects to inform power calculations, and (3) identify any logistical or social complexities, particularly concerning spillover dynamics between farmer groups. Prior studies indicate that people’s behaviors can have potential association with their irrigation related decisions (Duflo et al., 2011; Fehr & Gächter, 2000). By testing these mechanisms in a real-world agricultural setting, the study provides early experimental evidence on the feasibility and effectiveness of integrating sensor-based irrigation tools into smallholder farming systems. As a proof-of- concept, it aims not to generalize across regions but to establish the empirical foundation for scaling up soil moisture monitoring and farmer-centered learning interventions. The experiment also explores how simple, low-cost technologies can complement indigenous water management tradition, such as social norms of cooperation and local user groups, to address contemporary challenges of water stress and sustainability. The findings contribute to the growing literature on climate-smart irrigation, behavioral technology adoption, and collective decision-making in agriculture. They also align with the CIMMYT 2030 Strategy, the Sustainable Development Goals (particularly SDG 2 on zero hunger, SDG 6 on clean water and sanitation, and SDG 13 on climate action), and Nepal’s national priorities for resilient and inclusive agricultural transformation. Overall, this study demonstrates how innovation, social learning, and traditional cooperation can jointly enhance adaptive capacity in smallholder farming systems facing increasing climate uncertainty. Objectives This study seeks to evaluate the impact of the “Chameleon Moisture Sensor” on farmers’ irrigation intensity and crop yields, both among adopting households and their neighbors. To this end, we address two primary research questions. First, does the adoption of the sensor lead to higher irrigation intensities and yields compared to conventional irrigation technology? To test this, we formulate our first hypothesis (H1): Households receiving the sensor will exhibit higher irrigation intensity and crop yield than those using conventional methods. The corresponding null hypothesis (H0) posits no difference in these outcomes between sensor-adopting households (T1) and pure control households using conventional technology (T4). Second, are there spillover effects, whereby neighboring or peer households of sensor adopters also experience improved outcomes? This leads to our second hypothesis (H2): Neighboring households of sensor adopters (T2) will demonstrate higher irrigation intensity and crop yield than neighbors of households using conventional technology (T5). The null hypothesis (H0) states that there exists no difference between these spillover groups. These hypotheses will be tested using conventional technology (T4). Second, are there spillover effects, whereby neighboring or peer households of sensor adopters also experience improved outcomes? This leads to our second hypothesis (H2): Neighboring households of sensor adopters (T2) will demonstrate higher irrigation intensity and crop yield than neighbors of households using conventional technology (T5). The null hypothesis (H0) states that there exists no difference between these spillover groups. These hypotheses will be tested using regression analysis, where the key dependent variables are irrigation intensity (frequency/volume) and wheat yield (Mt/ha), and the primary independent variables are the treatment dummies (T1, T2, T4, T5), with sociodemographic and structural factors included as controls to isolate the causal effect of the sensor technology. Methods Experimental design We conducted experiments with 143 farming households distributed across five treatment arms with unequal allocation: Treatment 1 (T1, 21 Sensor HHs + orientation), Treatment 2 (T2, 63 Immediate neighbor of sensors receiving HHs), Treatment 3 (T3, 17 only orientation receiving HHs); Treatment 4 (T4, 21 HHs with conventional methods); and Treatment 5 (T5, 21 neighbors of conventional methods). This allocation was intentionally designed to maximize learning about specific mechanisms that the larger sample for T2 (Spillover) was prioritized to robustly capture the variability and strength of peer effects, a central and under-researched component of this study. The smaller samples for T1, T4, and T5 were deemed sufficient for a pilot to provide initial estimates of the direct treatment and control baselines. The smallest group, T3, was included to explore the potential for secondary spillovers, acknowledging that this is a highly preliminary investigation. Experimental setup We distributed sensors of cost to the selected farmers, chosen based on specific selection criteria rather than randomly. We assume that the benefits of the sensors extend beyond the selected farmers to their peers or neighbors. Consequently, we conducted additional surveys for the peers or neighbors of each selected farmer who received a sensor. To account for potential variations in the benefits received by neighbors based on the installed sensors of selected farmers, we carefully assigned different IDs and considered both dependent and independent variables. This approach enables us to estimate the "spillover effect" of the sensor beyond the selected households. To calculate the effects of the sensor on irrigation scheduling, irrigation intensities, and other outcome variables, we have separately identified farmers and their neighbors or peers who do not have such sensors. These individuals may be using alternative methods to assess irrigation scheduling. Due to complexities and interactions among the farmers who received orientation and training on the sensors, we were unable to assign households (HHs) fully Program Report | Irrigation scheduling chameleon sensor | Page 5 of 14 randomly among various treatments. As a result, there is some degree of selection bias affecting the household. To avoid these selection bias, we selected only those HHs who have adopted at least one conventional technology to monitor soil moisture. Treatments In the implemented quasi-experimental design, households (HHs) were assigned to five distinct groups to analyze the direct and indirect effects of soil moisture sensor adoption and related training. • Treatment 1: This group consists of households with direct intervention group who are provided with Chameleon Moisture Sensors and receive a comprehensive orientation on their use and benefits. This group serves as the core intervention to measure the direct impact of the technology on irrigation behavior and crop productivity. • Treatment 2: This group consists of immediate neighboring households who can potentially benefit from 1st -layer of spillover effects of the households in Treatment 1. The primary objective of this group is to capture knowledge and practice spillovers through casual social interaction and observation. These households do not receive sensors themselves, and their inclusion is designed to measure the diffusion of irrigation practices within tight-knit local networks. • Treatment 3 (2nd-layer spillover group): This group includes households who are not immediate neighbors of sensor recipients but are part of the broader peer network within the same village or community. This group is instituted to assess secondary spillover effects, understanding how knowledge and practices permeate beyond immediate proximity, potentially through wider social channels. • Treatment 4: This group can be considered pure control groups that consists of households from a similar agro-economic context but located far from the sensor distribution areas. These households receive neither sensors nor any orientation and are expected to rely on conventional irrigation methods. This group provides a baseline to measure the full effect of the intervention against standard practices. • Treatment 5: This group is designed for households with conventional practices to capture their spillover effects from neighbors in T4. This unique group is critical for isolating the spillover effects specifically attributable to the sensor technology, by controlling for any pre-existing spillover effects that might naturally occur from conventional irrigation methods within a community. Treatments 4 and 5 distinct control groups are designed to ensure robust counterfactual analysis. This multi-layered design allowed for a nuanced analysis, isolating the direct effects of the technology, the spillover effects through social networks, and the baseline influence of conventional practices, thereby providing a comprehensive understanding of the intervention's total impact. As shown in figure 1, we prepare five treatments that are as follows: Variables We selected a set of dependent variables that capture key agricultural outcomes to evaluate the impact of the treatments and control variables. We primarily measured irrigation water use, quantified as the total volume of water applied per hectare over the growing season. We analyzed crop yield (Kg/ha), to determine the intervention's effect on productivity. Furthermore, we assessed water productivity, which we calculated as the crop yield per unit of water consumed (kg/m³), to gauge the efficiency of water utilization. Finally, we examined the adoption of sensor-informed scheduling, a binary variable indicating whether farmers made irrigation decisions based on the sensor data. If they do, we asked their number of irrigations to wheat crops. In our analysis, we employed three categories of independent variables to assess their impact on the key outcomes. First, we incorporated a set of treatment dummies to evaluate the intervention's effect. We designated Farmers with Sensors (T1) as the base reference group and compared them against Neighbors of farmers with sensors (T2), Farmers without sensors but who received only training (T3), Farmers using conventional methods (T4), and Neighbors of farmers using conventional methods (T5). Program Report | Irrigation scheduling chameleon sensor | Page 6 of 14 Figure 1. Flow chart in implementing experiments in the study areas. Second, we controlled several sociodemographic and cognitive characteristics. We included binary variables for Gender (using Female as the base) and Caste (using "Others" as the base, compared to Tharu). We also accounted for continuous measures, including the Number of years of schooling, the Number of family members, and years of farming experience. Furthermore, we integrated a binary variable for Generativity (with "No" as the base) and the Number of generations a farmer considered in their decision-making. Finally, we included a suite of structural variables to capture contextual factors. We classified a farmer's Position in the irrigation stream as Head, Middle, or Tail, and used Head as the base category. We also accounted for binary indicators of whether a farmer had access to an Advisory service, used a Krishi meter, relied on a Pump set for irrigation (using Canal as the base), and cultivated Local wheat varieties (using High-yielding varieties as the base). Experimental procedures The experimental procedure was initiated with a targeted capacity development training workshop, a collaborative effort between the International Water Management Institute (IWMI) and the International Maize and Wheat Improvement Center (CIMMYT) under the Cereal Systems Initiative for South Asia (CSISA) project. The primary objective was to enhance farmers' capacity to monitor soil moisture conditions and make informed irrigation decisions, thereby enhancing water use efficiency and crop yields. The training focused specifically on the use of Chameleon Soil Moisture Sensors for irrigation scheduling of wheat in the command area of the Jamara Branch Canal, Rani Jamara Kuleria Irrigation Project (RJKIP). The two-day training event was conducted for 43 participants, including 35 farmers (8 female, 6 youth). The workshop commenced with theoretical sessions led by a principal trainer, who introduced fundamental concepts of soil water dynamics, including field capacity, permanent wilting point, and plant-available water using simplified language and pictorial aids. This was immediately followed by a practical demonstration in the field, where the Chameleon Sensor's color-coded system was explained: “Red” indicating dry soil (below wilting point), “Green” indicating adequate moisture, and “Blue” indicating saturated conditions. A live sensor installation was performed to showcase the proper setup and reading interpretation, fostering direct farmer engagement. Following the training, the operational phase involved the selection and deployment of sensors. Farmers voluntarily identified themselves into two groups and self-selected or selected others as the candidates as sensors recipients (Treatment 1). Because of this fact, the assignment of farmers for treatment 1 cannot be considered random as it poses some degree of self-selection biases. Since the sensors’ benefits are likely to be transferred to the nearest neighbors cultivating same crop with similar contexts, we additionally identified 1-3 neighbors with three neighboring plots (≥ neighbors) corresponding to each sensor distributed to the selected farmers in treatment 1. Note that farmers in treatment 1 were provided with three sensors. Thus, these neighbors are assigned to treatment 2. Those who received training from experimenters and/or experts but could not receive sensors were assigned to treatment 3. Program Report | Irrigation scheduling chameleon sensor | Page 7 of 14 Table 1. Variables and their descriptions. Variables Description Dependent variables Times of irrigation (N) Number of irrigations done by a farmer. Discharge rate (L/sec) A continuous variable indicating the farmer’s total use of water measured in liter/sec. Yield (Mt/ha) A continuous variable for a farmer’s wheat yield for a farmer measured at the end of the experiment. Independent variables Treatment dummies (Base group = Farmers with Sensor (Treatment 1)) Treatment 2 A dummy variable that takes value 1 for a neighboring farmer of a farmer with sensors (T1); Otherwise, 0. Treatment 3 A dummy variable that takes value 1 for a farmer who only takes training (no sensor); otherwise, 0. Treatment 4 A dummy variable that takes value 1 for a farmer practices with conventional methods to decide when to irrigate; Otherwise, 0. Treatment 5 A dummy variable that takes value 1 for a neighbor of farmers with conventional method of irrigation; Otherwise, 0. Sociodemographic variables Gender (Base group = female) Male A dummy variable that takes value 1 if a farmer is male; Otherwise, 0. Caste (Base group = Others) Tharu A dummy variable that takes value 1 if a farmer belongs to Tharu caste; otherwise, 0. Year of schooling (# Year) Continuous variable indicating a farmer’s education in years of schooling. Family members (No) Continuous variables indicating a farmer’s family size. Generativity (base group = No) A dummy variable that takes value 1 if a farmer reports that his/her next generation is willing to continue agriculture; Otherwise, 0. Experience (# Year) Continuous variable indicating a farmer’s agricultural experience in years. Structural and farm variables Position at irrigation stream (base group = Head) Middle A dummy variable that takes value 1 if a farmer’s plot lies in the middle of the nearest irrigation source; otherwise, 0. Tail A dummy variable that takes value 1 if a farmer’s plot lies at the end of the nearest irrigation source; otherwise, 0. Advisory (base group = No) A dummy variable that takes value 1 if a farmer receives agricultural advisory during wheat season; otherwise, 0. Krishi meter (base group = No) A dummy variable that takes value 1 if a farmer receives electricity price subsidy for irrigation; otherwise, 0. Irrigation source (base group = canal) Pump set A dummy variable that takes value 1 if a farmer uses pumping set for irrigation; otherwise; 0. Wheat variety (base group = High yielding variety) Local A dummy variable that takes value 1 if a farmer uses local wheat seeds for cultivation; Otherwise, 0. Experts after the orientation and demonstration of how sensors work, they helped support installing three sensors for each selected household under T1. Each sensor, containing three individual sensors, a card reader, and an instructional flyer, was installed across the farmers' plots to ensure comprehensive spatial coverage of soil moisture conditions and provide hands-on orientation on interpreting the sensor's color-coded data for irrigation decisions. To formalize the commitment and ensure accountability, a handover event was held at the WUA’s office, thereby concluding the deployment phase and initiating the monitoring period. Following the baseline, the specific treatments were implemented in a targeted fashion. A specialized team revisited households in Treatment 1 (T1) to install the Chameleon Soil Moisture Sensors in their wheat fields and provide hands-on orientation on interpreting the sensor's color-coded data for irrigation decisions. HHs in Treatment 3 (T3) received a separate visit where they were given general training on water management best practices, deliberately omitting any mention of sensor technology to isolate the effect of knowledge from the technology itself. Households in the spillover and control groups (T2, T4, T5) received no intervention visits, allowing their existing practices and potential social interactions to form the basis for measuring spillover effects. Table 2: Experimental stages for data collection across the treatments. Stage T1 T2 T3 T4 T5 Baseline Survey Survey Survey Survey Survey Rollout Install Sensor + Training No visit General Training Only No visit No visit Monitoring Survey + Sensor Reading Survey Survey Survey Survey Endline Crop cut + Survey Crop cut + Survey Crop cut + Survey Crop cut + Survey Crop cut + Survey Program Report | Irrigation scheduling chameleon sensor | Page 8 of 14 The pilot field experiment was rolled out in a sequential manner across the treatment groups and the integrity of data collection. The process began with a comprehensive baseline data collection phase, wherein research assistants (hereafter, RAs) visited all 143 pre-selected households across the five treatment arms before the treatments’ rolled out. During these visits, experimenters administered a detailed baseline survey to capture information on sociodemographic, farming experience, cognitive factors, and existing agricultural practices for all HHs belonging to each treatment. In doing so, they strictly maintained protocol by not disclosing household assignments at this stage, thereby preventing any anticipatory changes in farmers’ behaviors. Throughout the growing season, intensive monitoring and measurement were conducted. RAs carried out 35 rounds of periodic observations across all households, systematically recording irrigation activities, including timing, duration, and stated reasons for irrigation. In doing so, soil moisture data were collected at 3-day intervals through farmer visits or phone calls. During each visit, field staff recorded the color reading from the Chameleon soil moisture sensor, the installation and interpretation of which followed established guidelines, following Shrestha et al. (2023). Concurrently, irrigation events were monitored on an event-basis; whenever a farmer irrigated, data on the irrigation method, start and end time, and discharge rate (L/sec) were collected using standardized measurement techniques (CGIAR, n.d.). All data were recorded digitally on the same day of collection using pre-designed Kobo Toolbox forms to ensure consistency and real-time access. For T1 households, RAs additionally documented the real-time readings from the installed sensors during these visits. The procedure culminated at harvest with an endline data collection phase. A trained team performed standardized crop cuts in designated plots for every household to obtain accurate yield measurements, complemented by a final survey to account for any seasonal changes or confounding factors. Results and discussions Given the pilot nature of the study and its limited sample size, our analytical approach emphasizes more on the descriptive statistics over a reliance on statistical significance and associated econometric analyses. We acknowledge that the study is underpowered to detect small-to-moderate effects with conventional significance levels. Therefore, the results are interpreted as indicative of trends and promising directions for future research, rather than conclusive evidence due to the small sample sizes across the treatments. Table 3 presents the summary statistics of key variables across the five treatment groups, establishing the baseline characteristics of the sample. As shown, the sample is predominantly male (91 percent), with households having an average family size of 6.43 members. Farmers are observed to attain a formal education level of 7.70 years of schooling and have 21.50 years of farming experience. The averages for education are 7.81, 6.60, 9.76, 9.29, and 7.53 for T1 through T5, respectively, suggesting some variation in education levels across treatment groups. In terms of access to information and resources, 35 percent of farmers received agricultural advisory services. The primary sources of irrigation are canals and boreholes, with 76 percent of farmers using improved wheat varieties. Overall, the summary statistics suggest that the sample is broadly similar across key demographic and farm management characteristics, though some variation exists in education and access to advisory services, which will be controlled in the subsequent econometric analysis. Table 3 also presents the distributions of mean and median of observed irrigation responses and wheat yield. Results show that on average farmers irrigated their wheat plots 1.50 times and achieved a yield of 3.54 tons per hectare prior to the interventions. Based on the results, a clear hierarchy of irrigation intensity emerges across the treatments. Farmers who directly received sensors (T1) demonstrated the most frequent irrigation, with a mean of 1.81 times. In contrast, their neighbors (T2) showed the lowest irrigation frequency with mean of 1.30, suggesting a limited spillover effect for this specific practice. The intermediate groups, including those receiving IVR and deliberation (T3), exhibited moderate irrigation levels, indicating that while the socio-behavioral intervention had an effect, it was not as strong as the direct sensor-based technology in prompting increased irrigation. Overall, the results for irrigation response and wheat yield demonstrate the following hierarchical relationships - T1 (Sensor) > T4 (Conventional method) > T5 (Neighbors of conventional method) > T3 (Orientation only) > T2 (Sensors’ neighbors). Program Report | Irrigation scheduling chameleon sensor | Page 9 of 14 Figure 2: Box plots illustrating (a) time of irrigation and (b) yield (Mt/ha) by treatments. The box plots in Figure 1 visually summarize the distribution of irrigation frequency and wheat yield across the treatment groups following the intervention. A clear pattern emerges, showing a positive effect of the combined IVR and deliberation treatment (T3). The distribution of irrigation times for T3 is shifted upward compared to the control (T1) and IVR-only (T2) groups, indicating that farmers who received both information and a forum for discussion adopted more frequent irrigation practices. This pattern is mirrored in the yield outcomes, where the T3 group demonstrates a higher median yield and a more compact distribution of higher values compared to other groups. Visually, the T3 box plot shows less dispersion in the upper yield range than the T2 group, suggesting that the addition of deliberation not only increased yields but also made high performance more consistent among participants. These graphical results provide initial evidence that the synergistic effect of information and social learning is a powerful driver of improved agricultural outcomes. To statistically characterize the effects of the treatments on the farmers’ irrigation responses and wheat yield, we run Poisson and Ordinary Least Square (OLS) regression by taking observed times of irrigation and crop cutting yields from the surveys. Table 4 presents the estimated coefficients from Poisson and Ordinary Least Square (OLS) of key independent variables on the times of irrigation in numbers (Panel A) and wheat yield in Mt/ha (Panel B), respectively. Since we basically focus on the treatments’ effects on wheat yields, we only interpret the results from Panel B of Table 4. The results reveal patterns in wheat yield determinants across the specified treatment groups and control variables. As shown in Model 1, the coefficient for T2 shows a statistically significant yield of 3.9 Mt/ha which is equivalent to NPR 12,000, at average price of NPR 40/kg during research time. The negative coefficients for the other treatment groups, particularly T4 (Farmers with conventional methods) and T5 (Neighbors of farmers Program Report | Irrigation scheduling chameleon sensor | Page 10 of 14 Table 3: Summary statistics of key dependent variables and independent variables. Treatment Group Stats Irrigation times Wheat yield Gender Caste Education Experience Generativity Family size Advisory Plot location Krishi Meter Irrigation source Wheat Variety T1 N 21 21 21 21 21 21 21 21 21 19 21 21 21 Mean 1.81 3.89 2.00 5.52 7.81 22.19 0.67 6.52 1.67 2.58 1.38 0.43 0.67 Median 2.00 4.07 2.00 6.00 8.00 25.00 1.00 6.00 2.00 3.00 1.00 0.00 1.00 Std. 0.75 0.92 0.00 1.25 4.86 11.14 0.48 2.73 0.48 0.61 0.50 0.51 0.48 Min 1.00 2.07 2.00 2.00 0.00 4.00 0.00 3.00 1.00 1.00 1.00 0.00 0.00 Max 3.00 5.88 2.00 6.00 18.00 48.00 1.00 15.00 2.00 3.00 2.00 1.00 1.00 T2 N 61 57 62 62 62 61 62 62 62 58 62 62 62 Mean 1.30 3.43 1.89 5.71 6.60 21.85 0.65 6.03 1.15 2.55 1.27 0.29 0.84 Median 1.00 3.56 2.00 6.00 7.50 25.00 1.00 6.00 1.00 3.00 1.00 0.00 1.00 Std. 0.53 0.66 0.32 0.93 4.37 10.70 0.48 2.35 0.36 0.60 0.45 0.46 0.37 Min 1.00 2.00 1.00 2.00 0.00 3.00 0.00 1.00 1.00 1.00 1.00 0.00 0.00 Max 3.00 5.33 2.00 6.00 15.00 43.00 1.00 14.00 2.00 3.00 2.00 1.00 1.00 T3 N 17 17 17 17 17 17 17 17 17 16 17 17 17 Mean 1.47 3.45 1.82 5.12 9.76 22.94 0.88 7.47 1.71 2.13 1.53 0.18 0.76 Median 1.00 3.38 2.00 6.00 10.00 20.00 1.00 8.00 2.00 2.00 2.00 0.00 1.00 Std. 0.51 0.57 0.39 1.58 3.60 10.03 0.33 2.15 0.47 0.72 0.51 0.39 0.44 Min 1.00 1.88 1.00 2.00 2.00 5.00 0.00 3.00 1.00 1.00 1.00 0.00 0.00 Max 2.00 4.50 2.00 6.00 17.00 40.00 1.00 11.00 2.00 3.00 2.00 1.00 1.00 T4 N 21 21 21 21 21 21 21 21 21 20 21 21 21 Mean 1.76 3.58 1.90 5.67 9.29 20.48 0.86 7.48 1.43 2.15 1.38 0.05 0.67 Median 1.00 3.57 2.00 6.00 10.00 20.00 1.00 6.00 1.00 2.00 1.00 0.00 1.00 Std. 0.89 0.42 0.30 1.15 4.83 9.61 0.36 4.46 0.51 0.49 0.50 0.22 0.48 Min 1.00 2.90 1.00 1.00 0.00 5.00 0.00 3.00 1.00 1.00 1.00 0.00 0.00 Max 3.00 4.62 2.00 6.00 18.00 40.00 1.00 21.00 2.00 3.00 2.00 1.00 1.00 T5 N 17 17 17 17 17 17 17 17 17 17 17 17 17 Mean 1.59 3.49 2.00 6.00 7.53 19.24 0.88 5.41 1.24 2.24 1.29 0.12 0.71 Median 1.00 3.40 2.00 6.00 8.00 20.00 1.00 5.00 1.00 2.00 1.00 0.00 1.00 Std. 0.71 0.33 0.00 0.00 5.19 10.03 0.33 1.54 0.44 0.44 0.47 0.33 0.47 Min 1.00 3.15 2.00 6.00 0.00 5.00 0.00 3.00 1.00 2.00 1.00 0.00 0.00 Max 3.00 4.40 2.00 6.00 15.00 35.00 1.00 9.00 2.00 3.00 2.00 1.00 1.00 Total N 137 133 138 138 138 137 138 138 138 130 138 138 138 Mean 1.50 3.54 1.91 5.64 7.70 21.50 0.74 6.43 1.35 2.40 1.34 0.24 0.76 Median 1.00 3.50 2.00 6.00 8.00 22.00 1.00 6.00 1.00 2.00 1.00 0.00 1.00 Std. 0.68 0.65 0.28 1.07 4.64 10.36 0.44 2.79 0.48 0.60 0.48 0.43 0.43 Min 1.00 1.88 1.00 1.00 0.00 3.00 0.00 1.00 1.00 1.00 1.00 0.00 0.00 Max 3.00 5.88 2.00 6.00 18.00 48.00 1.00 21.00 2.00 3.00 2.00 1.00 1.00 Program Report | Irrigation scheduling chameleon sensor | Page 11 of 14 with conventional methods), suggest that farmers without sensor technology, especially those relying on conventional irrigation approaches, experienced significantly lower yields compared to the sensor-equipped baseline group. This model, however, explains only 6.1 percent of the variation in yield (R-squared = 0.061), indicating that treatment assignment alone captures a limited portion of the yield variability. With the incorporation of sociodemographic and cognitive variables in Model 2, the explanatory power of the model substantially improves, with the R-squared value increasing to 0.165. The constant term decreases to 3.173 Mt/ha, suggesting that some of the yield effect initially attributed to the treatment in Model 1 is explained by farmer-specific characteristics such as years of schooling, farming experience, or generational involvement. Although the coefficients of the treatment dummies are not fully reported here, the inclusion of these control variables likely refines the estimated treatment effects, helping to isolate the net impact of the sensor from confounding sociodemographic factors. Model 3, which further incorporates structural and functional variables such as wheat variety (comparing local to high-yielding varieties), demonstrates the highest explanatory power, with the R-squared reaching 0.301. This indicates that nearly 31 percent of the variation in wheat yield is explained by the combined influence of treatment, sociodemographic, and structural factors. The constant in this model is 3.3 Mt/ha. The significant negative coefficient associated with the adoption of local wheat varieties (base: high-yielding variety) underscores the importance of using improved seeds alongside advanced irrigation technology to maximize productivity. The increasing R-squared across the three models confirms that wheat yield is shaped by a complex interplay of technological, human, and structural factors, with sensor-based irrigation serving as a foundational, but not standalone, component of enhancing agricultural productivity. The results were validated with farmers through a learning workshop. Conclusion We designed a proof-of-concept quasi-experiments and conducted with five treatments: (1) Farmers using Chameleon Sensors (T1), (2) Neighboring farmers of T1 (T2), (3) non-neighboring farmers of T1 (T3), (4) Farmers using conventional moisture technology located at distance from T1 (T4), and (5) Neighboring farmers of T4 (T5) with a total of 138 wheat growing farmers of Kailali district of Nepal. We employed with and without subject design and collected periodic datasets for each five days, including baseline and endline survey, to evaluate the effects on irrigation practices and crop yields. Descriptive and regression analyses revealed that farmers using Chameleon Sensors (T1) increased their irrigation intensity by +1 event compared to their counterparts using conventional technology (T4). Furthermore, farmers with the sensors experienced a positive increase in yield by 0.43 MT/ha, equivalent to NPR 12,000 Mt/ha compared to conventional methods. It also appeared that indirect yield increment of 0.36 Mt/ha compared to neighbors (T2) of the sensors adopting farmers, indicating sensors’ potential spillover effects on yield. Overall, the results demonstrated that the adoption of Chameleon Moisture Sensors leads to increased irrigation intensities and yields among wheat farmers, implying their effectiveness over conventional technologies. These findings from the proof-of-concept highlight the potential of Chameleon Moisture Sensors to improve water and wheat yield, suggesting an emerged scope for promoting such technologies along with a full- phased experiment to draw insights to scale up. Program Report | Irrigation scheduling chameleon sensor | Page 12 of 14 Reference Dixit, A., Pokhrel, A., Rai, D. R., Dixit, K., & Upadhya, M. (2009). Living with Water Stress in the Hills of the Koshi Basin, Nepal (0 ed.). International Centre for Integrated Mountain Development (ICIMOD). https://doi.org/10.53055/ICIMOD.508 Duflo, E., Kremer, M., & Robinson, J. (2011). Nudging Farmers to Use Fertilizer: Theory and Experimental Evidence from Kenya. American Economic Review, 101(6), 2350–2390. https://doi.org/10.1257/aer.101.6.2350 Fehr, E., & Gächter, S. (2000). Cooperation and Punishment in Public Goods Experiments. The American Economic Review, 90(4), 980–994. JSTOR. Joshi, J. (2015). Preserving the Hiti, ancient water spout system of Nepal. Proceedings of the Fabric, Threads of Conservation, Australia ICOMOS Conference, Adelaide, Australia, 5–8. Kaini, S., Harrison, M. T., Gardner, T., Nepal, S., & Sharma, A. K. (2022). The Impacts of Climate Change on the Irrigation Water Demand, Grain Yield, and Biomass Yield of Wheat Crop in Nepal. Water, 14(17), 2728. https://doi.org/10.3390/w14172728 Kapri, K., & Ghimire, S. (2020). Migration, remittance, and agricultural productivity: Evidence from the Nepal Living Standard Survey. World Development Perspectives, 19, 100198. https://doi.org/10.1016/j.wdp.2020.100198 Kaveney, B., Khoi, C. M., Minh, D. D., Thi Kim Phuong, N., An Giang, C. D., Barrett-Lennard, E., Khanh, T. D., & Condon, J. (2025). Investigating the suitability of Chameleon soil moisture sensors to improve water use of upland crops under conditions of temporal salinity. Agricultural Water Management, 316, 109602. https://doi.org/10.1016/j.agwat.2025.109602 Koirala, P., Kotani, K., & Managi, S. (2022). How do farm size and perceptions matter for farmers’ adaptation responses to climate change in a developing country? Evidence from Nepal. Economic Analysis and Policy, 74, 188–204. https://doi.org/10.1016/j.eap.2022.01.014 Ladha, J. K., & Singh, Y. (2009). Integrated crop and resource management in the rice-wheat system of South Asia (O. Erenstein & B. Hardy, Eds.). International Rice Research Institute. Nepal, S., Neupane, N., Koirala, S., Lautze, J., Shrestha, R. N., Bhatt, D., Shrestha, N., Adhikari, M., Kaini, S., Karki, S., Yangkhurung, J. R., Gnawali, K., Singh Pradhan, A. M., Timsina, K., Pradhananga, S., & Khadka, M. (2024). Integrated assessment of irrigation and agriculture management challenges in Nepal: An interdisciplinary perspective. Heliyon, 10(9), e29407. https://doi.org/10.1016/j.heliyon.2024.e29407 Parajuli, J., Eakin, H., & Chhetri, N. (2021). Small irrigation users’ perceptions of environmental change, impacts, and response in Nepal. Climate and Development, 13(7), 563–580. https://doi.org/10.1080/17565529.2020.1836468 Porter, J. R., Howden, M., & Smith, P. (2017). Considering agriculture in IPCC assessments. Nature Climate Change, 7(10), 680–683. https://doi.org/10.1038/nclimate3404 Risal, A., Urfels, A., Srinivasan, R., Bayissa, Y., Shrestha, N., Paudel, G. P., & Krupnik, T. J. (2022a). Impact of Climate Change on Water Resources and Crop Production in Western Nepal: Implications and Adaptation Strategies. Hydrology, 9(8), 132. https://doi.org/10.3390/hydrology9080132 Risal, A., Urfels, A., Srinivasan, R., Bayissa, Y., Shrestha, N., Paudel, G. P., & Krupnik, T. J. (2022b). Impact of Climate Change on Water Resources and Crop Production in Western Nepal: Implications and Adaptation Strategies. Hydrology, 9(8), 132. https://doi.org/10.3390/hydrology9080132 Shrestha, S., Adhikari, S., Babel, M. S., Perret, S. R., & Dhakal, S. (2015). Evaluation of groundwater-based irrigation systems using a water–energy–food nexus approach: A case study from Southeast Nepal. Journal of Applied Water Engineering and Research, 3(2), 53–66. https://doi.org/10.1080/23249676.2014.1001881 Thapa, B., & Scott, C. A. (2019). Institutional Strategies for Adaptation to Water Stress in Farmer-Managed Irrigation Systems of Nepal. International Journal of the Commons, 13(2), 892–908. https://doi.org/10.5334/ijc.901 Program Report | Irrigation scheduling chameleon sensor | Page 13 of 14 Appendix Table A1. Estimated coefficients from Poisson and Ordinary Least Square (OLS) of key independent variables on the times of irrigation in numbers (Panel A) and wheat yield in Mt/ha (Panel B), respectively. Variables Times of irrigation (#No) Yield/ha Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Independent variables Treatment dummies (Base group = Farmers with Sensor (T1)) T2 -1.000*** -1.000*** -0.369 -0.451*** -0.389** -0.339* (0.186) (0.264) (0.292) (0.162) (0.165) (0.181) T3 -1.000*** -1.000*** -0.111 -0.435** -0.363 -0.463** (0.240) (0.354) (0.376) (0.207) (0.219) (0.232) T4 -1.000*** -1.000*** 0.092 -0.305 -0.300 -0.453** (0.227) (0.325) (0.356) (0.196) (0.204) (0.219) T5 -1.000*** -1.000*** -0.363 -0.394* -0.403* -0.593** (0.240) (0.343) (0.361) (0.207) (0.211) (0.227) Sociodemographic and cognitive Gender (Base group = female) Male (0.351) (0.366) 0.228 0.170 0.000 0.054 (0.219) (0.230) Caste (Base group = Others) (0.286) (0.300) Tharu -0.000 -0.004 0.284 0.308 (0.021) (0.022) (0.177) (0.189) Year of schooling (# Year) 0.000 -0.031 0.017 0.015 (0.034) (0.034) (0.013) (0.014) Family members (No) -0.000 0.256 -0.015 -0.025 (0.233) (0.266) (0.021) (0.021) Generativity (base group = No) -0.000 -0.000 -0.079 0.048 (0.233) (0.146) (0.168) Experience (# Year) 0.000 -0.001 0.000 (0.006) (0.006) Structural variables Position at irrigation stream (base group = Head) Middle -0.239 0.281 (0.418) (0.257) Tail -0.512 0.015 (0.402) (0.249) Advisory (base group = No) -0.011 0.056 (0.225) (0.143) Krishi meter (base group = No) -0.029 -0.065 (0.211) (0.132) Irrigation source (base group = canal) 0.349 Pump set (0.243) -0.106 -0.262 (0.155) Wheat variety (base group = Others) (0.217) Local -0.239 -0.406*** (0.418) (0.134) Constant 2.000*** 2.000*** 2.675*** 3.886*** 3.173*** 3.308*** (0.139) (0.766) (0.835) Observations 137 137 129 133 133 125 R-squared - - - 0.061 0.165 0.306 Program Report | Irrigation scheduling chameleon sensor | Page 14 of 14 Citation Koirala, P., Khadka, L., Acharya, H.P., Adhikari, S., Lamichhane, H.P., Adhikari, S., Adhikari, S., Shrestha, N., KC, J., Krupnik, T.J. 2025. Do Chameleon Moisture Sensors adopted by farmers increase their irrigation intensity and yields? A proof-of-concept field experiment. Working paper, CIMMYT, Lalitpur, Nepal. Acknowledgements The CGIAR Sustainable Science Program forms a part of CGIAR’s new Research Portfolio, addressing key challenges in agri-food systems by fostering efficient production of nutritious foods and safeguarding the environment to create fair employment opportunities, as we simultaneously tackle climate change, soil degradation, pests, diseases, and desertification. Its research is being implemented by CGIAR researchers from 13 CGIAR Research Centers CIMMYT and IWMI. We would like to thank all funders who supported this research through their contributions to the CGIAR Trust Fund: https://www.cgiar.org/funders/ CIMMYT, IWMI About CGIAR Sustainable Science Program Report This research was conducted as part of the CGIAR Sustainable Farming Science Program. This research is being implemented by CGIAR researchers from CIMMYT and IWMI. CGIAR is a global research partnership for a food-secure future. Its science is carried out by 15 Research Centers in close collaboration with hundreds of global partners. www.cgiar.org Photos @Pankaj Koirala Disclaimer This working paper has not been peer reviewed. Any opinions stated herein are those of the author(s) and do not necessarily reflect the policies or opinions of CIMMYT, IWMI, donors, or partners. This publication is copyrighted by CIMMYT It is licensed under a Creative Commons Attribution – Non-commercial 4.0 International License. To view this license, visit https://creativecommons.org/licenses/by/4.0. Unless otherwise noted, you are free to share (copy and redistribute the material in any medium or format), adapt (remix, transform, and build upon the material) for any purpose, even commercially, under the following conditions: ATTRIBUTION: The work must be attributed, but not in any way that suggests endorsement by CIMMYT, IWMI or the author(s). NOTICE: For any reuse or distribution, the license terms of this work must be made clear to others. Any of the above conditions can be waived if permission is obtained from the copyright holder. Nothing in this license impairs or author’s moral rights. Fair dealing and other rights are in no way affected by the above. The parts used must not misrepresent the meaning of the publication. CIMMYT would appreciate being sent a copy of any materials in which text, photos, etc., have been used. ©2025 CIMMYT Key Words: Chameleon Moisture Sensor; Agriculture; Irrigation Intensity; Yield; Farmers; Experiment; Nepal Partners About CGIAR Sustainable Farming Science Program The CGIAR Sustainable Farming Science Program will address key challenges in agrifood systems by fostering efficient production of nutritious foods and safeguarding the environment to create fair employment opportunities, as we simultaneously tackle climate change, soil degradation, pests, diseases, and desertification.