Assessing the Representativeness of Irrigation Adoption Studies: A Meta-Study of Global Research

Abstract

For decades, nations around the world have been promoting irrigation expansion as a method for improving agricultural growth, smoothing production risk, and alleviating rural poverty. Despite its apparent advantages, suboptimal adoption rates persist. According to the existing literature, determinants of irrigation adoption are often highly dependent on cultural, contextual, and/or local institutional factors. Yet, studies from diverse geographies identify a consistent set of factors. Thus, to be able to make generalizable inferences from such studies, a global geographic representativeness assessment of irrigation adoption studies was conducted to determine whether identified factors influencing irrigation were the result of geographic, epistemological, or disciplinary biases. The results indicate that multiple geographic biases exist with respect to studying farmers’ irrigation adoption decision-making. More research on this topic is being conducted in regions that have little to a high percentage of irrigation (>1%), are readily accessible, receive moderate amounts of average annual rainfall, and have moderate amounts of cropland cover. The results suggest the need to expand research efforts in areas with little to no irrigation to identify constraints and help accelerate economic growth, poverty reduction, and food and livelihood security for rural communities in these regions.

1. Introduction

One of the major global environmental issues confronting us today is climate change, which threatens our ability to meet the growing population demands for basic resources like food and water [1,2]. Due to its inherent link to natural resources, agriculture is highly sensitive to changing climatic conditions [3] and is among the most vulnerable sectors to climate change risks and impacts [4]. Changes in temperature and rainfall patterns will have direct and indirect impacts on our food systems, ranging from reduced crop production to volatility in markets and food prices [5,6]. Even though food production trends of the last 40 years have more or less kept pace with the rising food demands [7], pressure on our food systems will only intensify with changing consumption patterns, lifestyles, and diets in the coming years [1,8]. Additionally, in most developing countries, agriculture provides the main livelihood and employment opportunities for rural populations and contributes significantly to the national GDP [9]. Therefore, any reductions in production will impact agricultural economies and challenge the resilience of agricultural-dependent communities as well [9,10]. Hence, there is a need to strengthen local capacity to deal with forecasted and/or unexpected climatic changes [3], and this requires adaptation [11].

Adaptation is considered a vital component of any policy response to climate change in addition to mitigation [4], and often involves changes in processes, practices, or structures to reduce potential adverse impacts [3]. Sakschewski et al. (2014) in their assessment of agricultural production argued that production increases can be accomplished either by increasing land productivity or by increasing land resources, but since cropland expansion is limited, engineered or technological adaptive responses remain the most common in this sector [12]. One such adaptation strategy is to augment rainfed production with the use of irrigation [13]. Irrigation has the potential to buffer climate stress and increase production on existing agricultural lands, smooth production risks, and improve the growth of agricultural economies [9,13,14,15]. According to the UN Food and Agriculture Organization (FAO), the global area equipped for irrigation worldwide increased from 184 million ha in 1970 to 324 million ha in 2012 [16]. Much of this expansion has occurred in developing Asian countries [17], with China having the largest irrigated area in the world, followed by countries like India, the United States of America, Pakistan, and Iran in the top five [18,19].

Despite the multiple benefits, irrigation adoption among farming communities has been slow or the long-term investments needed delay its adoption [20]. This is because adoption of any technology, in general, is a complex sociological phenomenon [21] that involves a large number of factors affecting the adoption decision [21] and is seldom rapid [20]. Globally, many attempts from different disciplinary backgrounds have been made to identify the factors that act as barriers to irrigation technology adoption by farmers [22,23,24,25,26,27]. Studies from diverse geographies identify a consistent set of factors, with the cost of technology cited as the most common barrier to its adoption/uptake [28]. However, the existing literature also asserts that the determinants of irrigation adoption are often dependent on local culture, context, and/or policies [29]. For example, Alabama in the south-eastern U.S. receives an average of 55 inches of precipitation annually which allows for a long growing season in the state. However, the recent increase in flash drought instances within the state is a cause of worry for those practicing rainfed agriculture, especially the small farm owners, making them the most vulnerable to these changing climatic conditions (For more details see the U.S. Drought Monitor for Alabama from the year 2000–Present available at: https://www.drought.gov/states/alabama, accessed on 15 October 2022). Accordingly, this identification of factors influencing irrigation adoption across a wide range of geographic contexts will be useful when climate change necessitates adaptation in such unprecedented areas.

One explanation for this disconnect concerning the different factors affecting irrigation adoption, which we explore in this paper, is that the geographic contexts in which irrigation adoption studies are often conducted might be biased, and this bias has influenced the set of factors identified as having explanatory power. If such a bias exists, it would not be unique to irrigation adoption studies. For instance, Martin et al. (2012) found the global distribution and context of ecological field study sites to be biased toward more accessible locations with limited human influence. According to the authors, the geographical context of selected study sites greatly influenced the observations made within these locations [30]. Therefore, to better understand the reasons as to why a farmer chooses to adopt or not adopt irrigation, it is first essential to recognize the global extent and context-dependency of irrigation adoption. This can be achieved through a geographic representativeness analysis. Using this analytical approach, the representativeness of studies examining factors affecting the diffusion and adoption of irrigation by farmers from around the world will be assessed to determine whether the identified factors (influencing irrigation adoption) from a set of case studies selected through a systematic review were the result of certain geographic biases or not. Accordingly, to identify these potential biases, we test the following two hypotheses:

H₁. 

The geographic context of irrigation adoption studies is biased towards locations with substantial levels of existing irrigation, relatively low annual precipitation, and greater accessibility to markets.

H₂. 

The same factors (affecting farmers’ decision-making) are observed regardless of the geographic context of these studies.

Thus, the goal of this review is to understand whether the apparent consistency of factors influencing irrigation adoption is the result of the geographic contexts in which it is studied. Given the emerging challenges presented by climate change, we suspect that there are settings in which irrigation (and the study of its adoption) is currently limited but would be beneficial (i.e., improved yields, profits). If this is the case, then the set of factors influencing the irrigation adoption process may be different than in contexts with established irrigation practices. To answer this question, we narrowed our review to articles that explicitly addressed the irrigation adoption process, rather than broader investigations of the adoption of climate-smart agriculture or best management practices, e.g., [31] or those that assessed the benefits of irrigation adoption, e.g., [32].

This paper is structured as follows. Section 1.1, Section 1.2 and Section 1.3 give a brief overview of the motivations behind irrigation adoption and the technological and theoretical perspectives commonly used to study the adoption process. Section 2 describes the procedure followed for this systematic review, followed by the presentation and discussion of the results of the geographic representativeness and adaptation factors’ analyses in the subsequent sections and some concluding remarks.

1.1. Why Irrigation?

Irrigation refers to the systematic and artificial application of water to plants at regular intervals to assist in the growing of crops and maintenance of landscapes [33,34,35] and is usually classified as surface, sprinkler, and micro-irrigation [35,36]. Irrigated agriculture, which accounts for more than 70% of total global freshwater withdrawals [15,37], provides for about 40% of the world’s agricultural production [38] from less than 20% of its area [39,40,41]. Asia continues to contribute the largest share of total irrigated area, followed by Latin America, while sub-Saharan Africa only contributes 6% of its cultivated area to irrigation [18,38].

Irrigation use increases and stabilizes crop production in areas that do not receive enough precipitation [42,43,44], and has helped shape the economies of many semi-arid and arid regions around the world [45,46]. It also contributes toward income stabilization of dependent communities by improving agricultural growth and smoothing production risk [47,48,49]. According to a study by Bhattarai et al. (2007) [50], irrigation use can provide direct benefits like increased crop production that go to individual farm owners and/or entire community(s), and indirect benefits that are accrued to the wider sectors of the economy. There can also be spillover effects, which is brought by the increased household spending in the local economy due to enhanced income and employment as a result of increased land productivity made possible by irrigation [48].

1.2. Technology Adoption and Related Theories

Adoption is the decision to use a particular technology or innovation by an individual [51,52], which then leads to its diffusion or dissemination within a social system [52,53]. There exists a plethora of literature on factors that determine the adoption of a technology. Various researchers even define ‘technology’ itself in different ways and based on their definitions and disciplinary backgrounds use different theories or models to study its adoption [29,54,55,56,57]. For instance, in marketing research, the purchase of a technology is often the focus rather than its actual use [53]. Within agriculture, scholars have commonly used economic models and theories to explain individual technology adoption decisions [57,58], which allow for only rational and objective decision-making behaviors of farmers, rather than their perceptions, and assume that they adopt technology only for profit or utility maximization [59,60]. Alternatively, some research has shifted analytical focus to the role of individual knowledge, perceptions, and/or attitudes in the decision-making process, which in turn are conditioned by extrinsic factors, such as characteristics of the individual (~age, gender, education levels, etc.) and their external environment [29,61]. While others have taken a relatively more macro-perspective as they focus not only on the individual but also on the characteristics of the technology in question and the infrastructure needed for its successful diffusion to comprehensively understand its spread across the entire society (or market) over time [51,62,63]. A more recent strand of literature on agricultural technology adoption has also included the role of social networks in influencing the adoption of agricultural technologies [64,65,66,67,68]. Another key element associated with the adoption decisions is that of uncertainty or risk, which refers to the suitability of technology with an individual’s characteristics including his/her experience or skillset, and with their local conditions (~agronomic, economic, and/or climatic) [57,69].

Since there exists no single model for understanding the decision-making processes in which an individual engages before adopting a certain technology, adoption is examined through a combination of research paradigms [55,57]. Moreover, these studies mostly utilize regression models to explain the uptake of technology as a function of several independent variables [70,71] including personal characteristics, preferences, individual attitudes, economic or institutional constraints, that are gathered either through census data, surveys, or personal interviews or a combination of it [29].

1.3. Factors Affecting Irrigation Adoption

Studies suggest that uptake of a new technology is rarely rapid, particularly among small farmers in developing countries [61,72,73], and with a wide range of factors acting as possible deterrents [72,74,75]. For example, a study in Burkina Faso (West Africa) of 629 farmers highlighted the need for farmers’ training and information dissemination on irrigation to increase adoption [76]. Another study investigated factors affecting the adoption and application of sprinkler irrigation technology by farmers in the county of Famenin, Iran, and showed that the adoption was influenced by both environmental factors, such as farm size, access to water, water quality, and non-environmental factors, including workforce number in the family, employment diversity, and participation in extension education and courses on agricultural water management [77]. Another study investigated the sources of variation for on-farm irrigation systems across producer fields in Nebraska (USA) [42]. Their findings showed that biophysical factors such as soil, crop type, and weather, explained about half of the observed variations in field irrigation. However, the rest of the variations remained unexplained, suggesting that both producer behavior and skills played a significant role in shaping these decisions. Another study looked at the effect of production risk on irrigation technology choice among small-scale farmers in Chile and their results indicated that more educated farmers, with credit access, receiving extension services, and living in communes with other adopters were more likely to use modern irrigation techniques [78]. Another study of 1500 farmers from Henan Province in China found that the farmers who believed in climate change adopted adaptation measures like irrigation to respond to and mitigate its negative impacts [79]. Thus, understanding the kinds of factors influencing adoption decisions is crucial not only for the propagators of these technologies to increase the likelihood of their adoption but also for identifying the overall determinants of agricultural growth and development [80].

2. Materials and Methods

2.1. Data Collection

2.1.1. Literature Search Strategy

A literature search was conducted using Science Direct and Scopus databases. The search was limited to only research articles written in the English language and published in peer-reviewed journals between 2000 and 2021. Articles prior to 2000 were excluded as the global irrigation dataset used in this analysis is based on the nationally reported statistics from around the year 2000 (more details about this dataset are presented in

Table 1. Description and sources of all the datasets used in this analysis.

Dataset Name	Description	Source
Global Administrative Areas (GADM)	A spatial database of the location of administrative areas of all countries, at all levels of sub-division.	GADM (2018–2022) [82]
GLOBE Land Units (GLUs)	GLUs are equal-area hexagonal cells that cover the Earth’s land surface and are based on the geodesic Discrete Global Grid (DGG) system of Kevin Sahr (2003).	GLOBE (2012)
Average Annual Precipitation	Average annual precipitation (mm/year) from 1950–1999. Native resolution is 30 arcminutes projected in Geographic Coordinate System WGS 1984.	Willmott & Matsuura (2001) [83]
Percent Crop Area	Percent crop land cover area per grid cell derived from HYDE (History Database of the Global Environment) land cover data. Native resolution is 0.5° projected in Geographic Coordinate System WGS 1984.	Klein Goldewijk et al. (2011) [84]
Market Access Index	Global grid of a normalized market access index based on travel time to cities with populations of at least 50,000 and 750,000.	Verburg et al. (2011) [85]
Percent Area Equipped for Irrigation	Global map of irrigation areas showing the amount of area equipped for irrigation around the year 2000 in the percentage of the total area on a raster with a resolution of 5 min.	Siebert et al. (2005) [86]

). Moreover, this also reflects the broader trends in irrigation adoption globally because the percentage of reported data on irrigation use from around the world is largest from the year 2000 onwards compared to the earlier years [81]. The steps taken in the search and screening process are presented in Figure 1. First, a target set of 10 articles containing both ‘true positives’ and ‘true negatives’ was assembled from a wide range of disciplines to represent the full range of publications in this research domain and assemble a set of search keywords. Target set articles are listed in Appendix A. Different keywords such as irrigation, technology adoption, agriculture, farmer decisions, water management, and climate change adaptations were combined using Boolean operators to download relevant studies. The specific search terms used were: ((“irrigation”) AND (“technology” OR “adoption”) AND ((“reasons and constraints”) OR (“attitudes”) OR (“drivers”) OR (“perception”) OR (“barriers”)) AND (“climate change adaptation” OR “climate smart agriculture” OR “climate change” OR “adaptive capacity”) AND ((“drought”) OR (“water management practices”)) AND ((“farmers”) OR (“farmer decisions”))).

2.1.2. Selection of Case Studies

After the literature search, the resulting dataset consisted of 438 publications. The next step was article screening to identify case studies that should be used in this meta-study. Both study titles and abstracts were checked and critically reviewed for suitability for this analysis. Articles were excluded if they did not (1) investigate the different factors/reasons affecting technology adoption within the agricultural sector, and (2) present an assessment of farmers’ views or opinions. Conference proceedings, grey literature, reports, and duplicate articles were also excluded from the dataset. The initial screening reduced the number of eligible articles to 119. The second round of screening was performed using the full text of each remaining article. Articles were primarily screened to determine specifically if irrigation adoption by farmers was studied or not, irrespective of the type of irrigation system. For instance, many studies examined the adoption of several different agricultural practices together, in the form of climate change adaptation strategy, conservation agriculture, or as sustainable farming practices adopted by farmers including high-yielding crop varieties, different soil, and water management practices see, e.g., [87,88,89]. All the studies that did not include irrigation as one of the technologies or practices being studied were discarded. Moreover, studies that were conducted at a very large-scale and reported aggregated results (e.g., for entire U.S. mid-west region [90] or 11 African countries together [91]), were excluded to ensure comparability of results, since the goal was to examine the geographic contexts of these studies that would otherwise have been difficult to capture. Additionally, studies that investigated the benefits of irrigation adoption, assessed its impact on crop production under climate change, or estimated future adoption rates were also not considered, e.g., [92,93,94,95]. As a result, 50 case studies, which passed the inclusion and exclusion criterion were selected and used in this meta-study. A complete list of the studies included in this review is also provided in Appendix A.

2.2. Data Analysis

A representativeness analysis provides a robust statistical test to enable the user to investigate potential geographic biases within a collection of primary data observations (e.g., case studies) [96]. Using this analytical approach, for a given global variable of interest (e.g., average annual precipitation), the frequency distribution of the global variable within a user-specified geographic extent was compared with the frequency distribution of the observations in the sample collection, and the degree to which the sample collection’s distribution is representative of the distribution of the global variable was quantified [96,97]. The null hypothesis for this analysis was that the frequency distributions of the global variable and sample collection are not statistically different. If the null hypothesis can be rejected with a low probability of type I error, then the sample can be declared as significantly biased. To enable comparability between values of the global variable and sample collection observations, which might include case study geographies of diverse extents, the standardized, hexagonal, and equal-area geographic units from the GLOBE system were used, known as GLOBE land units (GLUs). The degree of representedness (r) was then computed with a chi-squared (χ2) test and was characterized as follows:

r = 0 if f_e(g_v) = f_o(g_v)
−(1 − p) if f_e(g_v) > f_o(g_v)
(1 − p) if f_e(g_v) ≤ f_o(g_v)
undefined if f_e(g_v) = 0 ∧ f_o ≠ 0

where f_e(g_v) was the expected frequency of the bin to which GLU g belonged (calculated from the population set), f_o(g_v) was the observed frequency of that bin (calculated from the sample set), and p was the p-value for the χ² test. The range of r is between [−1 to 1], with 0 indicating perfect representedness, negative numbers indicating under-representedness, and positive numbers indicating over-representedness [96].

Several data preparation steps were followed to produce the sample and global (~population) datasets. Table 1 describes all the datasets used for this analysis. After shortlisting the case studies, the locations of the study sites (total = 53) mentioned in each of the selected 50 articles were mapped using the shapefiles of administrative boundaries from the GADM dataset in ArcGIS Pro software (see Figure 2). Next, the global GLU feature layer obtained from GLOBE was filtered using several context variables (see

Table 2. Description of all the filters applied to the GLU layer obtained from GLOBE.

Variable Name	Description and Source	Filter(s) Applied
Olson Biomes	Terrestrial ecoregions of the world defined by climate, geology, and evolutionary history from Olson et al. (2001) [98]	Biomes—Boreal forests and Tundra removed.
Average Annual Temperature	Average annual temperature (°C) from 1951–2002. Values range from −28 °C to 31 °C. See [97] for more details.	Values greater than 28.57 °C and less than −12.2058 °C removed.
Average Annual Precipitation	Average annual precipitation (mm yr−1) from 1950–1999 [83]. Values range from 0–10,572 mm/year.	Values greater than 2948.79 mm/yr and less than 30.0 mm/yr removed.
Population Density	Global model of population density from HYDE population model 2000 [84]. Values range from 0–62,018.	Values equal to ‘0’ removed.
Percent Land Area	Percentage of land area contained within each GLU cell based on LandScan 2007 by Oak Ridge National Laboratory (2008). See [97] for details. Values range from 0–100%	Values less than 1 removed.
Percent Crop Area	Percent crop land cover area per grid cell derived from HYDE land cover data (2000) [84]. Values range from 0–100%	Values equal to ‘0’ removed.
Slope Suitability Class	Global grid of land suitability for agriculture based on combined slope constraints [99]. Total 8 classes.	Classes 7 and 8 corresponding to ‘Very Frequent Severe Constraints’ and ‘Unsuitable for Agriculture’, respectively, removed.

) to restrict the global dataset to the expected geographic extent of agricultural areas. Case study locations were also intersected with the filtered GLU layer to form the sample dataset and to maintain a similar unit of analysis for both the layers. For each GLU, values of three variables—average annual precipitation (mm/year), percent crop area, and market access index were calculated. For the area equipped for irrigation (%) variable, mean values were computed using zonal statistics within each GLU for both the above feature layers. The extent/range of the selected four variables within both the global and sample layers are shown in Figure 3, Figure 4, Figure 5 and Figure 6. For each of these four variables, these two datasets were divided into different intervals or bins. The binning strategy was kept the same as their source datasets (see Table 1 for dataset details) except for average annual precipitation variable for which a geometric interval was used. Finally, Pearson’s χ² test for the independence of two datasets was conducted to compare the frequency distributions of the sample and global datasets for each of the selected four variables to determine the geographic representativeness of the assembled case studies on irrigation adoption and answer the first hypothesis.

To test the second hypothesis, first a list of factors reported to influence irrigation adoption decisions of farmers was compiled from the selected case studies. Factors affecting farmers’ adoption decisions are often classified into broad clusters like—financial/economic, physical, institutional, and individual characteristics, but depending on the researchers’ preferences and disciplinary backgrounds this categorization can vary [57,70]. For our study, based on the background literature, the different (influential) factors were clustered into seven broad categories—biophysical, demographic, geographic, technology-specific, social capital, farm enterprise, and institutional factors (Figure 7). Individual factors were coded using these broad categories for frequency analysis. Next, the relationships between these seven factor categories and their corresponding geographical contexts were examined using correspondence analysis. Correspondence analysis (CA) is a multivariate statistical technique and a useful visualization tool for summarizing, examining, and displaying the relationships between categorical data in a contingency table [100,101]. No underlying distributional assumptions are needed for this analysis and therefore, it accommodates any type of categorical variable—binary, ordinal, or nominal [102]. Moreover, the row and column points from the contingency table are shown together on a multi-dimensional map called biplot, which allows for easier visualization of the associations among variables [103,104]. CA uses the chi-square statistic to measure the distance between points on the map, but it does not reveal whether these associations are statistically significant and is therefore used only as an exploratory method [104].

All the above-mentioned statistical tests were conducted and developed in the PyCharm IDE (Integrated Development Environment) using pandas, Matplotlib, Prince, and Scipy Stats libraries.

3. Results

3.1. Geographic Representativeness of Irrigation Adoption Studies

Geographic representativeness analyses were conducted for the percentage of GLU area equipped with irrigation, percentage of GLU area in cropland, average market accessibility, and average annual precipitation. Pearson’s χ² tests for independence for each of the four variables (

Table 3. Pearson’s χ² test results with percentage of area equipped for irrigation variable.

Bins	Frequency		χ2 Statistic	p-Value *	Representedness Degree
	Observed	Expected			r-Value **	Representedness
0.0–0.1	11469	20997	4951.870159	0.0	−1	Highly under
0.1–1.0	3634	5705	524.5160253	4.41 × 10−116	−1	Highly under
1.0–5.0	5375	4079	203.1462568	4.30 × 10−46	1	Highly over
5.0–10.0	3571	1769	654.2752213	2.63 × 10−144	1	Highly over
10.0–20.0	3875	1710	906.3700068	4.05 × 10−199	1	Highly over
20.0–35.0	3589	1236	1225.836112	1.48 × 10−268	1	Highly over
35.0–50.0	2482	719	1013.405549	2.19 × 10−222	1	Highly over
50.0–75.0	2536	694	1096.828508	1.62 × 10−240	1	Highly over
75.0–100	749	369	130.4361647	3.29 × 10−30	1	Highly over
Total Frequency	37280	37280	Dist.χ2= 24137.36522	Dist. p-value = 0.0	Diagnosis: Highly biased

* At 0.05 significance level; ** r-value calculation based on criteria defined in Section 3.2.

,

Table 4. Pearson’s χ² test results with percentage of cropland variable.

Bins	Frequency		χ2 Statistic	p-Value *	Representedness Degree
	Observed	Expected			r-Value **	Representedness
0.0–0.1	6219	15900	5962.41	0.0	−1	Highly under
0.1–0.2	4595	5556	104.6993983	1.42 × 10−24	−1	Highly under
0.2–0.3	5366	3863	278.0312679	2.02 × 10−62	1	Highly over
0.3–0.4	6522	3245	1259.874418	5.93 × 10−276	1	Highly over
0.4–0.5	6063	2856	1304.727198	1.06 × 10−285	1	Highly over
0.5–0.6	3990	2551	345.7343162	3.60 × 10−77	1	Highly over
0.6–0.7	2870	1954	185.2504582	3.46 × 10−42	1	Highly over
0.7–0.8	2002	1319	146.4197305	1.05 × 10−33	1	Highly over
0.8–0.9	203	416	73.20047746	1.17 × 10−17	−1	Highly under
0.9–1.0	372	543	31.96753669	1.57 × 10−08	−1	Highly under
Total Frequency	38202	38202	Dist.χ2= 15313.11155	Dist. p-value = 0.0	Diagnosis: Highly biased

* At 0.05 significance level; ** r-value calculation based on criteria defined in Section 3.2.

,

Table 5. Pearson’s χ² test results with market accessibility variable.

Bins	Frequency		χ2 Statistic	p-Value *	Representedness Degree
	Observed	Expected			r-Value **	Representedness
0.0–0.1	10902	20736	5215.919285	0.0	−1	Highly under
0.1–0.2	4340	3907	25.36746844	4.74 × 10−07	1	Highly over
0.2–0.3	4618	3186	292.2496414	1.61 × 10−65	1	Highly over
0.3–0.4	4448	2677	484.9267588	1.81 × 10−107	1	Highly over
0.4–0.5	3919	2075	614.9158442	9.54 × 10−136	1	Highly over
0.5–0.6	3187	1481	663.2799333	2.89 × 10−146	1	Highly over
0.6–0.7	2398	1335	317.6479106	4.71 × 10−71	1	Highly over
0.7–0.8	2228	1172	342.6063765	1.73 × 10−76	1	Highly over
0.8–0.9	1503	994	106.8413592	4.82 × 10−25	1	Highly over
0.9–1.0	659	640	0.253736593	0.61	0.4	Well-represented
Total Frequency	38202	38202	Dist.χ2 = 12191.37033	Dist. p-value = 0.0	Diagnosis: Highly biased

* At 0.05 significance level; ** r-value calculation based on criteria defined in Section 3.2.

and

Table 6. Pearson’s χ² test results with average annual precipitation (mm/year) variable.

Bins	Frequency		χ2 Statistic	p-Value *	Representedness Degree
	Observed	Expected			r-Value **	Representedness
30–287	1680	3293	558.91	1.45 × 10−123	−1	Highly under
287–463	3449	5289	437.02	4.83 × 10−97	−1	Highly under
463–584	6537	4397	488.33	3.28 × 10−108	1	Highly over
584–666	3444	2457	178.54	1.01 × 10−40	1	Highly over
666–786	3819	2999	108.02	2.66 × 10−25	1	Highly over
786–962	6864	3658	1132.16	3.39 × 10−248	1	Highly over
962–1219	6137	4917	157.17	4.71 × 10−36	1	Highly over
1219–1595	5319	5660	12.30	4.54 × 10−04	−1	Highly under
1595–2145	835	3261	1517.03	0.0	−1	Highly under
2145–2949	118	2270	2000.03	0.0	−1	Highly under
Total Frequency	38202	38202	Dist.χ2= 10070.33756	Dist. p-value = 0.0	Diagnosis: Highly biased

* At 0.05 significance level; ** r-value calculation based on criteria defined in Section 3.2.

) found that the observed (~sample) distributions were statistically different from the expected distributions.

The observed frequencies of the two lowest percent areas of irrigation were significantly lower than their expected frequencies (see Figure 8) and highly underrepresented (Table 3). Similarly, the remaining seven bins were highly over-represented in this collection as the observed frequencies of these bins were higher compared to their corresponding expected frequencies. Case studies of irrigation adoption were thus biased toward areas of existing agriculture, and studies were generally more over-represented as the area equipped for irrigation increased.

Similarly, in the case of the percent cropland variable (Table 4 and Figure 9), four out of ten bins (with very low and high cropland cover) were highly underrepresented. Irrigation adoption studies were more frequently conducted in areas with moderate extents of agricultural land use, and thus biased against areas of low or high cropland. This likely had implications for the irrigation adoption decisions studied. Locations that were dominantly or exclusively agricultural likely had better support services and infrastructure and did not compete with other land uses, which would presumably facilitate irrigation adoption. Conversely, farmers in low agricultural areas face the opposite conditions and may experience more barriers to irrigation adoption.

In the case of the market access index, most of the bins (8 out of 10) were highly over-represented (Table 5 and Figure 10) with a bias toward areas having moderate-high market access. Market signals that might favor irrigation adoption were likely dampened in low market accessibility areas, which may not have been enough to overcome economic barriers to irrigation adoption. Additionally, remote areas are generally understudied due to access difficulties for researchers [30]. As a result, irrigation adoption studies were skewed toward locations with greater accessibility, including a well-represented sample of the most accessible locations.

Finally, regions receiving moderate average annual rainfall (463–1219 mm/year) were highly over-represented, while regions with very low and high average annual rainfall were under-represented and understudied (Table 6 and Figure 11). The underrepresentation of low rainfall areas was surprising, but these may be neglected by irrigation adoption studies due to the necessity of irrigation and limited variability in decision-making. The limited sampling of high precipitation areas was not surprising, since areas receiving high average annual precipitation were more likely associated with rainfed agriculture. However, such areas may also include those in which seasonal drought is a concern despite high aggregate rainfall (e.g., humid southeast United States) and which potentially have unique sets of adoption decision factors.

3.2. Similarity of Irrigation Adoption Factors across Geographic Contexts

Most of the studies conducted in low irrigated regions of the world and that were highly underrepresented in this collection were from countries located in Africa and Latin America (see

Table 7. Distribution of number of cases based on percentage of irrigation.

Percentage of Irrigation	No. of Cases	Degree of Representedness
0.0–0.1	4	Highly under
0.1–1.0	14	Highly under
1.0–5.0	7	Highly over
5.0–10.0	9	Highly over
10.0–20.0	6	Highly over
20.0–35.0	7	Highly over
35.0–50.0	2	Highly over
50.0–75.0	4	Highly over

and Figure 12). Further,

Table 8. Frequency of broad clusters of factors affecting irrigation adoption.

	Study Sites	Asia (n = 20)	Africa (n = 20)	Australia (n = 1)	Europe (n = 2)	Latin America (n = 6)	North America (n = 2)	Near East (n = 2)	All Cases (n = 53)
Factors
SINGLE-FACTOR
B		0	1	0	0	0	0	0	1
D		1	0	0	0	0	0	0	1
TWO FACTORS
B, I		0	1	0	0	0	0	0	1
D, S		2	1	0	0	0	0	0	3
B, D		1	0	0	0	0	0	0	1
THREE FACTORS
D, S, T		0	0	0	1	0	0	0	1
B, D, G		0	1	0	0	0	0	0	1
I, S, T		0	1	0	1	0	0	0	2
B, D, F		0	0	0	0	1	0	0	1
D, F, S		1	0	0	0	0	0	1	2
B, D, T		0	1	0	0	1	0	0	2
B, S, T		0	1	0	0	0	0	0	1
D, G, S		1	0	0	0	0	0	0	1
B, D, S		1	0	0	0	1	0	0	2
FOUR FACTORS
B, D, F, S		3	2	0	0	0	0	1	6
B, D, F, G		1	0	0	0	0	0	0	1
B, D, I, S		0	0	0	0	0	1	0	1
D, G, I, S		0	1	0	0	0	0	0	1
D, G, S, T		0	1	0	0	0	0	0	1
B, F, G, S		0	1	0	0	0	0	0	1
D, F, I, S		0	1	0	0	0	0	0	1
D, F, S, T		0	0	1	0	1	0	0	2
B, D, G, T		0	1	0	0	0	0	0	1
D, F, G, S		1	1	0	0	1	0	0	3
FIVE FACTORS
B, D, G, I, S		0	1	0	0	0	0	0	1
B, D, F, I, S		2	2	0	0	0	1	0	5
B, D, F, G, S		2	0	0	0	1	0	0	3
B, D, G, S, T		0	1	0	0	0	0	0	1
B, F, G, I, S		0	1	0	0	0	0	0	1
B, D, F, G, T		1	0	0	0	0	0	0	1
SIX FACTORS
B, D, F, G, I, S		3	0	0	0	0	0	0	3
TOTAL CASES		20	20	1	2	6	2	2	53

B = Biophysical; D = Demographic; F = Farm Enterprise; G = Geographic; I = Institutional; S = Social Capital; T = Technology-specific.

lists the different clusters of factors affecting irrigation adoption identified from the case studies, broken down by world regions. The frequency of each of the causal factors as reported in the case studies are provided in this table as an absolute number (this method of frequency analysis is based on the Geist & Lambin (2004) study). Only two case studies had a single variable (factor category) that explained farmers’ decision-making regarding irrigation adoption, thus suggesting that the decision to adopt (or not) irrigation is best explained using a combination of factors (see Table 8). Dominating the broad clusters of factors affecting irrigation adoption decisions of farmers was the combination of—Biophysical, Demographic, Farm Enterprise, and Social Capital factors (B, D, F, S), followed by the cluster with Biophysical, Demographic, Farm Enterprise, Institutional, and Social Capital factors (B, D, F, I, S), with clear regional variations as both these clusters feature mainly in case studies from Asia and Africa. Cases from both these regions share a greater number of factors in common as compared to other regions. Demographic category that includes factors like age, gender, household size, and more (see Figure 7 for more details) featured the most, while both institutional and technology-related factor categories were least observed within these case studies. Further, demographic and social capital related factors together formed the most robust combination, although one that often occurred in combination with other clusters.

Additionally, the CA biplot between the study regions and set of causal factors (Figure 13) was also prepared to visually identify and understand these regional variations. In this symmetric scatterplot, component 0 was represented by the horizontal axis and component 1 by the vertical axis. Together both the components explained about 45.68% of the variance/inertia in this dataset. Europe had high positive values along component 0 (horizontal axis), while Australia had high positive values along the vertical axis. Similarly, North America had high negative values and low positive values along vertical and horizontal axis, respectively. Moreover, from just visually inspecting this biplot it was evident that the set of factors influencing irrigation adoption (of farmers) in cases from Europe, Australia and North America were very different from each other as they were placed in separate quadrants and were also far from the origin. Australia and Latin America study regions were placed in the same quadrant and thus, shared similar profiles, i.e., within both these regions similar combination of causal factors was observed as compared to say Europe or other regions (see Table 8 for more details). Further, the map also revealed that irrigation adoption by farmers from case studies in Europe was explained by a combination of only demographic, social capital, institutional, and technology-specific attributes. Whereas in case of North America, the strongest association was seen with factors like demographic, social capital, farm enterprise, institutional and biophysical.

4. Discussion

In this paper, we explored the geographic contexts where irrigation adoption studies were conducted and the set of causal factors that were reportedly associated with irrigation adoption decisions. Based on the results of the systematic review, our first hypothesis held true. That is, the geographic contexts in which irrigation adoption studies were often conducted were biased. Geographic regions with less than 1% area equipped for irrigation, very low (less than 0.2%) and high (above 0.8%) percent of cropland, low market accessibility index (less than 0.1), and average annual precipitation with less than 463 mm/year and greater than 1219 mm/year, were highly underrepresented in this collection of case studies. In other words, these case studies were significantly biased toward areas where at least some amount of irrigation was already being practiced. An explanation for this bias towards irrigated areas could be that the research was motivated by the need to identify challenges and/or opportunities associated with further expansion. Additionally, low cropland areas were also understudied, because research might have been focused more on areas having a moderate or higher amount of cropland cover to encourage further agricultural growth and development. Usually, farmers in areas with a high percentage of cropland cover, because of the limited scope for further (land) expansion, are more likely be using intensive agricultural practices (like irrigation) to increase their crop productivity, hence the focus was towards areas with moderate amount of cropland. Further, highly accessible regions were over-represented in this collection, because research is often conducted in locations (and with communities) that are easily accessible (or reachable) as compared to remote or hard to reach locations [105]. There is also evidence that farmers with greater market access had stronger incentives to adopt irrigation for market production [106]. Hence, regions with low market accessibility were understudied and accordingly underrepresented. Similarly, regions with low and high average annual rainfall were also underrepresented and this might be due to the overall ‘unsuitability’ of this technology within these regions. For instance, if a region receives abundant rainfall, farmers might have a natural inclination to rely on rainfall for agricultural activities rather than investing in new technology, as irrigation is generally a substitute for rainwater [107]. For regions with low average annual rainfall, although irrigation technology can be very useful nevertheless, reliable access to water might hinder its widespread diffusion and subsequent adoption [108].

The second hypothesis that we tested in this paper held partially true as only the Demographic category of factors was observed as the most common among all the case study regions. This indicated that demographic factors such as a farmer’s age, gender, household assets, income diversification options, and perceptions toward climate change (see Figure 3 for a complete list), significantly affected farmers’ decisions to adopt (or not) irrigation irrespective of the geographic context. However, some distinct regional variations were also seen. For instance, studies from North America explained irrigation adoption behavior of farmers using a combination of only demographic, biophysical, social capital, farm-enterprise, and institutional factors. Factors related to place or technology did not feature in the case studies from this region. Similarly, for cases from Near East, only categories of factors such as demographic, farm enterprise, biophysical and social capital were observed. Both institutional and technology related factors were least observed among all these case studies. Further, the highest frequency was of the cluster with Biophysical, Demographic, Farm Enterprise, and Social Capital factors (B, D, F, S), followed by the cluster with Biophysical, Demographic, Farm Enterprise, Institutional, and Social Capital factors (B, D, F, I, S), suggesting that irrigation adoption decisions around the world are best explained by the combination of multiple and coupled factors instead of a single variable.

Moreover, majority of the case studies in this collection were from geographic regions of Asia and Africa and were clustered with a greater (and often similar) number of factors as compared to the rest. This suggests that some common challenges might possibly exist with regard to irrigation technology diffusion and adoption within these regions, even though the study sites within these regions (See Appendix A for more information on study locations) were different from each other in many other aspects beyond just percentage of irrigation or average annual precipitation (national wealth, population densities, etc.). A recent study on understanding sustainability challenges in three different rural landscapes, namely, Australia, central Romania, and southwestern Ethiopia, found similarities among these three different social-ecological systems, even though the systems examined appear to be very different on the surface [109], thus, highlighting the need for a comprehensive analysis to identify and better comprehend such common challenges.

Although a nearly similar set of factors were observed from case studies of Asia and Africa, many of the study sites from Africa with little to no irrigation (less than 1%) were understudied, while all those from Asia were over-studied and hence over-represented in this collection (Figure 12). One explanation for this research bias could be that the farmers in the study sites within Africa might still be in their early adoption phase. Given the low percentage of irrigated areas, one can argue that in these sites only a few individuals are taking the risk of investing in this technology. Moreover, this technology might not have been completely diffused within these sub-regions of Africa (east, west, and south), and as a result, this topic might be highly understudied within these sites because there is first a need to properly introduce this technology to the people, make them aware of its use and benefits, and only then can the adoption process be studied. Furthermore, based on the results of the frequency analysis, institutional and social capital related factors were most commonly observed in cases from this study region compared to others. These categories include factors like access to informational services, credit facilities, extension services, skill development programs, supporting policies, incentives, and subsidies. A study by Wozniak (1987) [110] highlighted the important role played by education and information on the new technology, particularly for early adoption. Another study by Diederen et al. (2003) [111] presented empirical evidence for explaining the differences in adoption behavior of innovators, early adopters, and laggards. Their findings suggested that innovators (~first or early users of technology) made more use of external sources of information. In a more recent study on the adoption of improved seed varieties by farmers in Ethiopia, the findings suggested that farmers’ awareness about the available seed varieties is an important factor for the actual adoption to take place [112]. Teha & Jianjun (2021) [113] in their study on the adoption of small-scale irrigation found that ‘government promotion’ in the form of incentives and training positively affected a farmer’s irrigation adoption decision. Thus, some kind of external support like extension and credit services are vital for farmers for enhancing the diffusion and adaptation of successful technologies and practices [114,115]. With limited information and support, a farmer’s decision-making is primarily based on intuition and can be less efficient [116].

However, the results of this meta-study are limited in scope, since only peer-reviewed research articles that were available in the English language, in the two selected databases, and published on and after the year 2000 were considered for this analysis. Such a restriction on the publication date was imposed because the global irrigation dataset used in this analysis is based on the nationally reported statistics from around the year 2000. Further only articles that investigated the factors associated with irrigation adoption were selected for this analysis irrespective of the theoretical frameworks applied to examine a farmer’s adoption behavior. Due to this, certain factors might be emphasized more than others. For instance, a social network analysis approach was used to assess the barriers to climate change adaptations in Spain [117]. Because of the specific framework used in this study, the barriers identified were mostly categorized within social capital and institutional categories (see Appendix A for study details). Similarly, another case study from Nepal, used risk perception and motivation theory to understand farmers preparedness to cope with the impacts of climate-change hazards [118], and as a result, only the factors characterized as demographic were identified from this case study. Moreover, conference proceedings and grey literature were also excluded from the dataset due to inconsistent methodology and results reporting. Such sources may have contained useful and unique insights, but issues of comparability with information gathered from peer-reviewed would have unduly complicated the analysis.

Despite the limitations mentioned above, the global representativeness analysis highlights the multiple (geographic) biases that exist with respect to studying farmers’ irrigation adoption decision-making. More research on this topic is being conducted in regions that have little to high percentage of irrigation (>1%), are readily accessible, receive moderate amounts of average annual rainfall, and have moderate amounts of cropland cover. These results suggest the need to expand research efforts, particularly in areas with low irrigation and cropland cover to identify constraints to and help accelerate economic growth, poverty reduction, and food and livelihood security for rural communities in these regions.

5. Conclusions

Food production is still risky in many parts of the world, particularly in Sub-Saharan Africa, due to limited information about changing weather patterns, market access and demands, and unequal access to efficient technologies [116]. Additionally, this pressure on our global food systems will only intensify in the coming years with not only the changing consumption patterns but with the changing climatic conditions as well. For example, yield declines resulting from climate change (e.g., higher temperatures, increased seasonality, more frequent and severe hydroclimate events) have already occurred [119] and are expected to decrease the production of global consumable food calories by another 1% to 7% by the end of the century [120]. Irrigation currently remains one of the most critical inputs to farming today and is a key adaptation to variable precipitation and droughts resulting from changing climatic conditions [38]. New investments in irrigation infrastructure together with improved water management practices can not only minimize the impact of water scarcity but can also aid in meeting the water demands for global food production [121]. Further, managing and improving irrigation efficiency will, in turn, support global water, food, and energy goals [122]. Therefore, understanding the diverse reasons, motivations, and/or factors underlying the choices of producers regarding its adoption (or rejection), especially when climate change demands some kind of adaptation in unprecedented areas, will help better anticipate future food, energy, and water demands [123].

There is still much room left for improvements in both agricultural practices and water-use efficiency, but farmers’ reluctance to adopt new technologies needs to be better understood if such sustainability targets are to be achieved [72], and societal resilience must be built to mitigate the impacts of future climatic changes [11]. In this study, we identified multiple geographic biases that exist with respect to studying farmers’ irrigation adoption decision-making, thus, suggesting the need for extensive research even in areas with no irrigation and/or low cropland cover to identify opportunities for the implementation of other sustainable solutions to support agricultural development in these areas. Moreover, apart from these biases, some commonalities were observed in terms of constraints faced by farmers regarding irrigation technology adoption across different geographic landscapes. However, our findings also indicated that there may not be a ‘standard set’ of factors for understanding irrigation adoption, and nuances in the local context are just as important to identify as commonalities across settings. This suggests the need for more geographically comprehensive analyses that would enable comparative analysis of different landscapes, as well as studies that delve into the adoption process beyond individual technology adoption behaviors. Further, this kind of systems analysis will help unravel common challenges, drivers, and opportunities regarding agriculture development under changing climatic conditions across multiple systems, while also being attentive to local context offers the potential for co-learning [109,124].