Landscape resource management for sustainable crop intensification

Crop intensification is required to meet the food demands of an increasing population. This paper presents data from three paired scaling-up initiatives to compare the benefits of landscape-based interventions over individual plot-level interventions using evidence generated in the Indian semi-arid tropics. A range of soil and water conservation interventions were implemented in a decentralized manner following the landscape-based approach. The plot-level approach focused only on balanced fertilizer application and improved crop cultivars while the landscape-based interventions primarily addressed moisture availability, which was the key to reducing risks of crop failure besides aiding productivity gain and enhanced land and water-use efficiency. These interventions have additionally harvested 50–150 mm of surface runoff and facilitated groundwater recharge in 550–800 mm rainfall zones. Individual plot-level interventions also improved the crop yield significantly over the control plots. However, crop intensification was not achieved due to limited moisture availability. Landscape-based interventions produced 100%–300% higher crop production per year, greater income generation (>100%), and improved water productivity. Landscape-based interventions were also found to be beneficial in terms of reducing soil loss by 75%–90% and improving base flow availability additionally by 20–75 d in a year compared to untreated watersheds. With increased moisture availability, fallow lands in respective watersheds have been utilized for cultivation, thereby enhancing crop intensification. The findings of the study provide critical insights into the design of approaches suitable for scaling-up projects in order to both create impact and target the United Nations Sustainable Development Goals.


Introduction
Food, food security and the conservation of water resources are deeply embedded in the sustainable development goals that were set by the United Nations General Assembly in 2015.Global population is expected to touch 9.5 billion by 2050 and ensuring food security with a minimal water footprint is a major challenge that touches on many of these goals that are designed to achieve a sustainable future (Mekonnen and Hoekstra 2016, D'Ambrosio et al 2020, Gerten et al 2020, Hogeboom et al 2020).There is limited scope to expand agricultural land in many parts of the world as native grasslands, tropical rainforest, woodlands and wetlands have been converted to cultivation and planted pastures over past five decades (Rockstrom et al 2009, Niu et al 2019).Such changes have resulted in increased food grain production and ensured food self-sufficiency in different regions (Jägermeyr et al 2017).However, there are negative impacts of massive change in land use, including changes to hydrological cycles at global, national and regional scales, loss in biodiversity and alteration in bio-geochemical cycle of carbon nitrogen and phosphorus elements, and human induced climate change (Gleeson et al 2012, Famiglietti 2014, de Graaf et al 2019).Many of these control variables have either crossed or are fast approaching the planetary limits (Rockstrom et al 2009).The challenge, therefore, is how to bring a balance between increasing demand and dwindling resource availability without affecting ecosystem services (Ferrant et al 2014, MacDonald et al 2016).
A number of studies have indicated that there is a large yield gap in agriculture globally, especially in dryland areas.This includes India semiarid farming areas, where grain production typically below 1500 kg ha yr −1 , which is lower than the achievable potential of 3000-5000 kg ha yr −1 with available resources (Rockstrom and Falkenmark 2015, Anantha et al 2021a).It has been argued that improvements could be achieved through enhanced resource use efficiency, particularly by the introduction of improved management practices (Gerten et al 2020, Anantha et al 2021b).Such improvements in efficiency are possible as part of a second Green Revolution that incorporates adoption of an integrated genetic and natural resource management approach (Rockström 2003, Senapati andSemenov 2020).The first Green Revolution in the 1960s largely focused on irrigated ecologies, but the second Green Revolution needs to focus on both irrigated and dryland systems with enhanced resource use efficiency (Davis et al 2017, Shekhar et al 2020).Drylands in India are facing a number of challenges including water scarcity, land degradation, malnutrition, and poverty.With climate change looming large, these challenges are further exacerbated (Fritz et al 2019).Natural resource management interventions are one of the adaptation strategies for dealing with climate change challenges (Halofsky and Peterson 2010, Poff et al 2016, Strassburg et al 2020, Anantha et al 2021c).
A number of scaling-up programs designed to achieve United Nations sustainable development goals are being implemented across Asia and Africa, aiming to address issues of food security, land degradation, malnutrition and poverty through a combination of public and private investment (Jägermeyr et al 2017).Most of these programs are focusing on selected components such as soil fertility improvement, crop varietal replacement/promotion and other agronomic practices, which mostly operate on individual farmers' plots (Abera et al 2020, Kumar et al 2020).In this paper, these are referred to as plot-level interventions.In addition, several programs are being implemented at a landscape scale, which largely consider hydrological units as one entity, in which soil protections, crops, livestock and trees are integral parts of the interventions (Garg et al 2020).These are referred as landscape-based interventions.The landscapebased approach offers opportunities for resource augmentation by conserving available resources, and it also helps to enhance the productivity at the plotlevel by promoting improved management practices.
However, thus far various services generated exclusively by a landscape-based approach and a plot-level based approach have not been quantified and compared (Glendenning and Vervoort 2010, Fritz et al 2019, Mastrángelo et al 2019).The quantification and comparison presented here will help to prioritize the strategies towards achieving sustainable crop intensification.
There is increasing interest among researchers across Asia, Africa, Latin America, Europe and other parts of the world to explore opportunities to address food security, environmental challenges and social wellbeing through landscape resource management approaches (Sayer et al 2013, Estrada-Carmona et al 2014, Milder et al 2014, Freeman et al 2015, García-Martín et al 2016, Zanzanaini et al 2016, Carmenta et al 2020, Reed et al 2020, Wable et al 2021).A comprehensive review undertaken by Reed et al (2016) and Reed et al (2020) suggested that a landscape approach has considerable potential to meet social and environmental objectives at local scales while aiding national commitments to addressing ongoing global challenges.However, the evidence base within the scientific literature has remained limited despite considerable acceptance for landscape approaches (Reed et al 2020).
This paper synthesizes results obtained from three paired scaling-up initiatives carried out in semi-arid region of India that have been implemented since 1999 by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) and its consortium partners (figure 1).The plot-level approach was targeted to bridge the yield gap through the demonstration of a range of improved management practices (balanced fertilizer application and improved cultivars along with agronomic management practices) at an individual plot-level, along with capacity building.In parallel, a landscape-based approach followed another set of initiatives in which, along with productivity enhancement interventions, decentralized rainwater harvesting interventions were targeted (Garg et al 2012, 2020, 2021, Singh et al 2014, Karlberg et al 2015).For both approaches, biophysical, agronomic, hydrological, meteorological, socioeconomic parameters were intensively monitored (Garg et al 2020(Garg et al , 2021)).The overarching goal of this paper is to ascertain and compare differences between the landscape-based approach and the plotlevel approach in terms of (a) resource creation, (b) cropping intensity and (c) net income.

Description of study sites
The three paired project sites are located in the different parts of the Indian semi-arid tropics.These sites face challenges of water scarcity, land degradation, poor agricultural productivity and low resource  1. Bundi and Tonk received annual average rainfall of 600 mm (ranging 370-690 mm) and 525 mm (ranging 380-610 mm), respectively.Medek and Rangareddy received annual average rainfall of 750 mm (between 465 and 1160 mm); and Jhansi received average rainfall of 800 mm (between 400 and 1270 mm) during the study period.Out of total annual rainfall, 80%-85% is received during monsoon season (June and September).Among these sites, villages located in west and central India had sandy/loamy soils with low water retention capacity and villages in southern India are having high clay content with high water retention capacity.The land uses of these sites are also different as villages from southern and central India are agriculturally dominated (>90% agriculture land) whereas villages in Western India have high fallow/waste land and only 35% area was under agriculture (table 1).
Groundwater is the only source of freshwater for agriculture and domestic use in these villages.Shallow dug wells (5-12 m deep and a diameter of 2-5 m) fed by perched water table serve about 4-10 ha area for supplemental irrigation.As these regions are characterized by hard rock geology with poor aquifer storage capacity (1%-3%), these wells were drying soon after the monsoon season and farmers were facing water scarcity especially during dry years and also in summer season and were suffering with poor agriculture productivity and low intensification (Marechal et al 2006).

Description of landscape-based vs. plot-level initiatives
To address the above challenges, ICRISAT and partners undertook technology demonstrations in a cluster of villages between 1999 and 2016 by following two-pronged strategies viz., (a) landscapebased interventions; (b) plot-level interventions.In landscape-based initiatives, the targeted interventions were implemented for 5 years whereas for plotlevel initiatives, the interventions were undertaken between 3 and 5 years (table 1).

Landscape-based approach 2.2.1.1. Interventions implemented
Decentralized rainwater harvesting interventions were implemented by treating the land with appropriate landform treatments from ridge to valley.The landscape which has more than 2% slope was divided into smaller parts with earthen field bunds.About 1 ha area was divided into 3-4 parcels by forming earthen bunds to control runoff velocity and arrest soil loss.In addition, a range of gully control structures, farm ponds on the upstream fields were also Rainfall is the only source of water in the study watersheds, which is partitioned into various water balance components based on different biophysical factors (soil type, land use and landscape topography).A portion of the rainfall, which enters into the vadose zone through infiltration process, enhances soil moisture availability; out of this, the surplus amount contributes to the groundwater recharge.Available soil moisture of vadose zone is utilized by vegetation (trees/crops) through transpiration and some of it is evaporated, together this is termed as evapotranspiration.After saturating surface soil, an overland flow is generated (also called as surface runoff), which moves along the slope.Rainwater harvesting structures (e.g.check dams and storage structures) retain a fraction of surface runoff and the rest spilled over to downstream locations.The monsoonal water balance components of the study watershed is defined in equation (1): Rainfall (mm) = Surface runoff (mm) + Groundwater recharge (mm) + Actual Evapotranspiration (mm) .
(1) Rainfall was monitored by establishing meteorological stations and automatic runoff gauges were established at different micro watersheds of treated landscape ranging from 50 to 500 ha.For runoff gauging, a stilling well was constructed at the outlet of the watershed and a mechanical type stage recorder or an automatic pressure transducer, i.e.DIVER (pressure transducers for stage recording, Model DI801 TD) was placed at the bottom of the stilling well (Garg et al 2020, Singh et al 2021).The DIVER was programed to record pressure at the head at 15 min intervals.The measured pressure head was used to estimate spillway discharge (i.e.outflow from the watershed) by using equations ( 2) and ( 3) (2) where L is length of the rectangular weir and h t is depth of runoff layer passing from gauging station at a given time; (3) Further, the volume of water harvested in different structures was estimated from pressure head data collected through DIVER.The relationship between depth vs. storage capacity and depth vs surface area which was established by undertaking a topographic survey used to convert measured depth into storage volume.
The water table fluctuation (WTF) method is a well-accepted technique for estimating groundwater recharge in hard rock regions (Pavelic et al 2012, Garg and Wani 2013, Tilahun et al 2020).The water table of all the dug wells located in project villages (i.e.20-388 wells/site)were measured using water level indicators on a monthly time scale and groundwater recharge was estimated using equation (4) where R is the net groundwater recharge (mm), ∆h is the change in hydraulic head before and after the monsoon period (m), S is the aquifer storage (%), W is the water withdrawal during the monsoon period (mm).
The value of aquifer storage (S) was taken from earlier studies undertaken by National Geophysical Research Institute and other researchers (Marechal et al 2004, 2006, EPTRI and NGRI 2005).Whereas DIVERs were placed in selected wells to measure pumping hours in different cropping system for quantifying water withdrawal (refer., Singh et al 2021) in monsoon and post monsoon seasons.
Hydrological data were analyzed to understand the rainfall-runoff relationship under dry, normal and wet years.These categories of rainfall were differentiated as follows: (a) less than 20% of long term average = dry or deficit years; (b) greater than 20% of long term average = wet or surplus years; and (c) in between ±20% of long term average = normal year in respective ecologies (IMD 2010).The impact of surface runoff due to upstream rainwater harvesting interventions were estimated by comparing outflow in treated (with interventions) and untreated (nointerventions) landscapes in respective watersheds.Similarly, change in hydraulic head, which reflects the groundwater availability in shallow dug wells was compared in both treated and untreated watersheds.Surface runoff and groundwater recharge were estimated for respective pilot sites in different years.In addition, efforts were also made to monitor soil loss by integrating sediment samplers with runoff monitoring.During the surplus rainfall event, sediment samples at the gauging stations were automatically collected at hourly intervals and stored in separate containers.These samples were analyzed in a laboratory for sediment concentration and estimated soil loss (Pathak et al 2016).
The change in land use and cropping intensity due to landscape-based interventions was captured through ground survey before and after the project implementation.A calibrated Soil and Water Assessment Tool (SWAT, a semi-process based hydrological model, Arnold et al 2012) was used for estimating actual evapotranspiration (ET) from respective project sites (Garg et al 2012, 2021, Garg and Wani 2013).To understand the irrigation scheduling for major cropping system, the of pumping hours were measured in selected fields using DIVER.This data was provided as input into a model to simulate the ET for different years.
Data were analysed to estimate water productivity in different years (dry, normal and wet years) for both landscape-based and plot-level approach in respective study sites by using equation ( 5) where grain yield (kg ha −1 ) is measured for the field of 20 selected farmers during the kharif and rabi seasons in each of the watersheds.The detailed method of measuring crop yield using crop cutting is described in the next section.

Plot-level initiatives 2.2.2.1. Demonstrations details
In the plot-level initiatives, the focus was on optimizing available resources at individual plots for increased crop production through improved management practices.Soil test-based fertilizer application and use of improved crop cultivars were promoted.Soil samples (15-20 samples from 500 ha area) were collected by following stratified random sampling that considered the topography, landholding and cropping system from all the project sites and was analyzed for important plant available soil nutrients (available P and S, exchangeable K, B and Zn) and soil organic carbon (Walkley andBlack 1934, Hanway andHeidal 1952).Based on the soil analysis results, farmer participatory technology demonstrations on balanced fertilizer application were undertaken.Additionally, improved crop cultivars were introduced for major crops (cereals, oilseeds, pulses).
In this approach, a cluster of 3-5 villages were selected in respective districts and 200-250 farmer participatory technology demonstrations (>10 000 technology demonstrations in total) were laid out in each season to facilitate experiential learning for farmers to make them realize the productive potential of the cultivars and crop management practices.Technology demonstrations were undertaken over a minimum 3 year period in respective project sites in which about 50% of farmers were chosen as new participants every year to reach a maximum number of farmers, and the rest of the farmers were repeated at a minimum for two years.About 60% of the households in these villages are under small and marginal category with less than 2 ha of landholdings (table 1).Most of the demonstrations were associated with small and marginal farmers to enhance their capacity to adopt improved management practices.

Data monitoring
Under the plot-level approach, the farmers' fields were divided into two parts-a treated plot and a control plot to compare the impact of technology demonstrations.In the treated plot, balanced fertilizer application, improved crop cultivars or combination of both were demonstrated as per farmers' willingness and acceptability.Whereas in control plots, farmers followed their own practices.The technology demonstrations were undertaken under the close supervision of trained extension works and field scientists.About 1000 crop cutting studies (50-80 per site per season) were undertaken to assess the impact of improved management practices on crop yield (Tek et al 2016).In this method, a 3 × 3 m area was demarcated along with three replications.Crop was harvested during the maturity to measure the grain and biomass yield.In addition, cost of cultivation (farm inputs, irrigation, labour, energy cost) data was collected through household survey of selected farmers and net income was calculated using equation ( 6) where, NI a = net income (US$/HH/year); Y i is crop yield (kg ha −1 ) for plot i; A i is area of the plot i (ha); M i is market price (US$/kg); C i is cost of cultivation of plot i (US$/ha); n = number of plots farmers owning.

Statistical analysis 2.3.1. Landscape-based initiatives
Data of outflow, groundwater availability, soil loss and base flow measured on yearly time scale is compared among treated and untreated watershed using ANOVA.Further, a Chi-square test was performed to compare well-functioning status of dug wells among treated and untreated watersheds in respective project locations on monthly time scale.We divided all the dug wells into five categories based on water availability (i.e.hydraulic head, h): (a) dry; and (e) h > 5 m.The categorized data were used to check the level of significance between treated and untreated watersheds in respective sites.

Plot-level initiatives
Crop yield data of different crops obtained from treated and control plots were analyzed using ANOVA (Analysis of Variance) to understand the significance of crop yield due to the application of balanced fertilizer and crop cultivars over the control plots in respective project locations.Further, post-hoc analysis also performed to understand the impact among different treatment plots.

Comparing landscape-based vs plot-level impact
The total production and net income obtained from landscape-based and plot-level initiatives were compared over the untreated/control fields using ANOVA and post-hoc analysis.

Uncertainties of the results
Landscape hydrology is highly complex as is driven by various biophysical (topography and soil types) and land management factors.In the current study, emphasis was given on intensive hydrological monitoring; hence, surface runoff and groundwater recharge were measured using state-of-the art instrumentation.ET at landscape level was estimated in respective years as the balance closure of the water balance equation (refer equation ( 1)).However, ET at plot scale for different crops (e.g.maize, pearl millet, groundnut, etc) was estimated using simulation modeling, which may spatially vary due to inherent heterogeneity of the landscape and management practices.This may develop about 10%-15% uncertainty in ET estimation (i.e.∼30-50 mm) as there could be variations in soil properties, management factors and supplemental irrigation from plot to plot in respective watersheds.

Resource provision through landscape-based approach
The impact of interventions In Rangareddy (Telangana) and Jhansi (Uttar Pradesh) districts, which are 600-800 mm rainfall regions, the generated outflow largely decline during normal years.Runoff generated at 600 mm rainfall was negligible (<5% of total rainfall) as these villages are dominated by agricultural land use.Various landscape-based interventions in Rangareddy and Jhansi districts could harvest only marginal surface runoff in dry years.During normal years, the available runoff was 5%-15% of total rainfall in untreated watersheds, landscape-based interventions harvested about 40%-90% of generated runoff.Whereas the outflow generated with >1000 mm rainfall condition in wet years was in the range of 30% of total rainfall.Under this condition, the landscape-based interventions could partially harvest (10%-50%) generated runoff and rest was available for downstream uses.
Table 2 indicates that the landscape-based interventions have altered water balance components (outflow and groundwater availability) significantly (p < 0.05) between treated and untreated watersheds in the respective project sites.The harvested runoff due to landscape-based interventions primarily impacted on groundwater resource availability, both in terms of its amount and longevity.Figure 3 shows resource availability status in paired watersheds (with and without interventions) based on data collected from three study locations for dry, normal and wet years.Groundwater levels increased from 2.0 to 5.0 m and a minimum of 30% of the defunct wells were rejuvenated.In the watersheds without landscape-based interventions, the functioning percentage of shallow dug wells was less than 55% in most of the cases including in wet years.The runoff generated from the landscape quickly dissipates and therefore, in the untreated watershed, the groundwater recharge opportunities were limited Further, to describe the fluctuation in groundwater table on a spatial scale, we have presented data for one of the paired watersheds from Uttar Pradesh. Figure 4 summarized the water availability (hydraulic head) status of functioning wells over the year during 2014, 2015 and 2016 in treated and untreated watersheds along with the rainfall received.We categorized functioning wells into five groups based on available hydraulic head (water column in respective dug wells): (a) dry (b) poor (0-1 m); (c) moderate (1-3 m); (d) good (3-5 m) and (e) excellent (>5 m).In 2013 there was a total rainfall of 1276 mm, so it was classified as a wet year, whereas 2014, 2015 and 2016 received 520 mm, 404 mm and 768 mm respectively.As 2013 was a wet year (not shown in figure) both treated and untreated watersheds responded with 100% functioning well status.However, treated watersheds showed more than 85% of wells with hydraulic head status greater than 5 m compared to 32%-60% wells in untreated watershed under this category (up to October 2014).
In addition, about 10%-20% of dug wells were yielding excellent (>5 m head), 35%-40% good (3-5 m head); 20%-25% moderate (1-3 m); and less than 20% wells with less than 1 m hydraulic head in treated watersheds during June to September 2015.On the other hand, only 2%-5% of the total wells were in the excellent category (>5 m), less than 20% good (3-5 m head); 40% moderate; and 30%-50% wells were in poor yielding/nonfunctional category in untreated watershed during June to September 2015.The recharged water in shallow aquifers remained available for longer periods in treated watersheds compared to untreated watersheds.For example, despite receiving 520 mm and 404 mm rainfall in 2015 and 2016, a large percentage of shallow dug wells in treated watersheds were functioning.About 5%-18% wells were only dried in treated watersheds during October-December 2015 compared to 28%-38% in untreated watershed during the same period.
By February 2016, more than 40% of the wells were yielding, with moderate (1-3 m) to good (3-5 m) hydraulic head in treated watershed compared to only 17% in untreated watersheds.At any point of time, 20% of wells in treated watersheds were yielding a minimum amount with moderate status compared to 10% in untreated watersheds.During the beginning of the monsoon in July 2016, 35% of wells rejuvenated back to excellent, 40% to good and 33% to moderate in treated watersheds.Whereas, about 40% and 60% of wells were showing good and moderate yield status in untreated watersheds in July 2016.
The Chi-square statistics revealed that the functioning status of wells in treated watersheds of Jhansi (Uttar Pradesh) was significantly different from untreated watersheds (table 3).The difference in functioning wells was found highly significant among treated and untreated watersheds in all months except May 2016.As May 2016 was one of the driest months after two consecutive dry years i.e. 2014 and 2015.
Similar observations were also made in the other two project sites (Rajasthan and Telangana), with landscape-based interventions improving the functioning status of the dug wells with higher water availability (i.e.hydraulic head) (Garg et al 2012, 2021).
Figure 5 compares the area cultivated under different crops during dry, normal and wet years in kharif (monsoon) and rabi (post-monsoon) seasons in treated and untreated watersheds of Bundi (Rajasthan), Jhansi (Uttar Pradesh) and Rangareddy (Telangana) districts.With improved groundwater availability, the cropped area which was either single crop or left fallow has been brought into cultivation in both kharif and rabi season.Farmers in Bundi and Jhansi those largely prefer to cultivate rabi crops (mustard, chickpea, wheat, vegetables) were benefited to its full potential.The area under wheat (staple cereal crop in the region) doubled with assured water availability compared to nearby untreated watersheds.Whereas the area under mustard and   chickpea declined in treated watersheds.A minimum of 25% cultivable land, especially in uplands, that was always under fallow condition due to non-availability of water had been brought into productive use and only 5% land was remained fallow during normal and wet years.On the other hand, cropping system in the Rangareddy (Telangana) project site was kharif dominated and has also benefited with the improved availability of groundwater.Farmers started cultivating high-value vegetable crops during the monsoon and post-monsoon season along with cotton and pigeonpea.Similarly, cultivable fallow land has declined significantly after implementing landscape-based interventions.

Bridging yield gaps through improved management practices-plot-level approach
Figure 6 shows the soil fertility status of agricultural fields in selected villages of study districts.The soil sample analysis showed that these soils are deficient in micro and secondary nutrients along with poor organic carbon status (Chander et al 2013, Wani et al 2016Wani et al , 2017)).For example, more than 60% of farmers' fields were found deficient in available sulphur, boron and zinc in Jhansi (Uttar Pradesh) and Tonk (Rajasthan); and 30%-60% of fields at Telangana project sites.Whereas potassium (K) deficiency was found in less than 30% of fields across study villages.Organic carbon content was found below 0.5% in 70%, 30% and 60% of fields in the Rajasthan, Uttar Pradesh and Telangana pilot sites.Poor soil organic carbon also indicates likely mineral nitrogen deficiency in these soils and therefore, requires minimum doses of nitrogen supplement in the form of organic or inorganic fertilizers.
The crop cutting studies revealed that the crop yield was sensitive to the application of micronutrients.Application of micronutrients alone helped to increase crop grain yield by 10%-30% (figure 7).However, the maximum gain yield recorded was more than 100% in some cases when the initial productivity level was very low (<300-500 kg ha −1 ), indicating the high level of soil degradation.These initiatives helped to realize the potential of improved crop cultivars as the yield obtained in different crops was higher by 15%-30%.Some of these micronutrients activate several important metabolic reactions and play a direct role in photosynthesis and are helpful for better extraction of other nutrients from soils (Hussain et al 2012, Nadeema and Farooq 2019).
Farmers tend to use low yielding traditional crop cultivars due to poor awareness, non-availability of seeds of improved cultivars, and poor understanding about their suitability to their regions (Atlin et al 2017, Borg et al 2018, Singh et al 2020).Improved crop cultivars (including drought tolerance) along with the application of micronutrients have a compounding effect on crop yields (figure 7; table 4).With these integrated crop management practices, crop yield was significantly increased over control plots (p < 0.05).As prevailing crop yield levels were very low, these interventions showed higher incremental advantage and helped to bridge the yield gap.For example, the average maize yield in Tonk (Rajasthan) was 1800 kg ha −1 , which increased to as high as 2685 kg ha −1 with the application of micronutrients and 4640 kg ha −1 with a combination of micronutrients and improved crop cultivars.The average yield gain (difference between treated and control plots) in millets was found to be 320-420 kg ha −1 in response to the application of micronutrients and improved crop cultivars in Rajasthan.In the Jhansi project site, the average yield gain in groundnut, chickpea and wheat was 460 kg ha −1 , 700 kg ha −1 and 300 kg ha −1 respectively, with the application of micronutrients and improved crop cultivars.Similarly, the average chickpea yield in the Medak (Telangana) project site was increased from 1250 kg ha −1 to 1460 kg ha −1 with the introduction of improved cultivars.

Comparison of landscape-based vs. plot-level interventions
Figure 8 and table 5 compares the impact of plotlevel and landscape-based interventions on crop intensification and net income.The implementation of landscape-based interventions helped to enhance groundwater availability additionally by 50%-100% compared to untreated condition, which made it possible for farmers to convert their cultivable fallow lands into productive agricultural land.Whereas, no such impact was observed exclusively through the plot-level approach.A minimum of 15% of cultivable fallow land was converted into productive agriculture land with landscape-based interventions in different project sites.Agricultural production ranged from 800 to 1200 kg ha −1 before interventions, which increased to 1500-2000 kg ha −1 with the introduction of plot-level interventions i.e. improved crop cultivars and micronutrients.However, in landscapebased interventions, the increase was substantial (i.e.3500-5000 kg ha −1 ), largely due to reduced risks of crop failure with availability of supplemental irrigation and also due to opportunity for taking a second crop within a year.As a result, household incomes from agriculture increased from US$ 150 to US$ 450/household/year with plot-level intervention and increased to more than US$ 700/household/year with landscape-based interventions.Similarly, water productivity in agriculture improved on an average from 0.15 to 0.30 kg m −3 with plot-level interventions and 0.60 kg m −3 with landscape-based interventions.As indicated by F statistics and post-hoc analysis, the difference in agriculture production and net income was significant between control plots vs treated plots and

Discussion
In developing countries like India, sustainable crop intensification is required to feed the growing population, address malnutrition and fuel overall economic development (McLaughlin and Kinzelbach 2015).However, poor management of resources is one of the challenges which limits the opportunity to enhance crop intensification.Plot-level approaches to enhance crop productivity comprise a combination of improved management practices that helps to bridge the yield gap. Typically, with limited availability of moisture, their potential is not harnessed to its full capacity (Davis et al 2017).Landscapebased approaches first ensure moisture availability by harvesting additional runoff within the landscape in a decentralized manner (Abbasi et al 2019).These interventions are immensely important in the semiarid tropical climatic conditions in which rainfall is received only for few days (30-50 rainy days) in a year.In the current climate change scenario,   in summer, storing the equivalent amount of water in shallow aquifers preserves it for a longer period.Moreover, the surface water at downstream locations provides services only to a limited group of farmers, which often leads to conflict among stakeholders.In surface irrigation system, small holders and upland farmers were benefited least and found largely dependent on resource rich farmers for buying the water (Ajaz et al, 2019, Bajaj et al 2022).In contrast, the decentralized approach of groundwater recharge addresses larger groups of farmers including small and marginal farmer in an equitable manner.In these regions, groundwater is one of the reliable sources of freshwater as it is available on demand unlike surface water, which depends on predefined schedules and requires heavy investment in infrastructure (large dams and canal networks).In these areas, with innovation in pumping technologies, even marginal and small farmers can also access groundwater.
It is evident that when the water is made available, farmers were motivated to invest in new agricultural practices and explore new opportunities for crop intensification by investing in improved varieties, improved irrigation methods, farm machinery, and other farm inputs which enhanced their income and wellbeing (Ruzzante et al 2021).In contrast, under plot-level approach this self-motivation was relatively low as there was always uncertainty towards freshwater availability, which is the pre-requisite for crop intensification and reducing risks of crop failure.This is also one of the reasons for low adoption of improved management practices including crop cultivars especially in drylands (Cavatassi et al 2011, Di Falco and Bulte 2013, Ruzzante et al 2021); hence, agriculture growth is stagnant.Adoption of sustainable agriculture practices and outcomes are largely dependent on short term economic benefits (Piñeiro et al 2020).This is evident that agriculture production was about 1000 kg ha yr −1 which increased to 1500 kg ha yr −1 by introduction of plotlevel initiatives, but under landscape-based interventions it increased to 3500 kg ha yr −1 .The net water consumption under the non-intervention stage was about 60% of the rainfall received in the respective ecologies.Whereas, with landscape-based interventions, an additional 10%-15% of rainfall was harvested in terms of soil moisture and groundwater recharge, resulting in increased production at least by three times.This phenomenon was defined by the vapour shift concept outlined by Rockström (2003).A significant amount of water from fallow lands (either seasonal or permanent fallow), which is lost as non-productive evaporation under preor non-intervention stage, has shifted to productive transpiration.With this approach, more than 80% of farmers benefited in the respective project locations.However, we realized that there is a tradeoff between upstream development and downstream water availability, but it is always not negative (Garg et al 2020).Upstream development on the one hand reduces 40%-50% of freshwater availability in normal years, but at the same time it also controls floods and land degradation downstream during wet years (Garg et al 2021).During dry years, the generated runoff, however, is not sufficient to meet the water demand upstream and the downstream ecosystem as well, which should not be a concern.The base flow availability which has reduced significantly in last 3-4 decades in drylands and most of the perennial rivers of semi-arid tropics became seasonal (Sabzi et al 2019, Gerten et al 2020, Jägermeyr 2020), and the landscape-based approach provides an opportunity to rejuvenate the base flow with increased green cover and infiltrating more water into the soil.
The results from the plot-level approach showed positive impact in terms of enhanced productivity and knowledge of improved management practices among the community.However, these impacts are still limited to surplus production and achieving income security especially for small and marginal households that have less than 2 ha land area.Particularly during dry years, the net benefits accrued from plot-level interventions were so minimal and the impact of extreme events largely forced farmers to invest all their earnings to secure their livelihoods and they were found to be under the poverty trap (Hallegatte and Rozenberg 2017, Barbier and Hochard 2018, Marotzke et al 2020).In contrast, landscape-based interventions build system-level resilience against such shocks by resource creation, crop intensification and productivity enhancement (Singh et al 2014, Garg et al 2020).Even during dry years, farmers were able to harvest higher yields compared to their counterparts in untreated watersheds.Other than the direct benefits, landscape-based interventions also supported allied sectors such as livestock and water security for domestic uses (not discussed in this paper), which is one of the important challenges faced by the communities in rural areas of developing countries.With technological advancements in the areas of monitoring and evaluation, it has become possible to capture impact more accurately.It is important to put similar efforts for different agro-ecological regions to bridge the knowledge gap and to facilitate informed decisions.
The lessons of this study will be helpful to multiple stakeholders, particularly those seeking to achieve United Nations Sustainable Development Goals, as they inform the design of appropriate guidelines to promote integrated landscape approach towards addressing challenges of social wellbeing together with environmental sustainability.Policy makers should combine the results presented in this study with site-specific information to promote integration of landscape-based resource management and plot level technologies.Implementation will require a multi-disciplinary approach and extensive capacity and awareness building to address socio-economic and environmental tradeoffs facing people and nature in complex dryland systems.

Conclusions
This paper has compared the benefits of landscapebased approach over those accrued from a plot-level approach by examining examples from three paired initiatives undertaken since 1999.The landscapebased approach was found superior in terms of water resource creation, which translated into sustainable crop intensification, increased production, and household income over the plot-level interventions in semi-arid tropics.The total production and income accrued with landscape-based management approaches, even during extreme years, were much higher than those achieved during favorable years in plot-level approaches.This study also recommends that the plot-level approach should be converged with landscape-based initiatives for maximizing the benefits and achieving sustainability.These findings have important implications for the type and scale of measures that should be implemented at various scales by stakeholders, and which should be supported by government policy interventions.They provide important examples of practice that are suitable for deployment in the arid and semi-arid regions of India, elsewhere in South and greater Asia, and also Africa and Australia.Improvements in water resource creation, crop intensification, production, and household incomes have the potential to target a range of different sustainable development goals, including those related to poverty, food security, health and well-being, inequality, and responsible production.Similar efforts are required to intensify monitoring in different agro-ecological regions and create sciencebased evidence for scaling up climatic resilient technologies.

Figure 1 .
Figure 1.Location of study sites in the semi-arid tropical region of India following plot-level vs landscape-based approach; close-up maps shows land slope, stream network, location of rainwater harvesting structures and distribution of monitoring wells in treated watersheds under the landscape-based initiatives of the respective study districts; (Data source: climate classification in India by Raju et al 2013).

Figure 3 .
Figure 3. Resource availability status before and after landscape-based interventions-(a) generated outflow; (b) groundwater availability; (c) per cent functioning dug wells; and (d) cropping intensity.

Figure 4 .
Figure 4. Percentage of total wells with functioning status in (a) treated and (b) untreated watersheds between 2014 and 2016 along with (c) monthly rainfall.Number of wells monitored in treated and untreated watersheds were 388 and 150 at Jhansi, Uttar Pradesh, respectively.

Figure 5 .
Figure 5. Change in cropping pattern due to landscape-based interventions in pilot villages of Western, Central and southern India.

Figure 7 .
Figure 7. Impact of improved management practices on cereals, pulses and oilseeds in different environments, India.MN indicates micronutrients.

Figure 8 .
Figure 8. Comparing the impact of plot level based and landscape-based interventions on crop intensification (a) and (b), income (c) and water productivity (d).

Table 1 .
Location, rainfall, soil types and demographic details in landscape-based and plot-level initiatives in different states, India.

Table 2 .
ANOVA (F value)showing effects of landscape-based interventions on outflow, groundwater availability (hydraulic head in dug wells), soil loss and base flow (significant at p < 0.05).

Table 3 .
Measure of the functioning status between the treated and untreated watershed using Chi square statistics (significance level 5%).
a Not significant; the Chi-squared test was performed for 33 months between April 2014 and December 2016 (Data are based on 388 and 150 wells monitored in treated and control watersheds, respectively).

Table 4 .
ANOVA (F value)showing effects of micronutrients and improved crop cultivars on crop yield (significant at p < 0.05).
MN: Application of micro nutrients; Cultivars: Improved crop cultivars.a Post-hoc analysis indicate significant difference in crop yield among balanced fertilizer application and improved crop cultivars at all the project sites in different crops.

Table 5 .
ANOVA (F value)showing effects of different technologies on crop production, net income and water availability (significant at p < 0.05).
a Post-hoc analysis indicate significant difference in landscape-based and plot-level interventions in terms of crop production and net income in all the project sites.
Characterizing and evaluating the impacts of national land restoration initiatives on ecosystem services in Ethiopia Land Degrad.Dev.31 37-52 Ajaz A, Karimi P, Cai X, De Fraiture C and Akhter M S 2019Statistical data collection methodologies of irrigated areas and their limitations: a review Irrigation Drainage 68 702-13 Anantha K H, Garg K K, Barron J, Dixit S, Venkataradha A, Basu N B, Tate E and Wyckoff J 2014 Monsoon harvests: the living legacies of rainwater harvesting systems in South India Environ.Sci.Technol.48 4217-25 Meter K J V, Steiff M, McLaughlin D L and Basu N B 2016 The socioecohydrology of rainwater harvesting in India: understanding water storage and release dynamics across spatial scales Hydrol.Earth Syst.Sci.20 2629-47 Milder J C, Hart A K, Dobie P, Minai J and Zaleski C 2014 Integrated landscape initiatives for African agriculture, development, and conservation: a region-wide assessment World Dev.54 68-80 Nadeema F and Farooq M 2019 Application of micronutrients in rice-wheat cropping system of South Asia Rice Sci. 26 356-71 Niu G, Zheng Y, Han F and Qin H 2019 The nexus of water, ecosystems and agriculture in arid areas: a multi objective optimization study on system efficiencies Agric.Water Manage.223 105697 Pathak P, Chandrasekhar K, Wani S P, Sud R R and Budama N 2016 Integrated runoff and soil loss monitoring unit for small agricultural watersheds Comput.Electron.Agric.128 50-57 Pavelic P, Patankar U, Acharya S, Jella K and Gumma M K 2012 Role of groundwater in buffering irrigation production against climate variability at the basin scale in South-West India Agric.Water Manage.103 78-87 Piñeiro V et al 2020 A scoping review on incentives for adoption of sustainable agricultural practices and their outcomes Nat.Sustain. 3 809-20 Poff N et al 2016 Sustainable water management under future uncertainty with eco-engineering decision scaling Nat.Clim.Change 6 25-34 Raju B M K et al 2013 Revisiting climatic classification in India: a district-level analysis Curr.Sci.105 492-5 Reed J, Ickowitz A, Chervier C, Djoudi H, Moombe K, Ros-Tonen M, Yanou M, Yuliani L and Sunderland T 2020 Integrated landscape approaches in the tropics: a brief stock-take Land Use Policy 99 104822 Reed J, Vianen J V, Deakin E L, Barlow J and Sunderland T 2016 Integrated landscape approaches to managing social and environmental issues in the tropics: learning from the past to guide the future Glob.Change Biol.22 2540-54 Rego T J, Sahrawat K L, Wani S P and Pardhasaradhi G 2007 Widespread deficiencies of sulfur, boron, and zinc in Indian semi-arid tropical soils: on-farm crop responses J. Plant Nutr.30 1569-83 Rockström J 2003 Water for food and nature in drought-prone tropics: vapour shift in rain-fed agriculture Phil.Trans.R. Soc.B 358 1997-2009 Rockstrom J et al 2009 A safe operating space for humanity Nature 461 472-5 Rockstrom J and Falkenmark M 2015 Agriculture: increase water harvesting in Africa Nature 519 283-5 Ruzzante S, Labarta R and Bilton A 2021 Adoption of agricultural technology in the developing world: a meta-analysis of the empirical literature World Dev.146 105599 Sabzi H Z, Rezapour S, Fovargue R, Moreno H and Neeson T M 2019 Strategic allocation of water conservation incentives to balance environmental flows and societal outcomes Ecol.Eng.127 160-9 Sahrawat K L 2006 Plant nutrients: sufficiency and requirements Encyclopedia of Soil Science 2nd edn, ed R Lai (London: Taylor and Francis) pp 1306-10 Sahrawat K L, Wani S P, Rego T J, Pardhasaradhi G and Murthy K V S 2007 Widespread deficiencies of sulphur, boron and zinc in dryland soils of the Indian semi-arid tropics Curr.Sci.93 1428-32 Sayer J et al 2013 Ten principles for a landscape approach to reconciling agriculture, conservation, and other competing land uses Proc.Natl Acad.Sci.110 8349-56 Senapati N and Semenov M A 2020 Large genetic yield potential and genetic yield gap estimated for wheat in Europe Glob.Food Secur.24 100340 Shekhar S, Kumar S, Densmore A L, van Dijk W M, Sinha R, Kumar M, Joshi S K, Rai S P and Kumar D 2020 Modelling water levels of northwestern India in response to improved irrigation use efficiency Sci.Rep. 10 13452 Singh R P, Chintagunta A D, Agarwal D K, Kumar S P J and Kureel R S 2020 Varietal replacement rate: prospects and challenges for global food security Glob.Food Secur.25 100324 Singh R, Garg K K, Anantha K H, Venkataradha A, Dev I, Dixit S and Dhyani S K 2021 Building resilient agricultural system through groundwater management interventions in degraded landscapes of Bundelkhand region, Central India J. Hydrol.: Region.Stud.37 100929 Singh R, Garg K K, Wani S P, Tewari R K and Dhyani S K 2014 Impact of water management interventions on hydrology and ecosystem services in Garhkundar-Dabar watershed of Bundelkhand region, Central India J. Hydrol.509 132-49 Sishodia R P, Shukla S, Graham W D, Wani S P, Jones J M and Heaney J 2017 Current and future groundwater withdrawals: effects, management and energy policy options for a semi-arid Indian watershed Adv.Water Resour.110 459-75 Strassburg B B N et al 2020 Global priority areas for ecosystem restoration Nature 586 724-9 Tabari H 2020 Climate change impact on flood and extreme precipitation increases with water availability Sci.Rep. 10 13768 Tek B S, Jat M L, Jat R K, Kapoor P and Clare S 2016 Yield estimation of food and non-food crops in smallholder production systems Methods for Measuring Greenhouse Gas Balances and Evaluating Mitigation Options in Smallholder Agriculture ed T S Rosenstock, M C Rufino, K ButterbachBahl, E Wollenberg and M Richards (Berlin: Springer) pp 163-74 Tilahun S A, Yilak D L, Schmitter P, Langan S, Barron J, Parlange J Y and Steenhuis T S 2020 Establishing irrigation potential of a hillside aquifer in the African highlands Hydrol.Process.34 1741-53 Wable P S et al 2021 Modeling impact of agricultural water management interventions on streamflow in upper cauvery sub-basin of cauvery basin using SWAT Irrigation Drainage 1-21