Introduction

In 2015, the United Nations (UN) together with the heads of the state of 193 countries agreed to adopt the 2030 Agenda as a response to global challenges such as population growth, changing lifestyles, and climate change which deplete resources and stress the earth system (Bhaduri et al. 2016). They were built on the achievements of the Millennium Development Goals (MDGs) and are far more ambitious in scope of global development (Purvis et al. 2019). It consists of 17 Sustainable Development Goals (SDGs), 169 targets and 231 indicators (UN 2017; WWAP and UN-Water 2018; Eisenmenger et al. 2020). The 2030 Agenda now also incorporates a specific target focussing on achieving sustainable management of clean water and sanitation (SDG 6) (UN 2015). This is in recognition of the fact that water availability supports economic growth and the wellbeing of people and has been identified as one of the main challenges for human survival (Steffen et al. 2015).

SDG 6 consists of two indicators, one focusses on overall water scarcity (SDG indicator 6.4.2) and the other one focusses on efficient use of water from an economic perspective (SDG indicator 6.4.1). These indicators take into consideration water withdrawals of three main economic sectors – agriculture, industries and services and compare them to their economic contribution to the economy (6.4.1) and the total water available (6.4.2) (UN-Water 2021). SDG indicator 6.4.1 is a composite of the water use efficiency (WUE) of the three sectors, expressed into economic value added per unit of water withdrawn (in USD/m3). As a composite, this indicator is dominated by the high value, low water use sectors such as industry and therefore to be able to provide sector specific recommendations, literature suggests to evaluate each sector’s water use efficiency separately (Hellegers and van Halsema 2021).

Monitoring WUE for the agricultural sector, and thus Agricultural Water Use Efficiency (Awe), is also important for SDG indicator 6.4.2 as it is responsible for over 70% of the globally withdrawn water (FAO 2015). Even though the methodology for monitoring SDG indicator 6.4.1 has been clearly defined (FAO 2019), there is an ongoing debate on the rationale for using water withdrawn as it doesn’t take into consideration return flows which could be used elsewhere or is returned to the environment (Hellegers and van Halsema 2021; Vanham 2018). An alternative recommended approach therefore is to evaluate the agricultural productivity against the water consumed in the form of evapotranspiration (ET) (Hellegers and van Halsema 2021), which is also known as agricultural water productivity. One of the discrepancies between considering water withdrawals compared to water consumed is that the latter includes water stemming from rainfall. It is therefore important to consider the water consumed from irrigation only (also called blue ET (ETb) (Chukalla et al. 2015).

The progress of the SDGs is monitored by the countries themselves through comprehensive data collection (Connor 2015; Reyers et al. 2017). However, this is a costly effort and most countries do not collect this information on a regular basis (Roopnarine et al. 2019). In 2019, OECD (2019) reported that on average countries reported on just 55% of the SDG indicators since 2015, with only few countries providing data for all years. The low level of progress compared to the MDG monitoring, which was 68% in 2015, can be explained by the fact that the SDGs has more indicators (231 vs. 48) and more complex computing workflows (MacFeely 2018).

Remote Sensing (RS) is considered as a reliable source of data which can support the monitoring and reporting of the SDG indicators on a regular and near-real time basis (Calera et al. 2017; Hakimdavar et al. 2020; Kavvada et al. 2020). FAO, one of the custodian agencies for reporting several SDGs is therefore actively promoting the uptake of amongst other data, the use of Earth Observation (Gennari et al. 2019). Open data platforms such as Global surface water explorer (Pekel et al. 2016) and the Freshwater Ecosystems Explorer (Lu et al. 2021) contribute significantly to the periodical estimation of indicators related to SDG 6.Recent publications demonstrate the use of satellite data in mapping of wetlands contributing to SDG 6.6.1 (Weise et al. 2020), mapping land degradation to support SDG 15 (Giuliani et al. 2020), and collect evidence on slavery from space supporting SDG 8 (Foody et al. 2019; Boyd et al. 2018). RS based ET has been used to monitor various aspects of water management such as estimation of water allocation to various sectors (Droogers et al. 2010) and irrigation performance assessment (Karimi et al. 2019), however, the use of RS based ET for monitoring the SDGs is still relatively new (Biancalani and Marinelli 2021). There are various RS based ET products available which can be utilised to support reporting of SDG 6 indicators (e.g. SSEBop: Senay et al. 2013; CMRSET: Guerschman et al. 2009; SEBS: Chen et al. 2013; MOD16: Mu et al. 2011).

There are multiple global initiatives globally to coordinate the efforts of monitoring SDG’s using RS data. One of the major initiatives is Group on Earth Observations (GEO) which is a partnership of more than 100 national governments and organizations with major objective being main streaming Earth observation data for decisions and actions towards sustainable development (Anderson et al. 2017). GEO initiated the development of Global Earth Observation System of Systems (GEOSS: Nativi et al. 2020) with the Committee on Earth Observation Satellites (CEOS) contributing to the technical side of the development including initiatives to promote RS for SDG monitoring such as EO4SDG (Kavvada et al. 2022).

In this paper, we investigate how RS can support monitoring of SDG 6.4.1 on water use efficiency. We modelled the agricultural component of SDG 6.4.1. with RS data using two different approaches: (i) using volume of consumptive water use (actual evapotranspiration) and (ii) using the water productivity concept. We applied both approaches to the country Lebanon as a case study and compared the results to the standard calculations using AQUASTAT data.

Methodology

Case study: Lebanon

Lebanon has an area of 10,452 km2 and is divided into 7 governorates (Fig. 1). The annual rainfall ranges between 700 and 1,500 mm/year, averaging 910 mm/year. Lebanon is considered rich in water resources compared to its surrounding countries, with 12 permanent watercourses with a discharge of 3,542 million m3/year along with 2,000 major springs which provide 630 million m3/year (Shaban 2020).

Fig. 1
figure 1

Map of Lebanon showing distribution of irrigated areas in governorates in 2020. (source of data: WaPOR)

Full size image

Agricultural lands occupy more than one third of the area (3,600 km2), of which 1040 km2 is equipped for irrigation and 900 km2 was actually irrigated as recorded in the 1998 census (FAO 2008). About 42% this area is located in the Bekaa Valley (FAO 2008). The main irrigated crops are wheat (24%), potatoes (18%), citrus (16%) and vegetables (14%). The source of water for irrigation is evenly distributed between surface water (48%) and groundwater (52%) (Shaban 2014). The agricultural sector contributes 5% of GDP and employs 8% of the labour force (FAO-Lebanon 2022).

Methods

This study focuses on modelling the agricultural component of the SDG indicator 6.4.1 (Awe) using two different RS-based approaches (Awp) and compare the results to the standard equation of Awe. The analysis was conducted using annual timesteps for a 5-year period starting with the baseline for the SDG in 2015 to 2020. We used a geospatial modelling approach (Chen et al. 2020; Koubarakis 2023) to provide a comprehensive analysis of water use efficiency and productivity in agricultural systems.

The standard equation for the agricultural component of SDG indicator 6.4.1 (Awe) is defined in FAO (2019) as:

$${A_{we}} = \frac{{GV{A_a}(1 - {c_r})}}{{{V_a}}}$$
(1)

Where:

Awe represents the water use efficiency of irrigated agriculture in USD/m3.

GVAa is the Gross Value Added to gross domestic product (GDP) by agriculture in USD/year.

cr is the proportion of agricultural GVAa produced by rainfed agriculture.

Va is the volume of water used by the agricultural sector in m3/year.

The equation focusses on the gross value added to GDP by irrigated agriculture (represented by GVAa* (1 - cr)) as this is where the water withdrawals contribute to. This ensures that the Awe calculation focuses solely on the efficiency of the water withdrawals in agriculture.

For the RS-based approaches for modelling the agricultural component of SDG 6.4.1, we developed two approaches. The first approach modifies Eq. 1 by replacing the water withdrawn (Va) with the total volume of the blue water consumed (VETb) derived from water supplied through irrigation (Eq. 2).

$${A}_{wp1}=\frac{{GVA}_{a}\left(1-{C}_{r}\right)}{{V}_{ETb} }$$
(2)

Where:

Awp1 is the water productivity of irrigated agricultural in USD/m3 using approach 1

VETb is the volume of water consumed by irrigation (blue ET) in m3/year.

To estimate VETb the first step is to separate ET into water from rainfall (ET green) and irrigation (ETb) (Chukalla et al. 2015). Various models exist to separate ET, which have different level of complexity and without clear validation data it is not possible to identify the best model (Msigwa et al. 2021). Therefore, for this purpose, we used a rather simple model by considering effective rainfall as ET green with the remainder of ET to be attributed to ETb. The effective rainfall and ETb is calculated on a monthly basis and summed to an annual value using Eq. 3:

$${ET}_{b}={\sum }_{n=1}^{12}max(ET -{P}_{e},0)$$
(3)

Where:

ETb is in mm/year,

ET is in mm/month, and

Pe is effective rainfall in mm/month.

Pe is computed from monthly precipitation by applying the conditions in outlined in the Eq. 4, adapted from Brouwer and Heibloem (1986):

\(Pe = \max (0.8*P - 25,0)\), if P > 75 mm/month (4)

\(Pe = \max (0.6*P - 10,0),\) if P ≤ 75 mm/month

Annual ETb is then converted to volume (VETb) in m3/year by multiplying the average ETb of the irrigated areas (\({\stackrel{-}{\text{E}\text{T}}}_{\text{b}}\)) with the irrigated area (Airr) as given in Eq. 5.

$${V_{ETb}} = \overline {E{T_b}} \times {A_{irr}}$$
(5)

The second approach starts with calculating Biomass Water Productivity (WPb) at pixel scale in kg/m3. WPb is derived from dividing the Total Biomass Production or Land Productivity (LP in kg/ha) by the irrigation water consumed (ETb). This is represented by the following equation (Eq. 6):

$${WP}_{b}= \frac{LP }{{ET}_{b} }$$
(6)

WPb is converted to economic water productivity by multiplying the average price per unit of annual biomass production from irrigated agriculture as presented in the following equation (Eq. 7):

$${A}_{wp2}= {WP}_{b} \times \stackrel{-}{p}$$
(7)

Where:

Awp2 is the agriculture water productivity in USD/m3 using approach 2.

WPb is the Biomass Water Productivity in kg/m3.

\(\stackrel{-}{\text{p}}\) is the average value added in USD per kg of annual biomass production

The average price is obtained by dividing GVAa of irrigated agriculture with the total LP of the irrigated area. This second approach is modelled at the grid or pixel level, which fills the gap that national level indicators are not able to provide all the required information for decision makers (Giuopponi et al. 2018; Saner et al. 2020). The analyses enables the spatial disaggregation of the results to sub-national levels (governorates and villages), and thereby providing an assessment of spatial variability of Awp within a country.

Data sources

Table 1 shows an overview of the data used in this research and for each approach, it consists of both secondary data and RS derived data. Awe is calculated using solely the secondary data from AQUASTAT (FAO 2022). This secondary data consists of annual country-level data on agricultural water withdrawals (Va), gross value added to GDP from agriculture (GVAa), and the Proportion of GVAa produced by rainfed agriculture (cr). Except for Va, this data was also used to model the two RS-based approaches. For the RS-based approaches, the data is complemented by RS based ET and LP. This data was obtained from FAO’s portal to monitor water productivity through open access RS derived data (WaPOR) (FAO 2023). WaPOR data is available in three resolutions, for Lebanon the highest resolution available covering the entire country of 100 m (WaPOR level 2) was used in this study.

Table 1 Data and corresponding sources employed in each of the approaches
Full size table

Results and discussion

National level

The results of the two approaches for calculating Awp were compared to Awe (Fig. 2). The values of Awp1 range between 1.93 USD/m3 in 2020 to 2.46 USD/m3 in 2015, with Awp2 showing a slightly higher estimation compared to Awp1. The three curves show similar trends over time, with a decline from 2015 to 2016, a general increase from 2016 to 2019 and a sharp decline in 2020. The results show a high correlation between Awp1 and Awp2 (r2 = 0.9), and lower correlation between the two approaches for Awp and Awe (r2 = 0.5).

Fig. 2
figure 2

Comparison between Awe and the two approaches for estimating Awp

Full size image

In absolute values, however, Awp1 and Awp2 are almost a factor 2 higher than Awe. With the GVAa for Awe and Awp1 being the same, the difference is in the estimation of the water withdrawn or consumed. Comparing the annual irrigated areas obtained from the WaPOR and those from AQUASTAT, it shows a significant difference. The WaPOR LCC map estimates an area between 354 and 432 km2 as irrigated, which is less than half of the area reported by the Lebanese government in AQUASTAT (900 km2) (Fig. 2A). A major distinction between the two is that the AQUASTAT data is based on area equipped for irrigation, which is a static value, whereas WaPOR uses a water deficit index based on precipitation and evapotranspiration to estimate if a cropland is irrigated or not, which varies year by yearFootnote 2 (FAO 2020).

When comparing the volume of water used for irrigation per area (in mm/year) (Fig. 2B), the values from both sources have a closer match. The WaPOR ETb data ranges between 449 mm/year in 2020 to 604 mm/year in 2015 (Fig. 2B), while AQUASTAT reports the volume of water withdrawn by agriculture (WWagric) to be 0.7 km3/year (equivalent to 777 mm/year), and crop water requirement (CWR) of 0.529 km3/year (equivalent to 587 mm/year). The discrepancy in the last two years could be attributed to fields being abandoned because of the economic hardship faced by farmers in Lebanon. Despite these differences, the order of magnitude in estimated water depth estimated using the WaPOR data are reasonable.

As shown in Fig. 2, the results show a consistent drop in Awe and both Awp approaches between 2019 and 2020. Comparing WPb with Awp2 shows that the main culprit for the decline in Awp2 is not in efficient water use in the agricultural sector as WPb shows an overall increasing trend from 2015 to 2020, from 2.1 to 2.2 kg/m3 in 2015 to 3.18 kg/m3 in 2020 (Fig. 4). Awp2 followed the same trend as WPb until 2019, starting at 2.79 USD/m3 in 2015 to reach 3.21 USD/m3 in 2019. After 2019, the two indicators start deviating with WPb increasing to 3.18 kg/m3, and Awp2 dropping to 2.2 USD/m3. The drop in Awp2 is therefore not the result of a reduced performance of the agricultural sector but is due to factors affecting the economy in Lebanon 2020, with GVAa dropping from 634 million USD in 2018 to 339 million USD in 2020. At the same time, the results show that WPb continues to increase over time.

Fig. 3
figure 3

A) Area under irrigation from WaPOR and AQUASTAT and B) average ETb from WaPOR over irrigated areas versus agricultural water withdrawal (WWagric) and estimated crop water requirement (CWR) from AQUASTAT for Lebanon between 2015 and 2020

Full size image
Fig. 4
figure 4

Comparison between WPb, Awp2 and GVAa between 2015–2020, for irrigated agriculture in Lebanon

Full size image

Sub-national level

The spatial disaggregation of LP, WPb and Awp2 per governorates and villages provides additional information related to the performance of the irrigated areas in the different administrative zones (Fig. 5). The highest LP was found in the North and South with a LP between 20 and 25 ton/ha.

Fig. 5
figure 5

A) Water Productivity (WPb), B) Land Productivity (LP) and C) Awp2 per governorates in 2015; D) Awp2 per village in 2015

Full size image

The highest WPb during these five years was Bekaa governorate in 2020. However, the highest average WPb of all the governorates was Akkar with 2.8 kg/m3 and the lowest was recorded in the North Governorate with 2.4 kg/m3 (Fig. 5D). Overall, WPb per governorate was increasing over the years (Fig. 6A), however Awp2 was decreasing, especially in the last year of the analyses (Fig. 6B).

Fig. 6
figure 6

A) Annual WPb and B) annual Awp2 per governorate in 2015

Full size image

While Lebanon’s average AWp2 between 2015 and 2020 was 2.78 USD/m3, the spatial disaggregation per governorate or even village shows high variability, with Awp2 in the range between 1.8 and 3–3.5 USD/m3 (Fig. 6B). These results show that reporting of Awp2 (and therefore Awe) at country level masks the spatial variability within a country and thereby hiding the issues shown when data is disaggregated. Thus, the spatial disaggregation can actually identify hotspots and assist in identifying factors that can explain successes and thus measures for improvement.

Implications for SDG 6 reporting

In the context of reporting SDG indicator 6.4.1 for Lebanon, Fig. 7 shows the indicator increasing from 17.5 USD/m3 in 2000 to 25 USD/m3 in 2019. While the analysis shows that the indicator increased significantly over this period, the changes since 2015 (the baseline of the SDGs) were negligible. The contribution of agriculture to the WUE was relatively small (Fig. 7), although agriculture is a major consumer of water (Fig. 8a).

Fig. 7
figure 7

Contribution of the different sectors to SDG Indicator 6.4.1 water use efficiency for Lebanon over time between 2000 and 2019. (Source of data: UN-Water 2022)

Full size image
Fig. 8
figure 8

(a) water withdrawal per sector; (b) GVA per sector over time between 2000 and 2019. (Source of data: UN Water 2022)

Full size image

The data breakdown for SDG indicator 6.4.1 reveals that water withdrawals for the agricultural sector decreases over time while its contribution to the GVA remains relatively constant. The main jump in the overall SDG indicator 6.4.1 between 2000 and 2015 is thus clearly related to the increase in industrial water withdrawals, which has grown nine-fold from 2000 to 2017, surpassing agriculture, while its GVA contribution declines (Fig. 8A and B).

Conclusion

The results of the different approaches show that RS based data can be a valuable source of data for regular reporting and monitoring of SDG indicator 6.4.1. By replacing the water withdrawn by the water consumed, the latter of which can be observed by RS, a proxy to the conventional indicator is estimated. The results of these different approaches to estimate Awp show similar trends as Awe. While the overall results on trends are encouraging, the absolute values differ substantially. One main factor is the large inconsistencies in the data on area irrigated in Lebanon between the WaPOR data and the data reported in AQUASTAT. A possible explanation of this discrepancy can be related to the different definitions used in AQUASTAT where the irrigated area is the area equipped with irrigation, whereas WaPOR uses the water deficit index to identify areas that are actually irrigated (FAO 2020). Arguably, monitoring progress benefits more from information on actually irrigated areas and the water consumed in those areas compared to the potential irrigated areas and the estimated water supply to these areas as reported in AQUASTAT.

A second source of discrepancy is that the Awp approaches only estimate water productivity in irrigated areas and does not consider the aquaculture and livestock components as per the definition of SDG indicator 6.4.1. In the case of Lebanon, the volume of water consumed by irrigation contributes to an important part of the total water withdrawn by agriculture, and therefore will not affect the results.

The methodology used in this paper can be easily extended globally. The combination of spatial and temporal analysis can give a better understanding of the spatial variations and can help incorporate policy changes, decisions, and actions toward Agenda 2030. Before applying the methodology to other countries, it is recommended to validate the RS based estimates to improve the reliability of the analysis and to evaluate if the assumption that irrigated agriculture is the largest consumer of water in agriculture is valid.