Anticipating and monitoring water risks for agriculture

Annex C. Techniques, methods and models

Hazard monitoring and short-term forecasting

Water excess (flood)

Flood forecast rating curves (FRC) can be used to provide early flood warnings of up to several days. An FRC indicates the ratio of measured upstream water levels in comparison to estimated downstream discharge. This enables downstream water level forecasts by assessing upstream water levels. FRCs can be used to provide timely early warning of downstream flood events and their probable severity. Upstream flows and water levels can be measured in situ or via satellite remote sensing.

Water shortage (drought)

Drought indices are used to characterise the complex nature of droughts with an indicator. Monitoring and forecasting tools frequently rely on them to illustrate drought severity in a quantifiable way. Hundreds of indices have been developed to date, each of which is calculated based on different data available. While the first tools initially considered precipitation as a sole factor, later indices started to consider variables such as soil moisture, evapotranspiration, vegetation, etc. Initially, drought indices were mainly calculated based on meteorological (weather station) data; however, as satellite technology developed, this was later complemented with remote-sensing data. Drought indices may be based on a single variable or several variables (compound indices). Due to the complexity of droughts, compound indices are increasingly used. The relevance of drought indices varies based on space and time, as well as drought type (Table C.1.). As several drought indices were developed for geographically specific conditions, their reliability may differ if applied in a different region or ecosystem (Hanadé Houmma et al., 2022[177]) (Liu et al., 2016[16]) (Mullapudi et al., 2023[178]).

Table A C.1. Drought types and commonly used drought indices

Drought type	Commonly used drought indices
Meteorological drought	Palmer drought severity index (PDSI): monitors moisture deficiency Standardized Precipitation Index (SPI): compares precipitation deficit to the climatological record
Agricultural drought	Indices derived directly from soil moisture, such as the crop moisture index (CMI), standardised soil moisture index (SSMI), soil moisture percentile (SMP), normalised soil moisture (NSM).
Hydrological drought	Palmer hydrologic drought index (PHDI, measures hydrological drought impacts, e.g. reservoir and groundwater levels), runoff or streamflow percentile, and standardised runoff index (SRI)

Note: The table provides a non-exhaustive list of drought indices commonly used to detect various drought types.

Source: (Hanadé Houmma et al., 2022[177]) (Liu et al., 2016[16]) (Mullapudi et al., 2023[178]).

Water quality

Multiparameter, low-power buoys deliver near-real-time alerts of episodic pollution in freshwater at a fraction of the cost of conventional sondes. Such platforms point towards scalable, high-resolution monitoring networks. The prototype by Krklješ et al. (2024[179]) demonstrated stable operation, and an algorithm successfully mitigated interference from suspended solids in the coliform channel. The system provides early-warning capability downstream: by flagging spikes in turbidity, conductivity or thermotolerant coliforms within minutes, it lets water utilities or environmental agencies intervene before episodic pollution events propagate.

Combining remote sensing with machine learning, a convolutional neural networks model was employed to establish the relationship between satellite (Sentinel-2) images and in-situ water quality levels of Lake Dianchi (Meng et al., 2024[180]). This approach is useful for forecasting algal blooms and trophic status. It revealed month-to-month changes in water-quality grades, providing local authorities with seasonally explicit targets for phosphorus mitigation.

Salls et al (2024[181]) applied machine-learning calibration of spectral indices (Maximum Chlorophyll Index, Normalised Difference Chlorophyll Index) to over 100 lakes in the United States, predicting chlorophyll-a with high accuracy and classifying trophic states at 82%, making it useful for forecasting eutrophication risk.

Luo et al. (2022[182]) at Lake Qiandao, China, combined high-frequency in-situ monitoring with automated quality control and machine learning to detect anomalies and provide early warning of water-quality deterioration.

Irrigation advice

The DISPATCH method (DISaggregation based on Physical And Theoretical scale CHange) combines microwave and optical remotely sensed data to assess soil moisture levels. A study carried out in the Iberian Peninsula in Spain showed that using these high-resolution remote sensing data allows accurate estimation of irrigation water requirements, improving water use efficiency and optimising agricultural productivity (Dari et al., 2021[81]); (Parra-López et al., 2025[80]). Similarly, the SiAR tool in Spain combines data from field weather stations with satellite imagery to map irrigated areas and estimate their annual and monthly water requirements based on the daily water balance for the whole country (www.espaciosiar.es.).

Hazard and risk mapping

Water excess (floods)

Flood modelling approaches underpin flood hazard and risk mapping. Table C.2 summarises the strengths and weaknesses of different flood modelling approaches.

Table A C.2. The strengths and limitations of different flood modelling approaches

Method	Strength	Limitation	Suitability
Empirical methods	Relatively quick and easy to implement Based on observation Derived inundation estimate is independent Technology is rapidly improving	Non-predictive No/indirect linkage to hydrology (difficult to use in scenario modelling) Coarse spatial and temporal resolution (although improving) Engineering limitations (sensors, carriers, transmission devices) Environmental limitations (clouds, wind, damaging weather conditions, other natural constrains) Processing limitations (algorithm, artificial errors …)	Flood monitoring Flood damage assessment Serve as observations to support calibration, validation and data assimilation for other methods
Hydrodynamic models	Direct linkage to hydrology Detailed flood risk mapping Can account for hydraulic features/structures Quantify timing and duration of inundation with high accuracy	High data requirements Computationally intensive Input errors can propagate in time	Flood risk assessment Flood damage assessment Real-time flood forecasting Flood related engineering Water resources planning Riverbank erosion Floodplain sediment transport Contaminant transport Floodplain ecology River system hydrology Catchment hydrology
Simplified conceptual models	Computationally efficient	No inertia terms (not suitable for rapid varying flow) No/little flow dynamics representation	Flood risk assessment Water resources planning Floodplain ecology River system hydrology Catchment hydrology Scenario modelling

Source: Reproduced from Teng et al (2017[91]).

Height Above Nearest Drainage and on-the-fly Flood Mapping: Traditional flood maps require large amounts of data to complete complex flow calculations. For many parts of Canada, these data are not available. Natural Resources Canada researchers developed a simplified flood model covering the entire country, which only requires topographic data of a watershed and the shape and depth of the river network. By calculating the height difference between the land grid and the river grid, the Height Above Nearest Drainage (HAND) model produced accurate results in only a fraction of the processing time. This paves the way to complete on-the-fly flood maps that can be used to assist first responders during a flood emergency in any region (Natural Resources Canada, n.d.[183]).

Water shortage (drought)

16. Agricultural drought modelling is used for monitoring and anticipating water shortage and encompasses the following approaches i) multivariate descriptive drought modelling of drought parameters (e.g. onset, intensity), ii) multivariate predictive modelling of drought risk, and iii) predictive modelling of expected drought impacts (Table C.3.) (Hanadé Houmma et al., 2022[177]).

Table A C.3. Three drought modelling approaches

Approach	Description	Recent advancements
Multivariate descriptive drought modelling of drought parameters	Combines variables and attempts to get qualitative and quantitative information on drought parameters for drought assessment and monitoring. Used to monitor ongoing drought conditions	Improved accuracy in drought parameter estimates thanks to the availability of multi-source data through the joint use of remote sensing and machine learning models (as opposed to previous methods using spatial interpolation of in-situ measurements or multivariate statistical modelling)
Multivariate predictive modelling of drought risk	Predictive mapping of expected drought risk, its intensity and spatial extent. Used for forecasting and can facilitate effective early responses to drought. Modelling the frequency of drought parameters can facilitate proactive management	Relatively new technique based on machine and deep learning.
Predictive modelling of expected drought impacts	Predictive modelling primarily based on the history of several biophysical and hydroclimatic variables. Contextual socio-economic vulnerabilities are rarely considered. Highly challenging field, as drought impacts depends on other predicted variables (e.g. duration, drought intensity, etc.)	Self-learning methods based on deep and machine learning algorithms, but little accuracy so far. Integration of multi-sensor predictive modelling as a possible avenue

Source: (Hanadé Houmma et al., 2022[177]).

Impact assessment

Water excess (flood)

The floodam-agri tool allows the estimation of flood damage at the plot level. It models yield losses and short-term management strategies based on knowledge from agricultural experts from the region under study. The tool estimates the damage as a variation in added value resulting from a flood in terms of product loss (yield) and intermediate consumption. It was set up for the construction of damage functions on a national scale in France as part of cost-benefit analyses for the evaluation of flood management projects (Modjeska and Brémond, 2022[94]).

AGRIDE-c, a conceptual model for the estimation of flood damage to crops. AGRIDE-c is a synthetic, expert-based model relying on the investigation of flood damage mechanisms and related economic impacts to agricultural crops. It adopts a coupled approach considering the physical damage to crops (i.e. yield reduction as a function of hazard and vulnerability features) and its economic implications in terms of loss of revenue and variation in production costs, which also depend on farmer's alleviation strategy after the flood. The general framework can be adapted and replicated in different regions. Takes into account types of crops, climatic region, crop calendars, flood damage mechanisms, crop yields and prices, cultivation practices and related costs); this is one of the main added values of AGRIDE-c, which makes it suitable to be adapted and/or transferred to other regions (Molinari et al., 2019[95]).

The study “Agricultural flood vulnerability assessment and risk quantification in Iowa” uses the Agricultural Flood Damage Analysis Model (AGDAM), crop layer raster datasets and flood inundation maps to assess agricultural flood risk in the state, focusing on corn, soybean and alfalfa crops. It stresses the importance of high-quality crop information and flood inundation data for the accurate estimation of agricultural loss quantification. It finds that terrain-based flood mapping products (Height Above the Nearest Drainage, HAND) perform better than FEMA maps for agricultural flood loss analysis (Yildirim and Demir, 2022[184]).

A continuous consequence/probability diagram may be a suitable technique to assess risk. For the hazards considered (e.g. flood, drought, salinity), risk is divided in three levels in the consequence/probability diagram corresponding to different bands: a) high risk (red band), where the level of risk is considered intolerable and risk treatment is essential whatever its cost; b) medium risk (yellow band), where the risk is considered tolerable; c) low risk (green band), where the level of risk is considered negligible, so no risk treatment measures are needed. Risk tolerance limits depend on the study area characteristics and should be defined based on information from past events and risk owner judgement. After establishing the risk context for the area where the approach is applied, hazard scenarios based on historical data and stakeholder’s information have to be defined to support risk assessment. Consequence descriptors are evaluated for the defined scenarios and risk levels are determined, compared and evaluated against risk criteria and tolerance limits previously defined. Results provide scientifically supported information to help stakeholders and risk owners to discuss the acceptability of the risk magnitude. The consequence descriptors can be evaluated through the analysis of model results, and historical and monitoring data. The tool uses hazard scenarios and numerical models to assess risk. This is combined with historical data and stakeholders’ input to define the risk context including risk management objectives (Freire et al., 2021[185]).

Water resource assessment

Software, models

The Soil and Water Assessment Tool (SWAT) is one of the most widely used watershed scale tools to simulate the quality and quantity of surface and ground water (Bieger et al., 2016[186]). The tool can predict the effect of soil, land use and management on water quantity and quality across different spatial and temporal scales and under different climate scenarios. It is computationally efficient and requires input data that is usually available. As an open-source software, its algorithms are available to all, which has led to continued iteration and development of the tool. Despite its demonstrated flexibility and effectiveness, several weaknesses and limitations of the model remain.

Water Evaluation and Adaptation Planning (WEAP) is a software tool for integrated water resources planning. WEAP integrates demand-side factors such as water use patterns, equipment efficiencies, re-use strategies, costs, and water allocation schemes with supply-side components such as stream flow, groundwater resources, reservoirs and water transfers (WEAP, 2025[187]). As a result, the tool can be used to examine alternative water development, management and policy options. WEAP operates on the basic principle of a water balance and can be applied to diverse systems, including agricultural systems. WEAP provides a system for maintaining water demand and supply information as well as the capability to simulate a broad range of natural and engineered components, including rainfall runoff, baseflow, groundwater recharge, water conservation, allocation priorities, pollution, and ecosystem needs. A financial analysis module allows cost-benefit comparisons.

The Dynamic Water Resources Assessment Tool (DWAT) developed by the WMO assesses the sources, extent, dependability and quality of water resources. It can be used to assess land-use changes within a basin over time, and impacts on water availability under different scenarios including climate change. DWAT uses a distributed conceptual scheme for water cycle analysis and contains sub-algorithms, such as evapotranspiration, infiltration, watershed runoff, groundwater movement, and channel routing (World Meteorological Organization, 2025[188]).

Water footprint

Tuninetti et al (2019[189]) combine data on water use efficiencies (crop water footprint, CWF) and crop water consumption (water footprint) with a “water debt” (WD) indicator that captures the time needed for soil moisture, surface water and groundwater to replenish: the “water debt repayment time” indicator. The water debt repayment time quantifies the local mismatch between water use and availability. This Water Debt indicator “allows one to discriminate two countries showing the same water use efficiency and producing nearly the same amount of crop but generating a different impact on water resource depending on the local water availability”.