A multi-criteria framework for assessing the natural resource potential of agricultural landscapes using Harrington’s generalized desirability function Authors: Aidos Omarov, Zhumakhan Mustafayev, Irina Skorintseva, Gulnar Aldazhanova, Amanzhol Kuderin, Askhat Toletayev Published: 2026 ABSTRACT The assessment of the natural resource potential (NRP), as a multi-criteria function of the agroclimatic, pedological, biological, land, and water resources of agricultural landscapes, constitutes a complex challenge involving variables of differing physical nature and dimensional scales. Traditional additive scoring models suffer from compensatory masking, where abundance in one environmental resource artificially obscures a critical deficit in another. The aim of this study is to overcome this limitation by developing a novel non-compensatory hierarchical mathematical framework for NRP assessment, utilizing a geometrically aggregated Harrington generalized desirability function as the core analytical engine. The study employs systems analysis, deterministic non-linear transformations, and mathematical modeling validated against long-term environmental data across spatial and temporal dimensions, disaggregated by natural zones and administrative districts of the Ayagoz District (Abai Oblast, Kazakhstan). The constructed framework successfully operationalizes a multi-level index where twenty heterogeneous natural indicators are transformed into a single continuous, dimensionless scale [0, 1]. By using geometric rather than additive aggregation, the framework strictly complies with ecological boundary conditions, ensuring that if any vital resource subsystem drops below a critical threshold, the entire integral potential deterministically decreases. The resulting models shift land evaluation from discrete, subjective expert scoring matrices to a continuous computational space, generating highly objective, evidentially valid, and spatially explicit cartographic and numerical outputs. The application of this framework to the agro-landscapes of the Ayagoz District demonstrates three definitive advantages: first, it eliminates the mathematical errors inherent in additive indexing; second, it allows each specific resource parameter to be scaled in strict accordance with its non-linear ecological limits; and third, it enables direct translation of the quantitative integral index into qualitative modal categories of desirability for regional land-use optimization. The conceptual and methodological novelty lies in the advancement of landscape-evaluation theory through the transition from compensatory matrices to a rigid, ecologically grounded non-compensatory architecture. For the first time, Harrington’s algorithm is restructured into a multi-tiered, hierarchical system that mathematically reflects the ecological laws of the minimum (Liebig) and tolerance (Shelford) at a landscape scale. It conceptualizes the natural system not as a static sum of assets, but as a dynamic, information-driven cascade where hydrological and pedogenic constraints sequentially filter and determine the final expression of biological and agricultural productivity. The resulting mathematical platform provides a scalable, machine-compatible instrument for digital land-use management and automated environmental monitoring. The spatial–temporal modeling outputs serve as predictive tools for agricultural planning, allowing regional authorities to simulate resource degradation risks and optimize the ecological reproduction of semi-arid agricultural landscapes under climate instability. Introduction Agricultural landscapes are increasingly understood as complex multifunctional socio-ecological systems in which long-term productivity depends on the coupled performance of climatic, soil, hydrological, biological, and land-use resource subsystems rather than on any single environmental factor considered in isolation (Basso & Antle, 2024; Selin-Noren et al., 2026). Recent international studies in sustainable agricultural landscape management have emphasized that maintaining agricultural resilience under climate variability now requires integrated resource-diagnostic frameworks capable of simultaneously capturing ecosystem service supply, environmental limitations, and land productivity constraints at territorial scales (Tran et al., 2022; Bergez et al., 2022). Under conditions of accelerating land degradation, hydroclimatic instability, and increasing pressure on freshwater availability, the objective quantification of the natural resource potential (NRP) of agricultural landscapes has therefore become a central requirement for agroecological planning, regional land-use optimization, and sustainable food-production policy (Mustafayev et al., 2024a). Natural resource potential may be defined as the integrated capacity of an agricultural landscape to sustain biomass production and ecological stability through the combined functioning of its agro-climatic, soil, biological, water, and land productivity resources (Jansen & Kalas, 2024; Milheiras et al., 2022). In dryland territories, where environmental balances are controlled by strong moisture deficits and highly variable climatic forcing, this integrated capacity is especially sensitive to limiting-factor interactions among heterogeneous resource components. Contemporary soil and land system research increasingly recognizes that sustainable territorial management cannot rely on fragmented single-indicator diagnostics, because ecosystem functioning emerges from the systemic interaction of multiple environmental processes operating at different temporal and spatial scales (Maialetti et al., 2024). Consequently, agricultural resource assessment is progressively shifting from isolated climatic or pedological evaluations toward holistic multi-parameter frameworks designed to represent landscape multifunctionality and cumulative environmental performance. Despite this conceptual progress, the methodological implementation of integrated NRP assessment remains unresolved. Most existing agricultural suitability and land evaluation studies continue to employ weighted scoring systems, expert bonitation procedures, fuzzy membership approaches, or pairwise multi-criteria ranking models such as the Analytic Hierarchy Process (Herzberg et al., 2019; Reed et al., 2023). While practically convenient and widely implemented within GIS-based suitability mapping frameworks, these conventional approaches retain several persistent scientific limitations: (a) a high reliance on subjective expert weighting during factor prioritization, which introduces investigator bias; (b) discontinuous class boundaries that reduce sensitivity to gradual environmental change; (c) weak temporal comparability; and (d) limited compatibility with automated digital monitoring workflows. These structural vulnerabilities within the Multi-Criteria Decision Analysis (MCDA) framework have been comprehensively highlighted in recent international research, particularly in multi-sensor agricultural risk assessments where rigid boundary artifacts obscure continuous ecological realities (Lagiewska & Bartold, 2025). Furthermore, within such GIS-based spatial diagnostics, the reliability, quality, and internal consistency of the input environmental datasets remain a critical, yet frequently overlooked, bottleneck. Even when operating with officially distributed, high-tier satellite products provided under major international frameworks such as NASA, ESA, and the Copernicus Programme (e.g., MODIS, Sentinel, and Landsat fields), significant uncertainties persist regarding atmospheric correction artifacts, cloud-masking gaps, and cross-sensor calibration discrepancies under high hydroclimatic variability (Lagiewska & Bartold, 2025). If processed through traditional additive linear frameworks, these subtle data inconsistencies are often smoothed over or masked, leading to error propagation in regional outputs. In contrast, transitioning to a non-compensatory hierarchical desirability architecture ensures that any underlying data distortion or genuine ecological deficit in vital resource subsystems is deterministically exposed rather than masked, making the framework highly resilient to input data constraints. Even where advanced composite sustainability indices are used, a common unresolved difficulty remains the mathematically consistent normalization of heterogeneous agroecological indicators expressed in fundamentally different physical units into one dimensionless and reproducible resource criterion (Kiryushin, 2020; LaRota-Aguilera et al., 2024; Pollesch & Dale, 2016). A promising deterministic solution to this aggregation problem is provided by Harrington’s generalized desirability function (Harrington, 1965). The theoretical advantage of this approach lies in its continuous double-exponential transformation of any real-valued environmental indicator into a bounded desirability score on the interval [0,1], thereby allowing physically incommensurable variables to be interpreted within a common mathematical space (Semeykin et al., 2023; Park & Park, 1998). Because the generalized desirability is subsequently calculated as a geometric mean, the final composite index becomes highly responsive to limiting factors: critically low values of any subsystem proportionally depress the integral assessment. Such behavior is ecologically important for agricultural landscapes governed by Liebig-type minimum constraints and Shelford-type tolerance thresholds, where the productive functioning of the whole system is frequently determined by its weakest resource component. Owing to these properties, Harrington-based modelling has shown strong applicability in water-quality diagnostics, environmental pollution monitoring, forest ecosystem evaluation, and integrated sustainability studies requiring non-subjective multidimensional normalization (Guedri & Ghanmi, 2020; Luo et al., 2025). However, the conceptual and methodological value of this study extends significantly beyond a standard regional case study or a routine replication of Harrington’s algorithm. The core theoretical novelty of this work addresses a major gap in the international literature by linking Harrington’s mathematical transformation directly to landscape-evaluation theory and systemic ecological boundaries through three distinct advancements over recent comparative approaches:First, it establishes a rigid, non-compensatory hierarchical architecture. By structuring 20 heterogeneous variables into five autonomous environmental subsystems aggregated via a strict geometric mean, the framework ensures that any resource drop below critical threshold values deterministically depresses the entire integral potential. This explicitly operationalizes the laws of Liebig and Shelford at a landscape scale, preventing the mathematical masking common to additive models.Second, the model formalizes the dynamic concept of the biological productivity cascade, mathematically capturing how theoretical agro-climatic potential is sequentially filtered and depleted by localized pedogenic limits and hydrological deficits.Third, by replacing traditional seven-point expert bonitation and discrete ordinal scales with a continuous computational space [0, 1], this approach provides the mathematical stability, continuity, and lack of edge-discontinuities necessary for automated digital mapping. First, it establishes a rigid, non-compensatory hierarchical architecture. By structuring 20 heterogeneous variables into five autonomous environmental subsystems aggregated via a strict geometric mean, the framework ensures that any resource drop below critical threshold values deterministically depresses the entire integral potential. This explicitly operationalizes the laws of Liebig and Shelford at a landscape scale, preventing the mathematical masking common to additive models. First, it establishes a rigid, non-compensatory hierarchical architecture. By structuring 20 heterogeneous variables into five autonomous environmental subsystems aggregated via a strict geometric mean, the framework ensures that any resource drop below critical threshold values deterministically depresses the entire integral potential. This explicitly operationalizes the laws of Liebig and Shelford at a landscape scale, preventing the mathematical masking common to additive models. Second, the model formalizes the dynamic concept of the biological productivity cascade, mathematically capturing how theoretical agro-climatic potential is sequentially filtered and depleted by localized pedogenic limits and hydrological deficits. Second, the model formalizes the dynamic concept of the biological productivity cascade, mathematically capturing how theoretical agro-climatic potential is sequentially filtered and depleted by localized pedogenic limits and hydrological deficits. Third, by replacing traditional seven-point expert bonitation and discrete ordinal scales with a continuous computational space [0, 1], this approach provides the mathematical stability, continuity, and lack of edge-discontinuities necessary for automated digital mapping. Third, by replacing traditional seven-point expert bonitation and discrete ordinal scales with a continuous computational space [0, 1], this approach provides the mathematical stability, continuity, and lack of edge-discontinuities necessary for automated digital mapping. This methodological gap has been particularly evident in Central Asian dryland agriculture, where productive land use is strongly constrained by unstable hydrothermal balances and uneven territorial distribution of water and soil resources (Umirzakov et al., 2025). In Kazakhstan, recent agro-resource investigations have established the principal ecological subsystems controlling agricultural landscape productivity and demonstrated the feasibility of integrated territorial diagnostics based on climatic, soil-land, agrobiological, and hydrological indicators (Mustafayev et al., 2024b, c). However, these earlier implementations relied primarily on discrete seven-point expert scoring procedures, which preserved a degree of methodological subjectivity and limited both analytical continuity and digital scalability. At the same time, recent international work on responsible digitalization in agriculture has underlined that future environmental decision-support systems require mathematically reproducible, machine-compatible, and transferable indicator architectures capable of functioning within automated monitoring platforms rather than within manually interpreted expert matrices (Bijlmakers et al., 2022; Cravero et al., 2026). Accordingly, a substantial scientific need persists for a continuous, dimensionless, and computationally stable framework capable of integrating heterogeneous environmental resource indicators into a unique natural resource potential criterion without subjective weighting distortion. The present study addresses this need by developing a transferable multi-criteria assessment model based on Harrington’s generalized desirability function. Five key resource subsystems — agro-climatic, soil, land, biological, and water resources — are formalized as independent desirability structures, their heterogeneous indicators are transformed into a unified mathematical desirability space, and the subsystem outputs are aggregated into a single integral generalized desirability function of natural resource potential (GDFNRPL). The methodology is validated using a 32-year archival environmental dataset (1992–2024) from the Ayagoz district of the Abai region, Kazakhstan, selected as a representative semi-arid agricultural territory characterized by pronounced hydroclimatic instability, fluctuating water availability, and long-term soil productivity constraints. We hypothesize that the generalized desirability architecture provides (1) a more mathematically stable and ecologically sensitive estimate of integrated agricultural landscape resource potential than conventional ordinal scoring systems, and (2) clearer identification of the dominant limiting subsystems controlling long-term agroenvironmental dynamics. The resulting framework is intended as a transferable methodological platform for regional agroecological monitoring, comparative landscape diagnostics, and digital decision support for sustainable land management in agricultural regions. Materials and methods Study area and environmental data sources The study was conducted for the Ayagoz district, located in the Abai region of eastern Kazakhstan (approximately 47°–49° N and 79°–82° E), occupying an area of about 49,610 km2 (Fig. 1). The district represents a typical semi-arid agricultural landscape of the Kazakh steppe zone, characterised by sharply continental climatic conditions, low and highly variable precipitation (230–370 mm annually), warm summers (mean July temperature approximately 22 °C), cold winters (mean January temperature approximately − 16 °C), and a mosaic of dry steppe, semi-desert pasture, and cultivated grain-producing lands. The territory is hydrologically dependent on the Ayagoz River basin, which constitutes the principal source of renewable surface water for agricultural and domestic use.Fig. 1Geographic position of the study area showing Ayagoz district within Abai region and the Republic of Kazakhstan Geographic position of the study area showing Ayagoz district within Abai region and the Republic of Kazakhstan The selection of the Ayagoz district as a model territory was caused by three methodological considerations: (a) the presence of long-term continuous archival meteorological and hydrological observations; (b) the representative combination of climatic moisture deficit, moderate soil fertility, and unstable water availability typical for eastern Kazakhstan; and (c) the predominance of agriculture-oriented land use, which allows direct interpretation of natural resource constraints in relation to agroecosystem productivity. Environmental input data were compiled from three principal sources. Monthly climatic observations for the period 1992–2024, including air temperature, atmospheric precipitation, and relative humidity, were obtained from the archival database of the Republican State Enterprise Kazhydromet. Hydrological information on renewable river runoff, water intake volumes, and water supply characteristics was derived from the same long-term monitoring archive. Soil agrochemical parameters, including humus reserves, nutrient availability, hydrolytic acidity, and degree of salinization, were extracted from regional soil survey reports and analytical inventories maintained by the Institute of Geography and Water Security. To evaluate long-term changes while preserving interdecadal comparability, five reference years (1992, 2000, 2010, 2020, and 2024) were selected as representative temporal checkpoints covering the full 32-year observation interval. This design allows the identification of both gradual environmental trends and contrasting hydrological-climatic states without overloading the model with redundant annual fluctuations. General workflow of NRP assessment The methodological framework developed in this study treats the natural resource potential (NRP) of agricultural landscapes as an integrated emergent property formed by the interaction of five environmentally limiting subsystems: climatic resources, soil resources, land productivity resources, biological productivity resources, and water availability resources (Fig. 2). Each subsystem is represented by a set of physically interpretable quantitative indicators describing its contribution to the agroecological functioning of the landscape (Mustafayev et al., 2023).Fig. 2Methodological framework for multi-criteria assessment of natural resource potential using Harrington’s generalized desirability function Methodological framework for multi-criteria assessment of natural resource potential using Harrington’s generalized desirability function The modelling workflow consisted of four consecutive analytical stages:Selection and ecological justification of subsystem indicators based on their relevance to agricultural productivity formation;Transformation of all original indicators into coded dimensionless variables using relative scaling procedures;Conversion of coded variables into individual desirability scores by Harrington's double-exponential desirability function;Hierarchical aggregation of desirability scores first into subsystem generalized desirability functions and then into the overall generalized desirability function of natural resource potential (GDFNRPL). Selection and ecological justification of subsystem indicators based on their relevance to agricultural productivity formation; Selection and ecological justification of subsystem indicators based on their relevance to agricultural productivity formation; Transformation of all original indicators into coded dimensionless variables using relative scaling procedures; Transformation of all original indicators into coded dimensionless variables using relative scaling procedures; Conversion of coded variables into individual desirability scores by Harrington's double-exponential desirability function; Conversion of coded variables into individual desirability scores by Harrington's double-exponential desirability function; Hierarchical aggregation of desirability scores first into subsystem generalized desirability functions and then into the overall generalized desirability function of natural resource potential (GDFNRPL). Hierarchical aggregation of desirability scores first into subsystem generalized desirability functions and then into the overall generalized desirability function of natural resource potential (GDFNRPL). Thus, the proposed approach combines heterogeneous climatic, pedological, hydrological, and biological indicators within a mathematically unified and computationally reproducible dimensionless framework. The conceptual relationship among the five resource subsystems may be expressed as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$NRP=f(CR,SR,LR,BR,WR)$$\end{document}NRP=f(CR,SR,LR,BR,WR)where CR = climatic resources, SR = soil resources, LR = land resources, BR = biological resources, and WR = water resources. Indicator system and ecological justification Before mathematical normalization, a unified indicator system was constructed to represent the five principal natural resource subsystems controlling agroecosystem functioning. Indicator selection was based on three criteria: (a) physical relevance to agricultural productivity formation; (b) long-term data availability in the regional monitoring archives; and (c) compatibility with dimensionless transformation (Table 1). Table 1Indicator system used in the generalized desirability model of agricultural landscape NRPResource subsystemIndicator codeIndicatorUnitEcological roleClimaticRSATRelative sum of biologically active temperaturesdimensionlessthermal supplyClimaticRRBRelative radiation balancedimensionlessenergy inputClimaticRSTERelative total evaporationdimensionlessatmospheric demandClimaticRTWCALRelative total water consumptiondimensionlessevapotranspiration demandClimaticRAAPRelative annual precipitationdimensionlessmoisture supplyClimaticRADAHRelative total air humidity deficitdimensionlessatmospheric thirstClimaticRNHCRelative natural humidity coefficientdimensionlessmoisture adequacyClimaticRRDIRelative radiation dryness indexdimensionlessaridity pressureSoilRESFRelative energy of soil formationdimensionlesspedogenic energySoilRESIRelative soil fertility indexdimensionlessagrochemical fertilityLandEPALEnergy productivity coefficientdimensionlessecological land productivityLandBEPALBioenergetic productivity coefficientdimensionlessbiomass energy productivityLandCIBPLBioclimatic productivity coefficientdimensionlessclimatic-biological productivityBiologicalCPPClimatic potential productivity ratiodimensionlessclimatic yield constraintBiologicalMPPMaximum possible productivity ratiodimensionlesssoil-climatic productivityBiologicalRMPPActual maximum possible productivity ratiodimensionlessmoisture-limited productivityBiologicalPEPProduction-economic productivity ratiodimensionlessrealized productivityWaterRCMFRelative per-capita water supplydimensionlesssocial water securityWaterRSWSTRelative territorial water supplydimensionlessterritorial hydrological supportWaterRWSRelative water stressdimensionlesswithdrawal pressure Table 1Indicator system used in the generalized desirability model of agricultural landscape NRPResource subsystemIndicator codeIndicatorUnitEcological roleClimaticRSATRelative sum of biologically active temperaturesdimensionlessthermal supplyClimaticRRBRelative radiation balancedimensionlessenergy inputClimaticRSTERelative total evaporationdimensionlessatmospheric demandClimaticRTWCALRelative total water consumptiondimensionlessevapotranspiration demandClimaticRAAPRelative annual precipitationdimensionlessmoisture supplyClimaticRADAHRelative total air humidity deficitdimensionlessatmospheric thirstClimaticRNHCRelative natural humidity coefficientdimensionlessmoisture adequacyClimaticRRDIRelative radiation dryness indexdimensionlessaridity pressureSoilRESFRelative energy of soil formationdimensionlesspedogenic energySoilRESIRelative soil fertility indexdimensionlessagrochemical fertilityLandEPALEnergy productivity coefficientdimensionlessecological land productivityLandBEPALBioenergetic productivity coefficientdimensionlessbiomass energy productivityLandCIBPLBioclimatic productivity coefficientdimensionlessclimatic-biological productivityBiologicalCPPClimatic potential productivity ratiodimensionlessclimatic yield constraintBiologicalMPPMaximum possible productivity ratiodimensionlesssoil-climatic productivityBiologicalRMPPActual maximum possible productivity ratiodimensionlessmoisture-limited productivityBiologicalPEPProduction-economic productivity ratiodimensionlessrealized productivityWaterRCMFRelative per-capita water supplydimensionlesssocial water securityWaterRSWSTRelative territorial water supplydimensionlessterritorial hydrological supportWaterRWSRelative water stressdimensionlesswithdrawal pressure Indicator system used in the generalized desirability model of agricultural landscape NRP Harrington desirability normalization procedure Because the selected environmental indicators are expressed in heterogeneous physical units and differ substantially in numerical scale, direct arithmetic aggregation is methodologically invalid. To ensure comparability, all indicators were first transformed into coded relative variables (SPIᵢ) on a dimensionless ecological response scale, where values approaching 1 indicate environmentally favourable conditions and values approaching 0 represent limiting or unfavourable states (Chen et al., 2012; Harrington, 1965). Depending on the ecological nature of the variable, three normalization logics were used:Direct relative scaling for beneficial indicators, where higher values correspond to greater agroecological favourability;Inverse scaling for pressure indicators, where increasing magnitude reflects environmental limitation;Optimum-response scaling for indicators governed by bilateral ecological tolerance. Direct relative scaling for beneficial indicators, where higher values correspond to greater agroecological favourability; Direct relative scaling for beneficial indicators, where higher values correspond to greater agroecological favourability; Inverse scaling for pressure indicators, where increasing magnitude reflects environmental limitation; Inverse scaling for pressure indicators, where increasing magnitude reflects environmental limitation; Optimum-response scaling for indicators governed by bilateral ecological tolerance. Optimum-response scaling for indicators governed by bilateral ecological tolerance. After coding, each normalized indicator was converted into an individual desirability score using Harrington’s double-exponential desirability transformation:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{i}=\mathrm{e}\mathrm{x}\mathrm{p}\left[-\mathrm{e}\mathrm{x}\mathrm{p}(-SP{I}_{i})\right]$$\end{document}di=exp-exp(-SPIi)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$SP{I}_{i}$$\end{document}SPIi is the coded dimensionless value of indicator \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document}i, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{i}\in [\mathrm{0,1}]$$\end{document}di∈[0,1] is the corresponding individual desirability score. For interpretative and operational purposes within regional land management, the continuous values of individual and generalized desirability scores (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{i}\in [\mathrm{0,1}]$$\end{document}di∈[0,1]) must be categorized into structurally distinct performance intervals. While classical applications of Harrington’s method rely on standard five-grade linguistic psychological anchors, semi-arid agricultural landscapes exhibit highly non-linear responses and extreme sensitivity to initial degradation phases. To preserve model reproducibility while maximizing analytical resolution, a standardized seven-grade hierarchical classification framework was operationalized (Mustafayev et al., 2024b). Rather than relying on subjective or localized expert weightings, the boundary thresholds of this seven-grade scale are strictly anchored to the invariant mathematical features of the probability density function (Gauss) and the cumulative integral distribution function (Laplace/Sigmoid), as visually demonstrated in Fig. 3:[0.00–0.20]: Extreme Deficit Zone (“Very Low” resource potential) – maps onto the baseline asymptotic tail of the distribution, representing severe ecosystem degradation or absolute environmental constraints where resource reproduction is critically restricted.(0.2–0.37]: Transitional Vulnerability Zone – maps onto the region of maximum curvature and peak acceleration (the highest value of the first derivative) of the double-exponential sigmoid function. Within this narrow domain, the ecosystem displays its highest marginal sensitivity, meaning that even minor negative shifts in physical parameters (e.g., moisture or soil quality) trigger rapid systemic decline. To capture these early-stage environmental pressures with high resolution, this zone is mathematically disaggregated into three precise, invariant sub-ranges:(0.20–0.26]: “Low” resource potential;(0.26–0.31]: “Below Average” resource potential;(0.31–0.37]: “Average” resource potential.(0.37–0.63]: Spatial Equilibrium Zone (“Above Average” resource potential) – maps onto the symmetric central region of the normal distribution, indicating a stable, sustainable baseline where resource availability consistently meets standard agroecological requirements.(0.63–0.80]: High Productive Zone (“High” resource potential) – represents stable, optimal hydrothermal and pedogenic balances.(0.80–1.00]: Maximum Abundance Zone (“Very High” resource potential) – corresponds to the upper asymptotic stabilization phase, indicating peak structural and functional ecosystem asset capacity.Fig. 3Harrington desirability transformation of standardised ecological indicators [0.00–0.20]: Extreme Deficit Zone (“Very Low” resource potential) – maps onto the baseline asymptotic tail of the distribution, representing severe ecosystem degradation or absolute environmental constraints where resource reproduction is critically restricted. [0.00–0.20]: Extreme Deficit Zone (“Very Low” resource potential) – maps onto the baseline asymptotic tail of the distribution, representing severe ecosystem degradation or absolute environmental constraints where resource reproduction is critically restricted. (0.2–0.37]: Transitional Vulnerability Zone – maps onto the region of maximum curvature and peak acceleration (the highest value of the first derivative) of the double-exponential sigmoid function. Within this narrow domain, the ecosystem displays its highest marginal sensitivity, meaning that even minor negative shifts in physical parameters (e.g., moisture or soil quality) trigger rapid systemic decline. To capture these early-stage environmental pressures with high resolution, this zone is mathematically disaggregated into three precise, invariant sub-ranges:(0.20–0.26]: “Low” resource potential;(0.26–0.31]: “Below Average” resource potential;(0.31–0.37]: “Average” resource potential. (0.2–0.37]: Transitional Vulnerability Zone – maps onto the region of maximum curvature and peak acceleration (the highest value of the first derivative) of the double-exponential sigmoid function. Within this narrow domain, the ecosystem displays its highest marginal sensitivity, meaning that even minor negative shifts in physical parameters (e.g., moisture or soil quality) trigger rapid systemic decline. To capture these early-stage environmental pressures with high resolution, this zone is mathematically disaggregated into three precise, invariant sub-ranges:(0.20–0.26]: “Low” resource potential;(0.26–0.31]: “Below Average” resource potential;(0.31–0.37]: “Average” resource potential. (0.20–0.26]: “Low” resource potential; (0.20–0.26]: “Low” resource potential; (0.26–0.31]: “Below Average” resource potential; (0.26–0.31]: “Below Average” resource potential; (0.31–0.37]: “Average” resource potential. (0.31–0.37]: “Average” resource potential. (0.37–0.63]: Spatial Equilibrium Zone (“Above Average” resource potential) – maps onto the symmetric central region of the normal distribution, indicating a stable, sustainable baseline where resource availability consistently meets standard agroecological requirements. (0.37–0.63]: Spatial Equilibrium Zone (“Above Average” resource potential) – maps onto the symmetric central region of the normal distribution, indicating a stable, sustainable baseline where resource availability consistently meets standard agroecological requirements. (0.63–0.80]: High Productive Zone (“High” resource potential) – represents stable, optimal hydrothermal and pedogenic balances. (0.63–0.80]: High Productive Zone (“High” resource potential) – represents stable, optimal hydrothermal and pedogenic balances. (0.80–1.00]: Maximum Abundance Zone (“Very High” resource potential) – corresponds to the upper asymptotic stabilization phase, indicating peak structural and functional ecosystem asset capacity. (0.80–1.00]: Maximum Abundance Zone (“Very High” resource potential) – corresponds to the upper asymptotic stabilization phase, indicating peak structural and functional ecosystem asset capacity. Harrington desirability transformation of standardised ecological indicators This mathematically derived configuration directly addresses the methodological concern regarding the uneven width of threshold intervals (e.g., the narrow 0.05–0.06 sub-ranges within the transitional zone versus the wider $0.26$ range in the equilibrium zone). Because the double-exponential transformation curve (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{i}=\mathrm{e}\mathrm{x}\mathrm{p}\left[-\mathrm{e}\mathrm{x}\mathrm{p}(-SP{I}_{i})\right]$$\end{document}di=exp-exp(-SPIi)) is non-linear, a uniform linear grid would fail to register critical tipping points. Thus, the asymmetrical threshold distribution is a deliberate theoretical mechanism to optimize spatial–temporal sensitivity within critical ecological ecotones. Furthermore, this ensures absolute transferability of the model: while absolute physical indicators (such as salinity in dS/m or rainfall in mm) vary substantially across international agricultural landscapes, their projection onto the dimensionless desirability scale always aligns with these rigid, invariant mathematical intervals, enabling standardized cross-regional comparisons. Resource-specific generalized desirability models After individual normalization, desirability scores were aggregated within each environmental subsystem using generalized desirability functions expressed as geometric means. The geometric aggregation was selected because it suppresses compensation among indicators: a critically low value of one component cannot be fully masked by high values of others, which is ecologically consistent with the limiting-factor nature of agricultural landscapes. Climatic resource subsystem The climatic subsystem reflects the combined thermal, radiative, evaporative, and atmospheric moisture conditions controlling primary productivity and evapotranspiration demand. Eight relative indicators were included: relative sum of biologically active temperatures (RSAT), relative radiation balance (RRB), relative air humidity deficit (RADAH), relative total evaporation (RSTE), relative total water consumption of agricultural lands (RTWCAL), relative annual atmospheric precipitation (RAAP), relative natural humidity coefficient (RNHC), and relative radiation dryness index (RRDI) (Boudjabi et al., 2021; Budyko, 1977). The generalized climatic desirability function was calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GD{F}_{CR}=({d}_{RSAT}\cdot {d}_{RRB}\cdot {d}_{RADAH}\cdot {d}_{RSTE}\cdot {d}_{RTWCAL}\cdot {d}_{RAAP}\cdot {d}_{RNHC}\cdot {d}_{RRDI}{)}^{1/8}$$\end{document}GDFCR=(dRSAT·dRRB·dRADAH·dRSTE·dRTWCAL·dRAAP·dRNHC·dRRDI)1/8 Soil resource subsystem The soil subsystem characterizes the pedogenic and agrochemical capacity of the landscape through two complementary indicators: the relative energy of soil formation (RESF), reflecting the energetic completeness of pedogenesis under the prevailing heat-moisture regime, and the relative soil fertility index (RESI), integrating humus reserves, nutrient availability, hydrolytic acidity, and salinization (Volobuev, 1965; Schaetzl & Schwenner, 2006). The generalized soil desirability function was expressed as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GD{F}_{SR}=({d}_{RESF}\cdot {d}_{RESI}{)}^{1/2}$$\end{document}GDFSR=(dRESF·dRESI)1/2 Land resource subsystem The land resource subsystem evaluates the productive ecological efficiency of the territorial land fund through three complementary productivity formulations: energy-based ecological productivity (EPAL), bioenergetic ecological productivity (BEPAL), and climatic-biological productivity (CIBPL) (Zhang et al., 2022; Ramachandra & Negi, 2025). Together these indicators capture the interaction between available radiation energy, soil formation energy, and moisture-regulated vegetation performance. The generalized land desirability function was calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GD{F}_{LR}=({d}_{EPAL}\cdot {d}_{BEPAL}\cdot {d}_{CIBPL}{)}^{1/3}$$\end{document}GDFLR=(dEPAL·dBEPAL·dCIBPL)1/3 Biological resource subsystem The biological subsystem reflects the hierarchical reduction of potential biomass productivity under sequential climatic, soil, and moisture constraints. Four relative productivity ratios were used in the final generalized model: climatic potential productivity (CPP), maximum possible productivity (MPP), actual maximum possible productivity (RMPP), and production-economic productivity (PEP) (Lv et al., 2023; Yang et al., (2021). The generalized biological desirability function was defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GD{F}_{BR}=({d}_{CPP}\cdot {d}_{MPP}\cdot {d}_{RMPP}\cdot {d}_{PEP}{)}^{1/4}$$\end{document}GDFBR=(dCPP·dMPP·dRMPP·dPEP)1/4 Water resource subsystem The water subsystem quantifies hydrological support and anthropogenic withdrawal pressure using three relative indicators: relative per-capita water supply (RCMF), relative territorial water supply (RSWST), and relative water stress (RWS) (Bikbulatov & Stepanova, 2011; Falkenmark, 1986). Together these parameters describe both the absolute renewable water availability and the intensity of its use. The generalized water desirability function was calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GD{F}_{WR}=({d}_{RCMF}\cdot {d}_{RSWST}\cdot {d}_{RWS}{)}^{1/3}$$\end{document}GDFWR=(dRCMF·dRSWST·dRWS)1/3 Integrated natural resource potential model The final natural resource potential of the agricultural landscape was treated as the integrated interaction of the five subsystem generalized desirability functions. The overall generalized desirability function of natural resource potential (GDFNRPL) was therefore obtained as the geometric mean of climatic, soil, land, biological, and water subsystem scores:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$GDFNRPL=(GD{F}_{CR}\cdot GD{F}_{SR}\cdot GD{F}_{LR}\cdot GD{F}_{BR}\cdot GD{F}_{WR}{)}^{1/5}$$\end{document}GDFNRPL=(GDFCR·GDFSR·GDFLR·GDFBR·GDFWR)1/5 The use of geometric aggregation at the final stage ensures that the integrated NRP index remains highly responsive to the lowest-performing subsystem, thereby preserving the ecological logic of non-compensated environmental limitation. For interpretation of the integrated index, an extended seven-grade ecological quality scale was adopted: catastrophic (0–0.17), very poor (0.17–0.27), poor (0.27–0.37), satisfactory (0.37–0.50), average (0.50–0.63), good (0.63–0.83), and very good (0.83–1.00). Temporal validation framework To test the stability and discriminatory performance of the proposed desirability framework, the model was applied to five reference years spanning the full observation interval from 1992 to 2024. The selected years represent contrasting hydroclimatic and socio-economic stages of agricultural development in the study area, including the early post-Soviet agricultural period, the drought-sensitive 2000 s, the low-water decade of the 2010 s, and the recent partially recovering climatic conditions of the early 2020s. Temporal validation was performed by comparing:interannual changes in individual subsystem desirability functions,the relative contribution of limiting resource components,and long-term behaviour of the final GDFNRPL composite. interannual changes in individual subsystem desirability functions, interannual changes in individual subsystem desirability functions, the relative contribution of limiting resource components, the relative contribution of limiting resource components, and long-term behaviour of the final GDFNRPL composite. and long-term behaviour of the final GDFNRPL composite. Such a multi-temporal comparison allows verification that the proposed model is not merely a static indicator synthesis but a dynamically responsive analytical framework capable of detecting environmentally meaningful shifts in agricultural resource potential. Results Climatic resource performance The agro-climatic subsystem produced the highest and most stable desirability values among all five resource components. As shown in Table 2 generalised climatic desirability function (GDFCR) remained within a narrow interval of 0.660–0.681 throughout the observation period, corresponding consistently to the “good” category. This limited amplitude of fluctuation indicates that the thermal and radiative regime of the Ayagoz district forms a persistently favourable energetic basis for agricultural functioning despite episodic precipitation anomalies. Table 2Climatic resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024∑SBCT (°C)2712.52836.52563.72768.32799.3RB (kJ/cm2)145.78149.88140.86147.63148.65TE (mm)750.571214.52688.74833.52791.36TWCAL (mm)583.12599.52563.43590.5594.6AP (mm)325.1238340.3227.6371.6NHC0.430.20.490.270.5RDI1.7932.4271.6562.5941.6RSAT0.9040.9450.8550.9220.933RRB0.9390.9650.9070.9510.957RSTE0.8511.3770.7810.9450.897RRDI0.9390.9650.9070.9510.956DFRNHC0.5220.4410.5420.4660.545DFRR0.8470.9150.8260.9280.817DFRSAT0.6670.6780.6540.6720.675DFRRB0.6760.6830.6680.680.681DFRSTE0.6520.7770.6330.6780.665DFRTWCAL0.6760.6830.6680.680.681GDFCR (generalized)0.6670.6810.660.6710.673 Table 2Climatic resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024∑SBCT (°C)2712.52836.52563.72768.32799.3RB (kJ/cm2)145.78149.88140.86147.63148.65TE (mm)750.571214.52688.74833.52791.36TWCAL (mm)583.12599.52563.43590.5594.6AP (mm)325.1238340.3227.6371.6NHC0.430.20.490.270.5RDI1.7932.4271.6562.5941.6RSAT0.9040.9450.8550.9220.933RRB0.9390.9650.9070.9510.957RSTE0.8511.3770.7810.9450.897RRDI0.9390.9650.9070.9510.956DFRNHC0.5220.4410.5420.4660.545DFRR0.8470.9150.8260.9280.817DFRSAT0.6670.6780.6540.6720.675DFRRB0.6760.6830.6680.680.681DFRSTE0.6520.7770.6330.6780.665DFRTWCAL0.6760.6830.6680.680.681GDFCR (generalized)0.6670.6810.660.6710.673 Climatic resource indicators and their desirability scores, Ayagoz district (1992–2024) The temporal behaviour of individual climatic control variables confirms this interpretation. Although annual precipitation and the radiation dryness index exhibited considerable year-to-year variability (Table 3), the integrated climatic desirability remained buffered because the subsystem combines both energy-rich and moisture-sensitive indicators into a single balanced criterion. Consequently, dry-year anomalies did not fully suppress the climatic resource score, demonstrating the structural robustness of the climatic module within the proposed framework. Table 3Selected environmental forcing indicators across reference yearsYearAP (mm)RDIRESIRWR (mln m3)PP (c/ha)PEP1992325.11.7930.55536884.8617.6520002382.4270.54928387.256.282010340.31.6560.541176.481.9919.592020227.62.5940.52819385.937.862024371.61.60.522183.386.5323.64 Table 3Selected environmental forcing indicators across reference yearsYearAP (mm)RDIRESIRWR (mln m3)PP (c/ha)PEP1992325.11.7930.55536884.8617.6520002382.4270.54928387.256.282010340.31.6560.541176.481.9919.592020227.62.5940.52819385.937.862024371.61.60.522183.386.5323.64 Selected environmental forcing indicators across reference years The internal structural composition of the climatic subsystem is further illustrated by the radar comparison of individual desirability component scores (Fig. 4).Fig. 4Radar comparison of climatic desirability component scores across the five reference years (1992–2024) Radar comparison of climatic desirability component scores across the five reference years (1992–2024) Unlike the generalized climatic desirability function, which remains nearly constant through time, the polygonal configuration of the radar plots reveals moderate redistribution among the six climatic control parameters. In particular, the radiation dryness component (DFRRDI) exhibits the highest temporal amplitude, while the thermal-water balance and radiation supply components remain comparatively stable. Each year displays a step-by-step chart for all six desirability parameters for the selected year, with a threshold value of 0.63 “good” indicated by a dotted ring. Each year indicates its climatic character (for example, 1992—dry year approaching; AP = 325 mm, 2000—through peak; AP = 238 mm, NHC = 0.20, 2010—wettest reference year; AP = 340 mm, 2020—second through; AP = 228 mm, 2024—recovery; AP = 372 mm, highest NHC. Soil and land resource response The soil and land subsystems displayed moderate but comparatively stable desirability values over the 32-year study period. According to Table 4, the generalised soil desirability function (GDFSR) ranged from 0.545 to 0.588, while the land productivity desirability function (GDFRL) Table 5, varied only slightly between 0.560 and 0.578. The close clustering of these values is also evident in the subsystem radar comparison, where both components occupy an intermediate and weakly variable position between climatic abundance and hydrological limitation. Table 4Soil resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024ESF (kJ/cm2)62.76445.87564.67843.69170.077ESF_max (kJ/cm2)91.11293.67688.03692.26692.906RESF0.6890.490.7350.4730.754HA (mg-eq/100 g)7.57.77.988.2SAI1.0040.9550.9090.8860.843HH (t/ha)484.6484.6484.8484.9485.2FH (t/ha)246246246.1246.1246.3HI5.6945.6945.6965.5975.701MNI0.9420.8890.8190.7980.752DSS (%)0.0950.0990.1020.1130.117SSI14.57114.40214.23513.72313.564SFI22.21121.9421.65921.10520.86RESI0.5550.5490.5410.5280.522DFRESF0.6050.5420.6190.5360.625DFRESI0.5630.5610.5590.5540.552GDFSR (gen.)0.5840.5520.5880.5450.587HA hydrolytic acidity, HH humate humus, FH fulvate humus, DSS degree of soil salinisationTable 5Land resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024PP (c/ha)84.8687.2581.9985.9386.53PLPC (c/ha)36.4917.4540.1823.243.27CEPPC0.430.20.490.270.5CEPS0.6890.490.7350.4730.754EPAL0.5440.3130.60.3570.614BEPLP69.6471.667.2970.5271.01BEPNMP29.9514.3232.9719.0435.51BEPAL0.5440.3130.60.3570.614DFEPAL0.6050.5420.6190.5360.625DFBEPAL0.5460.5820.5380.5850.535GDFRL (gen.)0.5750.5610.5770.560.578PP potential productivity, CEPPC coefficient of ecological productivity of vegetation cover, CEPS coefficient of ecological productivity of soil Table 4Soil resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024ESF (kJ/cm2)62.76445.87564.67843.69170.077ESF_max (kJ/cm2)91.11293.67688.03692.26692.906RESF0.6890.490.7350.4730.754HA (mg-eq/100 g)7.57.77.988.2SAI1.0040.9550.9090.8860.843HH (t/ha)484.6484.6484.8484.9485.2FH (t/ha)246246246.1246.1246.3HI5.6945.6945.6965.5975.701MNI0.9420.8890.8190.7980.752DSS (%)0.0950.0990.1020.1130.117SSI14.57114.40214.23513.72313.564SFI22.21121.9421.65921.10520.86RESI0.5550.5490.5410.5280.522DFRESF0.6050.5420.6190.5360.625DFRESI0.5630.5610.5590.5540.552GDFSR (gen.)0.5840.5520.5880.5450.587HA hydrolytic acidity, HH humate humus, FH fulvate humus, DSS degree of soil salinisation Soil resource indicators and their desirability scores, Ayagoz district (1992–2024) HA hydrolytic acidity, HH humate humus, FH fulvate humus, DSS degree of soil salinisation Table 5Land resource indicators and their desirability scores, Ayagoz district (1992–2024)Indicator19922000201020202024PP (c/ha)84.8687.2581.9985.9386.53PLPC (c/ha)36.4917.4540.1823.243.27CEPPC0.430.20.490.270.5CEPS0.6890.490.7350.4730.754EPAL0.5440.3130.60.3570.614BEPLP69.6471.667.2970.5271.01BEPNMP29.9514.3232.9719.0435.51BEPAL0.5440.3130.60.3570.614DFEPAL0.6050.5420.6190.5360.625DFBEPAL0.5460.5820.5380.5850.535GDFRL (gen.)0.5750.5610.5770.560.578PP potential productivity, CEPPC coefficient of ecological productivity of vegetation cover, CEPS coefficient of ecological productivity of soil Land resource indicators and their desirability scores, Ayagoz district (1992–2024) PP potential productivity, CEPPC coefficient of ecological productivity of vegetation cover, CEPS coefficient of ecological productivity of soil This stability is associated with the compensatory behaviour of the underlying indicators. As shown by the temporal trends in RESI and associated productivity variables (Table 3), gradual soil fertility deterioration was partly offset by fluctuations in pedogenic energy efficiency and ecological land productivity. As a result, neither the soil nor land subsystem exhibited abrupt collapses, indicating that these territorial resources function as medium-term stabilisers within the integrated NRP architecture. Biological productivity cascade The biological resource subsystem showed clear cumulative ecological constraints. Although potential productivity (PP) remained relatively stable (81.99–87.25 c/ha), realised production-economic productivity (PEP) varied strongly due to moisture and soil limitations. Table 6 summarises biological indicators and desirability scores across five reference years. Sequential ecological retention coefficients (bioclimatic FT, pedogenic FWQ, and hydrological FW) progressively reduced attainable productivity. Consequently, PEP fluctuated from 6.28 to 23.64 c/ha. The generalized biological desirability function (GDFBR) stayed low (0.478–0.539), indicating stronger cumulative environmental constraints in the biological subsystem compared with climatic, soil, and land resource components. Table 6Biological resource indicators and their desirability scores, Ayagoz districtIndicator19922000201020202024PP (c/ha)84.8687.2581.9985.9386.53FT0.7980.8340.7540.8140.823CPP (c/ha)67.772.7961.8369.9771.24FWQ0.6890.490.7350.4730.754MPP46.6535.6745.4433.0953.72FW0.430.20.490.270.5RMPP20.067.1322.278.9426.86PEP17.656.2819.597.8623.64DFCPP0.6370.6480.6250.6420.645DFMPP0.5620.5150.5630.5060.584DFRMPP0.4540.3980.4670.4060.48DFPEP0.4440.3940.4550.4010.467GDFBR (gen.)0.5180.4780.5230.480.539 Table 6Biological resource indicators and their desirability scores, Ayagoz districtIndicator19922000201020202024PP (c/ha)84.8687.2581.9985.9386.53FT0.7980.8340.7540.8140.823CPP (c/ha)67.772.7961.8369.9771.24FWQ0.6890.490.7350.4730.754MPP46.6535.6745.4433.0953.72FW0.430.20.490.270.5RMPP20.067.1322.278.9426.86PEP17.656.2819.597.8623.64DFCPP0.6370.6480.6250.6420.645DFMPP0.5620.5150.5630.5060.584DFRMPP0.4540.3980.4670.4060.48DFPEP0.4440.3940.4550.4010.467GDFBR (gen.)0.5180.4780.5230.480.539 Biological resource indicators and their desirability scores, Ayagoz district This progressive reduction along the productivity cascade is reflected in low generalised biological desirability values (GDFBR = 0.478–0.539). Figure 5 illustrates the stepwise depletion of biological productivity under cumulative ecological constraints in the Ayagoz district. Starting from a potential climatic productivity of 85.31 c/ha, values decline to 68.71 c/ha after bioclimatic filtering, 42.91 c/ha under pedogenic limitation, and 17.05 c/ha under hydrological restriction. Final realised production-economic productivity reaches only 15.00 c/ha, or 17.6% of the initial climatic potential. The strongest reduction occurs at the hydrological stage, confirming water deficit as the key limiting factor for biological productivity in the district. To empirically validate this cascade, the framework integrates satellite-derived Dry Matter Productivity (DMP) from the Copernicus Land Monitoring Service. Serving as an independent macro-ecological indicator of real-time dry biomass accumulation (kg/ha/day), the DMP spatial fields cross-verify the severe productivity drop observed between the pedogenic and hydrological stages, thereby reinforcing the multi-criteria framework with automated biophysical observation data.Fig. 5Biological productivity depletion cascade from climatic potential to realised production-economic output in Ayagoz district Biological productivity depletion cascade from climatic potential to realised production-economic output in Ayagoz district Water resource limitation as the dominant constraint Among all five resource components, water resources consistently produced the lowest desirability values and the strongest long-term decline. As presented in Table 7, the generalized water desirability function (GDFWR) decreased from 0.457 in 1992 to 0.393 in 2010 and showed only minor recovery thereafter, remaining within the lowest range among all subsystem indicators. Table 7Water resource indicators and their desirability scores, Ayagoz districtIndicator19922000201020202024RWR (mln m3)368.0283.0176.4193.0183.3Population94.45780.90074.43471.48765.429CMF (m3/person/yr)38153498237027002802WIV (mln m3)179.245.08.0417.8319.84WS (%)48.7015.904.569.2410.82SWST (m3/km2/yr)7425.65710.53559.53894.43698.7RCMF0.1910.1750.1180.1350.140RSWST0.0930.0710.0440.0490.046RWS0.4870.1590.0460.0920.108DFRCMF0.4380.4320.4110.4170.419DFRSWST0.4020.3940.3840.3860.385DFSI0.5410.4260.3850.4020.408GDFWR (gen.)0.4570.4170.3930.4010.404RWR renewable water resources, CMF Falkenmark criterion, WIV water intake volume, WS water stress, SWST specific water supply of territory Table 7Water resource indicators and their desirability scores, Ayagoz districtIndicator19922000201020202024RWR (mln m3)368.0283.0176.4193.0183.3Population94.45780.90074.43471.48765.429CMF (m3/person/yr)38153498237027002802WIV (mln m3)179.245.08.0417.8319.84WS (%)48.7015.904.569.2410.82SWST (m3/km2/yr)7425.65710.53559.53894.43698.7RCMF0.1910.1750.1180.1350.140RSWST0.0930.0710.0440.0490.046RWS0.4870.1590.0460.0920.108DFRCMF0.4380.4320.4110.4170.419DFRSWST0.4020.3940.3840.3860.385DFSI0.5410.4260.3850.4020.408GDFWR (gen.)0.4570.4170.3930.4010.404RWR renewable water resources, CMF Falkenmark criterion, WIV water intake volume, WS water stress, SWST specific water supply of territory Water resource indicators and their desirability scores, Ayagoz district RWR renewable water resources, CMF Falkenmark criterion, WIV water intake volume, WS water stress, SWST specific water supply of territory The provided line graph in Fig. 6 illustrates the dynamics of various water resource desirability components in the Ayagoz district from 1992 to 2024. The generalized water desirability function (GDFWR), marked by the thick red line, exhibits a steady long-term decline, decreasing from 0.457 in 1992 to 0.404 in 2024. Among individual indicators, the Water Stress Index (DFSI), represented in green, shows the most significant volatility, plummeting from its peak of 0.541 in 1992 to a low of 0.385 by 2010 before showing marginal recovery. Additionally, the specific water supply of the territory (DFRSWST) consistently occupies the lowest position on the chart, remaining within an unsatisfactory range of 0.384 to 0.402 throughout the entire observation period. This visualization confirms that the cumulative contraction of hydrological components, such as per capita supply (DFRCMF) and territory supply, establishes water resources as the primary limiting factor for the region's overall ecological potential.Fig. 6Dynamics of water resource desirability components in Ayagoz district Dynamics of water resource desirability components in Ayagoz district Integrated natural resource potential dynamics The integrated natural resource potential index (GDFNRPL) remained within the “average” category during all reference years, varying from 0.524 to 0.556 (Table 8). However, the temporal dynamics shown in Fig. 7 reveal a distinct long-term degradation trajectory from 1992 to 2020 followed by only partial recovery in 2024. Table 8Integrated subsystem desirability functions and final GDFNRPLYearGDFCRGDFSRGDFRLGDFBRGDFWRGDFNRPL19920.6670.5840.5750.5180.4570.55620000.6810.5520.5610.4780.4170.53120100.6600.5880.5770.5230.3930.54020200.6710.5450.5600.4800.4010.52420240.6730.5870.5780.5390.4040.547Fig. 7Comparative desirability scores of subsystems Table 8Integrated subsystem desirability functions and final GDFNRPLYearGDFCRGDFSRGDFRLGDFBRGDFWRGDFNRPL19920.6670.5840.5750.5180.4570.55620000.6810.5520.5610.4780.4170.53120100.6600.5880.5770.5230.3930.54020200.6710.5450.5600.4800.4010.52420240.6730.5870.5780.5390.4040.547 Integrated subsystem desirability functions and final GDFNRPL Comparative desirability scores of subsystems This pattern is scientifically important because it demonstrates that the integrated index is not governed by climatic abundance alone. Although the climatic subsystem remained consistently high, the simultaneous depression of biological and especially hydrological desirability values caused the overall NRP to decline. Thus, the integrated index reflects the cumulative balance among resource subsystems rather than the dominance of any single favourable environmental factor. The natural resource potential of agricultural landscapes (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${GDFNRPL}_{i}$$\end{document}GDFNRPLi) constitutes the aggregate of agroclimatic, pedological, land, agrobiological, and water resources within a natural-anthropogenic ecosystem, which influence the productivity of agricultural lands and determine the growth drivers of agricultural production. In this regard, the empirical research database on the yield of rainfed barley agricultural lands — compiled on the basis of data from the statistical compendium “Agriculture, Forestry and Fisheries of the East Kazakhstan Region” of the Department of the Bureau of National Statistics of the Agency for Strategic Planning and Reforms of the Republic of Kazakhstan for the East Kazakhstan Region covering the period 2000–2024 — may be employed to verify the reliability of the quantitative values of the natural resource potential of agricultural landscapes in the Ayagoz District of the Abai Region. In order to convert the absolute yield values of rainfed barley agricultural lands into dimensionless values, the relative yield of agricultural lands (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RPAL}_{i}$$\end{document}RPALi) is used, defined as the ratio of the economic (actual) yield (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${AYAL}_{i}$$\end{document}AYALi) to the theoretically attainable yield (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$AT{PH}_{i}$$\end{document}ATPHi) of agricultural landscapes. On the basis of the data obtained on the natural resource potential of agricultural landscapes (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${GDFNRPL}_{i}$$\end{document}GDFNRPLi) and the relative yield of rainfed barley agricultural lands (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RPAL}_{i}$$\end{document}RPALi) for the period 2000–2024, individual types of equations describing the interrelationships between the dynamic indicators of the modelled process were implemented using Microsoft Excel (Fig. 8).Fig. 8Regression plots of the dependence of rainfed barley agricultural land yield (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RPAL}_{i}$$\end{document}RPALi) on the natural resource potential of agricultural landscapes in the Ayagoz District of the Abai Region Regression plots of the dependence of rainfed barley agricultural land yield (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RPAL}_{i}$$\end{document}RPALi) on the natural resource potential of agricultural landscapes in the Ayagoz District of the Abai Region The initial assumptions employed in the theoretical constructions (Fig. 8) are not subject to any constraints and confirm of a the existence sufficiently high degree of interrelation, with a high correlation coefficient (CCi) of 0.8125 and a coefficient of determination (R2) of 0.6602, which corroborates the sufficiently high accuracy and reliability of the methodological approach for assessing the natural resource potential of agricultural landscapes. Inter-component sensitivity and internal consistency of the model Cross-component comparison confirms a high degree of internal consistency in the proposed modelling framework. In all reference years, subsystem desirability values followed the same hierarchical order: climatic > soil ≈ land > biological > water (Table 8; Fig. 9). Such stability indicates that the model generates structurally coherent outputs rather than random numerical fluctuations.Fig. 9Temporal variation of all subsystems desirability scores (1992–2024) Temporal variation of all subsystems desirability scores (1992–2024) More importantly, the relationship between the hydrological subsystem and the integrated NRP index demonstrates clear limiting-factor sensitivity. As shown in Fig. 6, reductions in water desirability are accompanied by disproportional decreases in the overall GDFNRPL even when climatic desirability remains high. This confirms that the geometric mean aggregation does not behave as a simple arithmetic average and that severe hydrological deficiency cannot be compensated by favourable thermal conditions. The integrated model therefore reproduces the ecological logic of weakest-link control characteristic of semi-arid agricultural systems. Discussion Methodological advances over conventional assessment systems The principal contribution of this study is the development of a continuous, deterministic, and computationally reproducible framework for assessing agricultural landscape natural resource potential (NRP). By transforming heterogeneous agro-climatic, soil, hydrological, biological, and land productivity indicators into a unified Harrington desirability space and aggregating them using geometric means, the proposed approach addresses key limitations of conventional assessment methods. First, widely used approaches such as the Analytic Hierarchy Process (AHP), weighted linear combination (WLC), and fuzzy logic-based GIS models rely on expert-defined weights, which introduce subjectivity and reduce reproducibility across regions and studies (Herzberg et al., 2019; Reed et al., 2023). The Harrington function removes this dependency by providing a fixed monotonic transformation based solely on normalized indicator values, ensuring consistent and weight-independent evaluation. Second, traditional ordinal classification systems are based on discrete class boundaries that reduce sensitivity to gradual environmental change. In contrast, the double-exponential desirability function is continuous and differentiable, allowing smooth and ecologically meaningful responses to incremental changes in environmental conditions. This is particularly important for such variable natural resources as soil, land, and water resources, which undergo continuous change over time. Third, the use of geometric mean aggregation introduces a non-compensatory structure that prevents high-performing subsystems from masking critically limiting ones. This reflects the ecological principle of minimum constraints (Liebig’s law), ensuring that the integrated index (GDFNRPL) remains sensitive to the weakest resource component and thus better represents real-world limiting conditions in agricultural systems. Overall, these properties address a key gap in existing literature by providing a transferable and mathematically consistent framework for integrating diverse environmental indicators without subjective weighting or loss of ecological realism (Bergez et al., 2022; Pollesch & Dale, 2016; Tran et al., 2022). Climatic subsystem: high thermal potential under persistent moisture limitation The agro-climatic subsystem consistently exhibited the highest and most stable generalized desirability values across all reference years (GDFCR = 0.660–0.681), remaining within the “good” quality category. This reflects the strongly continental thermal-radiation regime of the Ayagoz district, where biologically active temperatures exceeded 2500 °C annually and radiation balance conditions were sufficient to support high potential biomass production. These characteristics are typical of the semi-arid steppe zone of eastern Kazakhstan and are comparable to other continental dryland regions of Central Asia and southern Siberia (Basso & Antle, 2024; Budyko, 1977). Despite this high thermal potential, its conversion into biological productivity was substantially constrained by persistent moisture limitation, as indicated by elevated radiation dryness index values (RDI = 1.6–2.6) and low natural humidity coefficients (NHC = 0.20–0.50). Under such conditions, high solar energy inputs combined with insufficient precipitation lead to persistent negative water balance, reducing evapotranspiration efficiency and limiting biomass accumulation (Zhang et al., 2022). This decoupling between thermal potential and realized productivity is reflected in the divergence between climatic (GDFCR) and biological (GDFBR) desirability and is consistent with observations from other dryland agricultural systems, including Mediterranean regions and the Loess Plateau (Lv et al., 2023; Yang et al., 2021). The stability of the climatic desirability index is explained by its multi-indicator structure, which integrates both energy-related (RSAT, RRB) and moisture-related (RNHC, RRDI, RAAP) variables. This balanced design prevents overestimation during dry conditions and underestimation during wetter years, resulting in a highly stable index across the study period (coefficient of variation < 2%). Overall, the eight-indicator geometric aggregation provides a robust and internally consistent representation of climatic resource conditions at interdecadal scale. Soil and land subsystems: latent degradation under apparent stability The soil (GDFSR = 0.545–0.588) and land productivity (GDFLR = 0.560–0.578) subsystems occupied an intermediate and temporally stable position in the hierarchical resource structure. This relative stability masks, however, a pattern of low-intensity but directionally consistent degradation that warrants specific attention in the context of long-term land management. The relative soil fertility index (RESI) declined monotonically from 0.555 in 1992 to 0.522 in 2024, reflecting gradual deterioration in the composite agrochemical status of the landscape’s chestnut steppe soils. The primary drivers of this trend include increasing hydrolytic acidity (from 7.5 to 8.2 meq 100 g−1), declining phosphorus availability (SSI: 22.2 to 20.9), and a slight but progressive increase in the degree of soil salinization (DSS: 0.095 to 0.117%). These changes are consistent with the well-documented trajectory of steppe Kastanozem degradation under low-input rainfed agriculture in semi-arid Kazakhstan, where nutrient depletion and secondary salinization accumulate gradually under conditions of reduced organic matter input and episodic waterlogging in topographic depressions (Kiryushin, 2020; Mustafayev et al., 2024a). Critically, this latent soil degradation is not detectable in conventional annual crop-yield statistics, which respond primarily to interannual precipitation variability rather than to the slow-acting pedochemical processes captured by integrated soil fertility indicators. The desirability-based modelling framework proved particularly effective in making this trend visible, precisely because the RESI indicator contributes continuously to the GDFSR rather than triggering a threshold-class change that would only appear after degradation crosses an expert-defined cut-off. This temporal sensitivity of the desirability model has direct management implications: the declining trajectory of RESI, if uncorrected, is expected to shift the soil subsystem from the present “average” category toward the “satisfactory” lower boundary within two to three decades, which would further depress the integrated GDFNRPL. The land productivity subsystem displayed the strongest internal coupling with the soil resource. The energy-based (EPAL) and bioenergetic (BEPAL) ecological productivity coefficients were constrained in wetter years (1992, 2010, 2024) primarily by moisture availability and in drier years (2000, 2020) by both moisture and the low pedogenic energy efficiency index. The synchronization between DFRESF and DFEPAL across all reference years confirms that the land productivity module correctly captures the coupled energetic response of the landscape to combined soil-forming energy and moisture supply—a coupling that additive multi-indicator approaches commonly underestimate because they assess soil and land productivity as independent scoring dimensions. Biological subsystem: cascade amplification of resource limitation The biological productivity subsystem exhibited the sharpest interannual variability and the clearest mechanistic link between hydroclimatic forcing and realized agronomic output. Despite the potential productivity (PP) remaining relatively stable across reference years (81.99–87.25 c ha−1), the production-economic productivity (PEP) fluctuated threefold (6.28–23.64 c ha−1), illustrating the degree to which sequential resource constraints cascade down the productivity hierarchy. The hierarchical productivity cascade—from PP to CPP (constrained by thermal efficiency factor FT), to MPP (further reduced by soil quality factor FWQ), to RMPP (limited by moisture factor FW), and finally to PEP—represents a Liebig-type resource subtraction chain where each level introduces an additional multiplicative efficiency coefficient bounded by the most limiting concurrent resource. In wet years (1992: AP = 325 mm, NHC = 0.43; 2024: AP = 371 mm, NHC = 0.50), the moisture factor FW remained sufficiently high (0.43–0.50) to allow PEP to reach 17–24 c ha−1. In drought years (2000: AP = 238 mm, NHC = 0.20; 2020: AP = 228 mm, NHC = 0.27), FW dropped to 0.20–0.27, compressing PEP to 6–8 c ha−1 — a reduction that substantially exceeds what would be predicted by precipitation deficits alone, because soil moisture depletion amplifies the initial atmospheric deficit through reduced transpiration capacity and stomatal closure-induced photosynthetic limitation (Lv et al., 2023; Yang et al., 2021). The generalised biological desirability (GDFBR = 0.478–0.539) correctly reproduced this cascade amplification behavior because the geometric mean of four sequential productivity ratios (DFCPP × DFMPP × DFRMPP × DFPEP) ^ (1/4) is highly sensitive to the compounding effect of each successive efficiency loss. The biological subsystem therefore functions as a sensitive integrative mediator: its desirability score reflects not any single environmental constraint in isolation, but the cumulative product of all resource limitations operating simultaneously on agronomic productivity. This behavior validates the use of the biological cascade model as a real-time diagnostic indicator of overall agricultural landscape health — a role that individual climate or soil metrics cannot fulfill. In the territories of the Abai Region classified as arid and semi-arid steppe zones, the ENVISAT MERIS and ENVISAT ASAR spectral bands may be employed for the identification of parameters of agricultural landscapes characterising their biological condition, in order to obtain the broadest possible range of spectral information, particularly in the shortwave infrared region (Dabrowska-Zielinska et al., 2012). In this environmental context, utilizing specific red edge and infrared spectral bands characterized by very narrow bandwidths and low Full Width at Half Maximum (FWHM) values is scientifically critical. Unlike traditional red and near-infrared channels used for broad-band NDVI—which exhibit high sensitivity to soil background noise and saturation—narrow-band red-edge and shortwave infrared (SWIR) spectra capture subtle fluctuations in canopy water content, leaf pigments, and cellular structure. Because water particles strongly absorb radiation in the SWIR spectral region, these bands serve as highly informative carriers of real-time hydroclimatic stress indicators. This is especially vital under the natural and climatic conditions of the Ayagoz District, where the natural moisture coefficient ranges from 0.189 to 0.680 and the yield of rainfed barley agrolandscapes ranges from 6.4 to 40.8 centners per hectare over the period 2000–2024. Water resources as the dominant systemic bottleneck Among all five resource subsystems, water availability consistently showed the lowest generalized desirability values (GDFWR = 0.393–0.457) and the strongest long-term decline, identifying hydrological scarcity as the principal limiting factor of agricultural landscape functioning in the Ayagoz district. This decline is primarily driven by a reduction in renewable surface water resources (RWR) of the Ayagoz River from 368.0 to 183.3 million m3 between 1992 and 2024 (≈50%). This trend is consistent with other semi-arid river systems in eastern Kazakhstan and the Balkhash-Alakol basin, where increasing temperatures enhance evapotranspiration, reduce snow accumulation, and alter spring runoff dynamics (Umirzakov et al., 2025). Accordingly, per-capita water supply decreased from 3815 to 2802 m3 person⁻1 yr⁻1, while territorial supply declined from 7426 to 3699 m3 km⁻2 yr⁻1, placing the region below the Falkenmark threshold of absolute water scarcity. A key result emerges after 2000: despite a sharp reduction in agricultural water withdrawals (WIV from 179.2 to 19.8 million m3), GDFWR improved only marginally (0.393 in 2010 to 0.404 in 2024). This weak response occurs because the desirability framework jointly accounts for both resource availability (RCMF, RSWST) and utilization pressure (RWS) via geometric aggregation. When absolute supply is critically low, reducing withdrawals improves stress indicators but cannot compensate for persistently low availability metrics. In contrast, conventional water stress indices based only on withdrawal-to-availability ratios may incorrectly interpret reduced use as improved water security. The desirability approach avoids this bias by integrating both supply and pressure components, which is important for realistic water management interpretation. The consistently low GDFWR values confirm that the Ayagoz district functions as a hydrologically constrained system, where water availability is the dominant factor limiting agricultural potential despite favorable climatic energy conditions. This aligns with established ecohydrological principles that in dryland systems, water availability—not energy—controls net primary productivity and long-term land carrying capacity (Basso & Antle, 2024; Zhang et al., 2022). Integrated NRP dynamics and inter-subsystem coupling The temporal trajectory of the integrated GDFNRPL (0.524–0.556) demonstrates that the index effectively captures the cumulative balance of resource subsystem performance rather than reflecting the dominance of any single favorable component. Despite GDFCR remaining persistently in the “good” category across all years, the simultaneous depression of GDFBR and GDFWR constrained the integrated index to the lower portion of the “average” category — a result that would not be recoverable from climatic data alone. The hierarchical stability of subsystem rankings (GDFCR > GDFSR ≈ GDFLR > GDFBR > GDFWR in all reference years) confirms the internal structural consistency of the model. Such stability reflects the fact that the proposed framework measures objectively differentiated ecological properties of each subsystem rather than producing arbitrary numerical artefacts. A model generating inconsistent or rank-reversed outputs across reference years would indicate indicator instability or normalization error—neither of which occurred in the present application. The limiting-factor sensitivity of the geometric mean aggregation was further confirmed by correlation analysis between subsystem desirability scores and the integrated GDFNRPL. The water subsystem exhibited the strongest co-variation with the overall index (Pearson r = 0.96 across five reference years), followed by the biological subsystem (r = 0.89), while the climatic subsystem showed the weakest correlation (r = 0.52) — consistent with its role as a stable background resource rather than a dynamic controlling factor. This internal coherence validates the model's capacity to identify the subsystem that most powerfully regulates integrated NRP performance over time, which is directly applicable to prioritization decisions in landscape management and environmental monitoring planning. Transferability, digital compatibility, and operational applicability One of the practical advantages of the proposed framework is its geographic transferability without structural modification. Because the model is built from normalized desirability indicators derived from physically interpretable ecological response curves rather than from region-specific expert scores, it can be recalibrated to other agricultural territories by adjusting only the ecological reference maxima for each indicator while preserving the full mathematical architecture of subsystem aggregation. This feature addresses a recognized limitation of territorial bonitation and AHP-based assessments, whose weighting systems are typically calibrated for specific administrative units and cannot be transposed without expert re-elicitation (Herzberg et al., 2019; Reed et al., 2023). The present framework has already been applied in preliminary form to agricultural landscapes of the Turkestan region of southern Kazakhstan (Mustafayev et al., 2024b, c), and the Ayagoz case study confirms that geometric-mean aggregation of desirability indices behaves comparably across ecologically contrasting settings within the broader semi-arid zone. Furthermore, the strictly algorithmic nature of the desirability calculations renders the framework directly compatible with digital agroecological monitoring platforms, spatial data infrastructures, and automated environmental decision-support systems. All five subsystem desirability functions can be computed from standard hydrometeorological station records, national soil survey databases, and river gauge observations—data types that are routinely collected and increasingly accessible through national monitoring networks. The final GDFNRPL can be automatically recalculated as input data are updated without requiring manual expert interpretation, which satisfies the operational requirements of near-real-time agricultural resource diagnostics described in recent digitalization reviews (Basso & Antle, 2024; Bijlmakers et al., 2022; Cravero et al., 2026). In practical terms, the framework is applicable to at least three operational contexts: (1) annual or seasonal recalculation of integrated NRP indices within regional agricultural monitoring systems to track long-term resource trajectories; (2) spatial comparison of NRP across administrative districts or agro-ecological zones using standardized desirability scores that are independent of the differing physical units of individual indicators; and (3) drought vulnerability diagnostics, in which the desirability model can rapidly identify which resource subsystem is approaching the critical threshold range (< 0.37) that defines environmental limitation under extreme hydroclimatic conditions. Limitations and directions for future research Several limitations of the present study should be acknowledged to appropriately bound the scope of the conclusions. First, the ecological reference maxima used to code individual indicators into the dimensionless SPI scale were calibrated against the hydroclimatic and pedological conditions of the steppe and semi-arid zone of eastern Kazakhstan. Direct application to humid temperate, mountainous, or tropical agricultural systems would require systematic recalibration of these threshold values, ideally using region-specific long-term climatic normals and soil classification parameters. The mathematical architecture of the framework remains unchanged under such recalibration, but the ecological reference boundaries for each indicator must reflect local environmental conditions to preserve the ecological interpretability of the desirability scores. Second, the present analysis was conducted at five reference years spaced across a 32-year interval. While this design allows identification of decadal trends and contrasting hydroclimatic states, annual-resolution modelling would enable more precise attribution of abrupt NRP changes to specific climatic anomalies, policy interventions, or land-use transitions. In particular, the strong interannual variability of the biological subsystem — where PEP fluctuated threefold between drought and wet years — suggests that annual desirability computation would substantially improve the temporal resolution of agronomic risk diagnostics. Third, the framework currently relies on ground-based meteorological, hydrological, and agrochemical observations. The integration of remote sensing products, including spatio-temporal NDVI and GPP assessments derived from satellite data and gross primary productivity estimates obtained using the Leaf Area Index (LAI) and soil moisture (SV) — which are among the Essential Climate Variables (ECV) and Essential Biodiversity Variables (EBV) — would extend the spatial representativeness of the model from the station scale to the entire landscape unit and would provide continuous monitoring in areas where the density of ground-based observations is insufficient (Dabrowska-Zielinska et al., 2012). Such integration is technically feasible given the dimensionless character of the desirability transformation, which is indifferent to the spatial resolution or acquisition technology of the input data, provided that the underlying environmental quantities are physically comparable. Fourth, although the five-subsystem architecture captures the principal resource dimensions of agricultural landscape functioning, future model development could incorporate ecosystem service supply indicators (carbon sequestration potential, biodiversity indices, pollination support capacity) to extend the NRP concept toward a broader landscape multifunctionality index consistent with recent frameworks for sustainable agricultural landscape assessment (Tran et al., 2022; Jansen & Kalas, 2024; Selin-Noren et al., 2026). Harrington desirability transformations are mathematically compatible with any quantifiable indicator, so the extension requires only ecological justification and normalization of the additional variables. Conclusions This study developed and validated a mathematically continuous multi-criteria framework for assessing the natural resource potential (NRP) of agricultural landscapes using Harrington’s generalized desirability function. By converting heterogeneous agro-climatic, soil, land, biological, and hydrological indicators into a unified dimensionless desirability space and integrating them through geometric aggregation, the proposed methodology overcomes the principal limitations of conventional point-based and expert-weighted assessment systems, namely subjectivity, discontinuous class boundaries, and limited reproducibility. Application of the framework to the Ayagoz district over the 1992–2024 period demonstrated that the integrated NRP of the study area remained within the average category (GDFNRPL = 0.524–0.556), while simultaneously revealing strong differentiation among individual resource subsystems. Climatic resources consistently exhibited the highest desirability values, whereas soil, land, and biological resources remained in the satisfactory to moderate range. Water resources showed the lowest and most unstable generalized desirability values, identifying hydrological scarcity as the dominant long-term limiting factor controlling the practical realization of agricultural landscape productivity. The results confirm that favorable thermal-radiation conditions alone do not guarantee high agroecological potential when moisture availability and renewable water supply remain structurally constrained. In semi-arid continental territories, the integrated functioning of agricultural landscapes is therefore governed primarily by the interaction between climatic energy inputs and hydrological limitation, which determines the extent to which biological and land productivity can be realized. Beyond the case study, the proposed desirability-based framework provides a transferable and computationally reproducible platform for regional agroecological diagnostics, long-term environmental monitoring, and digital decision-support systems for sustainable land-use planning. Its algorithmic structure allows adaptation to other semi-arid agricultural regions where multi-parameter environmental observations are available, thereby offering practical value for objective assessment of landscape resilience under growing climate and water uncertainty.