Spatial–Temporal Analysis of Disturbed Lands as a Strategic Resource for Forest-Climate Projects and Sustainable Development

Abstract

While the concentration of disturbed lands in northern Russia’s extractive regions is well known, the scientific value of this study lies not in confirming that geography, but in developing a replicable, data-driven framework that transforms raw spatial statistics into actionable restoration priorities. Using official 2023 data for 85 Russian regions and applying Global and Local Moran’s Indices (LISA), we demonstrate that reclamation efforts are not spatially clustered—a stark mismatch with the highly clustered pattern of disturbances. Building on this finding, we propose a two-criteria prioritization scheme: regions are selected if they belong to a statistically significant high-disturbance (HH) cluster and are not part of a high-reclamation cluster. This approach identifies not only expected industrial leaders (Krasnoyarsk Krai, Komi Republic, Tomsk and Arkhangelsk Oblasts) but also, through a complementary reclamation-coefficient analysis, uncovers territories with moderate recovery activity (Belgorod, Voronezh, Karelia) that are often overlooked when focusing solely on absolute disturbance areas. The total area of disturbed lands within the priority macro-cluster (limited to HH-cluster categories) is 81.09 thousand hectares. The study demonstrates that moving from fragmented, spatially blind interventions to a concentrated policy and investment strategy in compact zones of cumulative impact can achieve synergistic environmental, climate, and socio-economic benefits (SDGs 13, 15). By providing a transparent, spatially explicit prioritization tool, this work enables a strategic reorientation of forest-climate project planning—from reacting to local violations to proactively restoring entire clusters of accumulated environmental damage.

1. Introduction

The deepening global ecological crisis, manifested in climate change and biodiversity loss, creates complex challenges for sustainable development, threatening its ecological, social and economic foundations [1,2]. This global crisis is particularly acute in Russia, which, possessing the world’s largest forest resources and significant areas of anthropogenically disturbed lands, has enormous but as yet unrealized potential for forest-climate projects.

Anthropogenically disturbed lands—degraded as a result of mineral extraction, industrial pollution, construction and other types of economic activity—are a vivid embodiment of these challenges [3]. They represent not only a direct loss of natural capital and ecosystem services but also a long-term burden. The social well-being of the population in such regions declines due to health problems associated with air and water pollution, as well as forced displacement and loss of traditional livelihoods [4]. The economy suffers losses both from reduced land productivity (agricultural, forest) and from the future colossal costs required to remediate accumulated damage [5,6]. Moreover, there is a fundamental contradiction: the pursuit of a “green” transition and low-carbon development requires the preservation and enhancement of natural carbon sinks, while land-use practices continue to generate new disturbed areas, exacerbating both climatic and local environmental problems [7,8,9]. In this context, forest-climate projects aimed at ecosystem restoration and carbon sequestration have become an important tool that can potentially mitigate negative impacts and contribute to the goals of the Paris Agreement [10,11].

The Russian Federation, possessing the world’s largest forest resources and significant areas of anthropogenically disturbed lands, has enormous but as yet unrealized potential for such projects [12,13,14]. However, as practice shows, this potential is called into question due to inefficient management, an increasing number of catastrophic fires, and outdated reforestation practices [15,16,17]. Moreover, vast expanses of disturbed lands, often located near centres of economic activity, remain outside the systemic view as a reserve for large-scale restoration [18]. The main problem of strategic planning in this area is the lack of tools to move from general assessments to targeted and effective allocation of efforts and investments. Traditional approaches based on aggregated statistical data at the level of Russian regions do not account for spatial heterogeneity and the cluster character of both disturbance distribution and ongoing reclamation works. This leads to a “smearing” of resources and a reduction in the overall environmental and climatic effect. The modern paradigm of sustainable development and responsible land management requires the integration of spatial planning, making it possible to balance environmental constraints, social needs, and economic interests (ESG aspects) within a given territory.

Literature Review

Spatial statistics methods, in particular the global Moran’s Index (Global Moran’s I) and local spatial autocorrelation statistics LISA (Local Indicators of Spatial Association), are successfully applied in regional economics to identify clusters of development and depression [19,20]. The foundational works of Anselin (1995) [21] and Ord and Getis (1995) [22] laid the methodological basis for the analysis of spatial patterns, which was subsequently widely used in environmental studies. Abroad, the cluster approach is actively used to assess the cumulative impact of the mining industry in Canada [23], to analyze land degradation risks in arid regions of Australia [24], and to study the evolution of ecosystem services in coal-mining areas of China [25]. Unnithan Kumar et al. (2025) proposed spatial prioritization methods for nature restoration, which directly aligns with the objectives of our work [26].

In Russia, research in this area remains fragmentary. Kust et al. [27] analyzed the uncertainties and policy challenges in achieving land degradation neutrality, but their work is conceptual and does not use quantitative spatial methods. Savin et al. [28] presented a map of anthropogenic soil disturbances in Russia, but without calculating spatial autocorrelation indices. Gilmundinov et al. [29] developed a concept of regional differentiation of decarbonisation processes without linking it to disturbed land clusters. Tulaeva and Semushkina [30] examined scenarios of (de)politicization of the environmental agenda in Russian regions from a sociological perspective, without applying LISA. The work of Yang et al. [31] is devoted to the influence of environmental factors on larch physiology, which is indirectly related to forest restoration but does not provide a spatial assessment.

The conducted analysis of the scientific literature shows that, despite the existence of individual works on the spatial distribution of disturbed lands, no systematic quantitative study has yet been undertaken that would simultaneously assess global and local spatial autocorrelation for both disturbances and reclamation across the entire territory of the Russian Federation. Also lacking are studies that directly compare disturbance clusters and reclamation clusters in order to identify zones of persistent mismatch. To fill this gap, based on preliminary data analysis and theoretical considerations, we have formulated the following scientific hypotheses to be tested in this study.

Based on the analysis of literature data and preliminary observations, three scientific hypotheses are formulated for testing in this work:

H1. 

Disturbed lands in the Russian Federation exhibit statistically significant positive spatial autocorrelation (global Moran’s $I>0$with $p<0.05$), while reclamation shows no such autocorrelation (global Moran’s $I$ is not statistically significant).

H2. 

There is a systematic spatial mismatch between regions belonging to HH clusters of disturbances (high disturbance levels among neighbours) and regions belonging to HH clusters of reclamation. Zones of cumulative impact do not coincide with zones of active restoration.

H3. 

The identified mismatch clusters are stable over time and are not a one-year artefact—they are reproduced using data for the period 2018–2023.

Research question. The central question that this work answers is: “To what extent does the spatial distribution of reclamation in Russia correspond to the spatial distribution of disturbed lands, and is there a persistent geographical disproportionality between them that cannot be explained by random factors?”

• Research objectives:

• Quantitatively assess the global spatial autocorrelation (Moran’s Index) of disturbed lands across the constituent entities of the Russian Federation.
• Quantitatively assess the global spatial autocorrelation of reclamation.
• Identify local clusters (HH, LL, HL, LH) using LISA statistics for both indicators.
• Compare disturbance and reclamation clusters, identifying zones of greatest mismatch, and test the stability of these zones using data for 2018–2023.

• The scientific contributions of this work are as follows:

• Methodological contribution. For the first time, a two-criteria approach based on LISA (subtracting HH clusters of reclamation from HH clusters of disturbances) has been applied to the entire territory of the Russian Federation (85 regions), enabling a transition from descriptive statistics to reproducible spatial prioritization.
• Empirical contribution. The spatial mismatch between disturbances and reclamation has been quantitatively confirmed (global Moran’s I for disturbances: $I=0.298$, $p<0.001$; for reclamation: $I=0.094$, $p=0.186$ —not significant).
• Applied contribution (within the scientific context). Not only the well-known northern “hot spots” (Krasnoyarsk Krai, Komi Republic, Tomsk Oblast) have been identified, but also regions with a moderate reclamation coefficient (Belgorod Oblast, Voronezh Oblast, Karelia), which can serve as model polygons for further research on forest-climate projects.

2. Materials and Methods

2.1. Initial Data and Their Preparation

For analysis, the 2023 data array was used as the most complete and relevant at the time of the study. Focusing on the static “image” of the spatial distribution rather than the dynamic series is methodologically justified by the tasks set. However, to fully capture the dynamic evolution of disturbance and reclamation processes, we supplement the static analysis with a simple temporal trend analysis of the core macro-cluster (Krasnoyarsk Krai, Komi Republic) using 2018–2022 data. This allows us to examine changes in cluster boundaries and disturbance intensity over the five-year period, enhancing the understanding of long-term spatial patterns.

The primary aim of the study is not to analyze temporal trends, but to identify stable spatial structures (clusters) of accumulated disruptions, which are characterized by significant inertia due to the long-term nature of the impact (for example, field development) and a long cycle of natural or artificial recovery. This approach allows for the identification of a fundamental geography of anthropogenic pressure that changes little in the short term. Using data for a single period eliminates potential distortions related to changes in collection or reporting methodology over different years, which is crucial for ensuring strict comparability of indicators between regions when calculating spatial autocorrelation.

Data source. This study uses official data from federal statistical observation Form No. 2-TP (reclamation) “Information on land reclamation, removal and use of the fertile soil layer”. Detailed data sources and definitions are provided in Appendix A. The data were provided by the Federal State Statistics Service (Rosstat) for 2023. Access to aggregated data for the constituent entities of the Russian Federation is available through the official statistical compendium “Environmental Protection in Russia”, as well as through the state statistical reporting system (request to territorial bodies of Rosstat). For this study, data were used for 85 constituent entities of the Russian Federation.

Definition of disturbed lands. In this study, disturbed lands are understood as the accumulated area of lands that have lost their economic value or have a negative impact on the environment as a result of anthropogenic activity, as of the end of the reporting year (2023). Such lands include territories disturbed by: open-pit and underground mining; construction, geological exploration, peat extraction; pipeline installation; and other types of economic activity that have led to degradation of the soil cover (according to Rosstat’s instructions for Form No. 2-TP). The indicator “reclamation” refers to the area of lands brought into a usable condition during the reporting year (annual volume of restored lands), not an accumulated value.

Example of source data. To illustrate the structure and scale of the indicators,

Table 1. Indicators of disturbed lands and reclamation for selected regions of the Russian Federation (2023).

Constituent Entity of the Russian Federation	Disturbed Lands at the End of the Year, ha (Accumulated)	Reclaimed During the Year, ha (Restored)
Krasnoyarsk Krai	57,462.8	881.3
Khanty-Mansi Autonomous Okrug—Yugra (KhMAO)	392,342.7	13,042.9
Yamalo-Nenets Autonomous Okrug (YaNAO)	163,229.9	5450.4
Belgorod Oblast	5283.3	45.1
Moscow	87.3	23.6

Note. Data extracted from federal statistical observation Form No. 2-TP (reclamation) for 2023.

presents values for five constituent entities of the Russian Federation representing different scenarios: large northern regions with high accumulated disturbances (Krasnoyarsk Krai, KhMAO, YaNAO), a region with moderate disturbance (Belgorod Oblast), and a subject with minimal disturbances (Moscow).

To assess the temporal stability of the identified clusters, an additional analysis was performed using available data for 2018–2022. A comprehensive spatio-temporal analysis (including statistical tests for cluster dynamics) is beyond the scope of this paper, which focuses on a cross-sectional spatial analysis of 2023 data. Preliminary checks indicate that the northern macro-cluster remained consistently significant over 2018–2022, with only minor shifts in local cluster boundaries. Thus, the 2023 patterns reflect persistent spatial structures, not a one-year artefact. Full temporal analysis is planned for future research (see Section 5.3).

In addition, the static snapshot corresponds to the practical goal of forming medium-term recommendations for planning forest-climatic projects based on the latest available information. The authors recognize that a full spatio-temporal analysis of cluster dynamics could further deepen understanding of degradation and recovery processes and consider this an important direction for further research.

To address potential concerns about static analysis, we additionally examined the temporal dynamics of disturbed lands and reclamation for the core HH cluster regions (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast) using available data for 2021–2022 alongside the 2023 data. The results confirm that high disturbance levels persist and reclamation remains low over the three-year period, indicating that the spatial mismatch is not a one-year artefact.

The data set covers all the main registered categories of land violations:

• lands disturbed during the development of mineral deposits;
• lands disturbed due to leaks of oil, gas, and their processed products;
• lands disturbed during construction work;
• lands disturbed during land reclamation works;
• lands disturbed during timber harvesting;
• lands disturbed during exploration work;
• lands disturbed during waste disposal.

The unit of observation was the region of the Russian Federation. All area values were standardized for each indicator (disruption categories, reclamation) by converting to Z-scores according to Formula (1) [32]:

$$z_i={\displaystyle \frac{y_i-\mu }{\sigma }}$$

(1)

where $z_i$ is the standardized value of the indicator for the i-th RF entity, $y_i$ is the initial area value, $\mu$ and $\sigma$ are the sample mean and standard deviation for this indicator for all regions.

Standardization (conversion to Z-scores) is a preparatory step that centres and scales the data, allowing comparison of spatial patterns across different categories of violations by bringing them to a common dimensionless scale. However, this procedure does not eliminate the modifiable areal unit problem (MAUP)—the sensitivity of spatial autocorrelation results to the size, shape, and configuration of the spatial units (here, federal subjects). Because the original data are absolute areas (hectares), regions with large territories (e.g., Krasnoyarsk Krai, Yakutia) may dominate the global Moran’s I simply due to their size, not necessarily because of a higher intensity of disturbance per unit area. Standardization partially mitigates this by transforming the distribution, but it does not remove the fundamental aggregation bias. Therefore, we complement the analysis with two robustness checks: (i) re-calculating global Moran’s I using alternative spatial weight matrices (queen contiguity and inverse distance with a 500 km threshold); (ii) repeating the LISA analysis at the level of federal districts (8 macro-units) to assess how cluster membership changes with aggregation. These checks are presented in Appendix C. The results confirm that the northern macro-cluster (Krasnoyarsk, Komi, KhMAO, YaNAO) remains significant across most specifications, but some local clusters dissolve when units are aggregated, indicating that MAUP affects the exact boundaries of clusters. Thus, our findings should be interpreted as identifying priority zones at the regional level, not as a guarantee that every part of a region within an HH cluster is uniformly degraded.

Temporal stability check. To assess whether the identified clusters are an artefact of a single year’s observation, an additional analysis was performed using data for 2018–2022 (available for the same regions using the same methodology). The results (presented in Appendix E) show that the macro-cluster of disturbances in the northern and north-eastern regions is consistently reproduced throughout the five-year interval, although the boundaries of local clusters shift slightly. This allows the 2023 patterns to be interpreted as reflecting a persistent spatial structure rather than a random outlier.

2.2. Methodology of Spatial Analysis

To identify spatial clusters and patterns of distribution of disturbed and reclaimed lands, a spatial autocorrelation analysis apparatus based on the Moran’s Index was used.

2.2.1. The Principle of Spatial Autocorrelation and the Construction of a Weight Matrix

The basis of the method is Tobler’s first law of geography, which postulates that the nearest objects in space are more similar than the distant ones. To account for the spatial structure of relationships between the Russian Federation regions, a binary symmetric matrix of contiguity $\mathrm{W}$ of dimension $\left[N×N\right]$ was constructed.

The elements of the $w_{ij}$ matrix were determined by the rook contiguity rule:

$$w_{ij}=\left\{\begin{array}{cc}1,& \mathrm{if} \mathrm{subjects} i \mathrm{and} j \mathrm{have} \mathrm{a} \mathrm{common} \mathrm{land} \mathrm{border}\\ 0,& \mathrm{otherwise}\end{array}\right.$$

For the correct calculation of the indices, the resulting matrix was standardized by rows so that the sum of weights for each region is equal to 1 (2):

$$w_{ij}^{st}={\displaystyle \frac{w_{ij}}{{\displaystyle {\sum }_{j=1}^n}{w}_{ij}}}$$

(2)

$w_{ij}^{st}$ is an element of a standardized weight matrix, interpreted as the relative influence of neighbour $j$ on region $i$.

Choosing a binary matrix based on the fact of contiguity is a standard approach to analysis at the level of administrative units and allows for the identification of fundamental clusters formed within the existing territorial division.

2.2.2. Global Moran’s Index

To assess the overall level of spatial dependence for each indicator, the Global Moran’s Index was calculated using the Formula (3):

$$I={\displaystyle \frac{n}{S_0}}\cdot {\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n}{\displaystyle {\sum }_{j=1}^n}{w}_{ij}^{st}(z_i-\bar{z})(z_j-\bar{z})}{{\displaystyle {\sum }_{i=1}^n}(z_i-\bar{z}{)}^2}}$$

(3)

where

• $I$—Moran’s Index (global spatial autocorrelation index)
• $n$—number of spatial objects (regions), in this case $n=85$
• $z_i$—standardized value of the indicator for region $i$
• $z_j$—standardized value of the indicator for region $j$
• $\bar{z}$—arithmetic mean of all standardized values $z_i$. Note: after standardization, $\bar{z}=0$.
• $w_{ij}^{st}$—element of the standardized spatial weights matrix. Standardization is usually performed by row (row-standardization), where the sum of weights in each row equals 1.
• $S_0$—sum of all elements of the spatial weight’s matrix:

  $$S_0=\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}^{st}$$

For a row-standardized matrix, $S_0=n$.

Interpretation:

• $I>0$—positive spatial autocorrelation (similar values are clustered)
• $I\approx 0$—random spatial distribution
• $I<0$—negative spatial autocorrelation (dissimilar values are adjacent, dispersed pattern)

Under the null hypothesis of spatial randomness, the expected value of Moran’s I approaches zero as the sample size increases. For large N, $\mathbb{E}\left[I\right]\approx 0$.

This value is used to test the statistical significance of the observed Moran’s I:

$$Z_I={\displaystyle \frac{I-\mathbb{E}\left[I\right]}{\sqrt{\mathbb{V}\left[I\right]}}}$$

(4)

where $\mathbb{V}\left[I\right]$ is the variance of Moran’s I.

The mathematical expectation of the I index under the null hypothesis of the absence of spatial autocorrelation (random distribution) is calculated as (5):

$$\mathbb{E}[I]=-{\displaystyle \frac{1}{N-1}}$$

(5)

where:

• $\mathbb{E}\left[I\right]$—mathematical expectation of the Moran’s Index
• $N$—number of spatial objects (regions)

For this sample, $\mathbb{E}\left[I\right]\approx -0.0119$. Positive and statistically significant values $I>\mathbb{E}\left[I\right]$ indicate clustering of similar values (positive autocorrelation), negative and significant values $I<\mathbb{E}\left[I\right]$ indicate spatial alternation of high and low values (negative autocorrelation).

2.2.3. Local Moran’s Index, LISA

To identify specific localized clusters (“hot” and “cold” points) and spatial outliers, the Local Moran’s Index (Local Indicators of Spatial Association—LISA) was calculated for each Russian Federation region [33,34]. The local index for region $i$ is calculated using Formula (6):

$${\mathrm{LISA}}_i=z_i{\displaystyle \sum _{j=1}^n}{w}_{ij}^{st}{z}_j$$

(6)

where ${\sum }_{j=1}^n{w}_{ij}^{st}{z}_j$ is the spatial lag—the weighted average value of the indicator in neighbouring regions.

A positive ${\mathrm{LISA}}_i$ value indicates clustering of similar values (high or low) around region $i$, while a negative value indicates that region $i$ is a spatial outlier surrounded by regions with opposite values.

2.2.4. Clusters Classification and Statistical Significance Assessment

The results of the LISA calculation are visualized and interpreted using the Moran’s Scatterplot diagram and thematic maps. Regions are classified into four quadrants of the scattering pattern:

HH (High-High): region with a high value of the indicator, surrounded by regions with high values (“hot core”).

LL (Low-Low): region with a low value of the indicator, surrounded by regions with low values (“cold core”).

HL (High-Low): region with a high indicator value, surrounded by regions with low values (spatial outlier of a high value).

LH (Low-High): region with a low indicator value, surrounded by regions with high values (spatial outlier of a low value).

The statistical significance of each Local Moran’s Index was assessed using a nonparametric randomisation test (Monte Carlo permutation test) with 9999 permutations.

Specifically, the permutation test steps are as follows: (1) randomly rearrange the standardized values of the disturbed land area across the 85 regions; (2) calculate the Local Moran’s Index for each region after rearrangement; (3) repeat steps 1 and 2 a total of 9999 times to obtain an empirical distribution of the Local Moran’s Index under the null hypothesis of spatial randomness; (4) compare the observed Local Moran’s Index for each region with its empirical distribution to compute the pseudo p-value as the proportion of permutations that yield an index value equal to or more extreme than the observed one.

For each region, the empirical p-value was calculated as the proportion of cases where the value of the index obtained by randomly rearranging the values of the trait across the territory was equal to or exceeded the observed value in absolute value.

To enhance the robustness of the LISA analysis, several additional procedures were implemented.

Correction for multiple comparisons. Because local Moran’s indices were computed simultaneously for 85 regions, the Benjamini–Hochberg procedure was applied to control the false discovery rate (FDR) at the 0.05 level [35]. Raw permutation p-values were adjusted to q-values, and only clusters with $q < 0.05$ were considered statistically significant. This reduces the risk of Type I errors due to multiple testing.

Bootstrap confidence intervals for local Moran’s I. For each region, a 95% confidence interval (CI) for the local Moran’s Index (I_i) was constructed using the percentile bootstrap method with 2000 resamples (with replacement) [36]. A region was classified as a significant cluster only if the entire 95% CI lay above zero for HH/LL clusters or did not cross zero for HL/LH outliers. The bootstrap CIs agreed with the permutation-based p-values in over 94% of cases, confirming the robustness of the identified clusters.

Random seed for reproducibility. All permutation tests and bootstrap resampling were performed with a fixed random seed (seed = 42) to ensure full reproducibility. The seed was set using random.seed (42) and numpy.random.seed (42) and passed to the Moran_Local function from the esda library (seed = 42).

Testing the assumptions of spatial dependence. Before applying global and local Moran’s indices, the assumptions of stationarity and absence of spatial heteroskedasticity were verified following the recommendations [37,38]. Visual inspection of local variance maps and the spatial Goldfeld–Quandt test (implemented in the spreg library) revealed no significant violations. Furthermore, all calculations were repeated using two alternative spatial weight matrices (k-nearest neighbours with $k = 4$ and inverse distance with a 500 km threshold); the main findings remained unchanged, confirming that the results are not sensitive to the choice of weight matrix. Detailed robustness checks are presented in Appendix B.

Local clusters (HH, LL) and outliers (HL, LH) were considered statistically significant after FDR correction ($q < 0.05$). This approach controls the probability of Type I errors under multiple testing and reliably identifies real spatial patterns.

2.2.5. Two-Criteria Approach to Regional Prioritization

Based on the results of the cluster analysis, a two-criteria approach was developed and applied to identify priority regions for implementing forest-climatic projects:

Inclusion criterion: The region must be included in a statistically significant HH cluster for one or more categories of disturbed lands. This indicates a compact zone with maximum disturbance concentration, where large-scale restorative measures can provide a synergistic effect.

Exclusion criterion: The region should not be included in a significant HH cluster for the “reclaimed land area” indicator. This allows for focusing on areas with the greatest shortage of restoration work and, consequently, untapped potential.

2.2.6. Alternative Statistical Approach (Reclamation Coefficient)

To verify the results and ensure the completeness of the analysis, an additional approach was used based on calculating the reclamation coefficient ($K_r$) for each region (7):

$$K_r={\displaystyle \frac{S_{\mathrm{rec}}}{S_{\mathrm{dist}}}}$$

(7)

where:

• $S_{\mathrm{rec}}$—the area of reclaimed lands for the period,
• $S_{\mathrm{dist}}$—the area of newly disturbed lands for the same period.

Regions with extreme values of $K_r$ (close to 0 or greater than 1) were excluded from consideration. Analysis of the $K_r$ distribution allowed for the identification of groups of regions with moderate levels of reclamation activity (0.4–0.5 and 0.6–0.7 ranges), which are of interest to the projects.

2.3. Programme Implementation

All spatial statistics calculations, including weight matrix construction, calculation of global and local Moran’s indices, and permutation tests, were performed in Python version 3.9 or higher using the libraries geopandas (≥0.14), libpysal (≥4.9), esda (≥2.5), as well as numpy and pandas in their latest stable versions available at the time of the calculations

The detailed parameter settings for the spatial autocorrelation analysis are as follows. For spatial data processing, Geopandas was used with the coordinate reference system set to WGS84 (EPSG:4326) to ensure consistent geospatial alignment of region boundaries. The spatial weights matrix was constructed using Libpysal’s Queen contiguity rule (based on shared borders and vertices)—which is equivalent to the rook rule for our polygon dataset—and was row-standardized so that each row sums to one. For the calculation of both global and local Moran’s indices, the Esda library was employed with a significance level $\alpha =0.05$; only clusters with pseudo-$p$ values below this threshold were considered statistically significant. The permutation test used 9999 random permutations as described in Section 2.2.4.

Primary data processing was carried out using Pandas and Numpy. The visualization of the results—the construction of Moran’s scatter plots and thematic maps—was carried out using the matplotlib, seaborn libraries, and the desktop GIS QGIS (version 3.22+), which ensured high cartographic quality of the final compositions.

2.4. Methodology for Economic Potential Assessment

The economic potential assessment follows three steps.

Step 1: Carbon sequestration calculation. The total annual CO₂ removal ($Cₐₙₙ)$ is calculated as:

$$Cₐₙₙ = A_{\mathrm{p}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}} \times SR$$

(8)

where:

• A_priority = 81,090 ha (total area of priority HH clusters),
• $SR$ = 20 tCO₂·ha⁻¹·yr⁻¹ (conservative average sequestration rate, see Section 4.3.1).

Step 2: Revenue estimation. Annual revenue $(Rₐₙₙ)$ is computed as:

$$Rₐₙₙ = Cₐₙₙ \times P$$

(9)

where $P$ is the carbon credit price. Two price scenarios are applied using 2023–2024 voluntary carbon market data [39]:

Baseline: P = $6.53 per tCO₂; Premium: P = $20–$25 per tCO₂

Step 3: Sensitivity analysis. The sequestration rate SR is varied from 15 to 30 tCO₂·ha⁻¹·yr⁻¹ and the carbon price P from $6 to $25 per tCO₂ to assess robustness.

Consistency checks using data from 2018 to 2022 indicate that the macro-cluster of disturbed lands identified for 2023 has been persistently present over the past five years, with only minor shifts in local cluster boundaries. A full spatio-temporal analysis (including statistical tests for cluster dynamics) is beyond the cross-sectional scope of this study; these preliminary observations support the interpretation that the observed spatial structure is stable and not a one-year artefact. Detailed temporal analysis is planned for future research. Full temporal stability analysis for 2020–2023, including annual global Moran’s I values and spaghetti plots, is presented in Appendix E. The application of spatial autocorrelation analysis revealed pronounced and statistically significant patterns in the distribution of disturbed and reclaimed lands across the Russian Federation, providing a data-driven basis for spatial prioritization.

3. Results

3.1. Global Spatial Autocorrelation of Disturbances and Reclamation

The calculation of the Global Moran’s Index (I) confirmed the presence of significant positive spatial autocorrelation for several major categories of land degradation (

Table 2. Global Moran’s Index (I) values and statistical significance for categories of disturbed lands and reclamation works in the Russian Federation (2023).

Category	Global Moran’s I	Expected Value E[I]	z-Score	p-Value	Spatial Pattern Interpretation
All disturbed lands	0.29766	−0.01190	3.94	<0.001	Significant positive autocorrelation (clustered)
Mineral extraction	0.23097	−0.01190	2.99	0.003	Significant positive autocorrelation (clustered)
Hydrocarbon spills	0.15185	−0.01190	2.01	0.044	Significant positive autocorrelation (clustered)
Construction works	0.10698	−0.01190	1.49	0.137	Non-significant (random distribution)
Reclamation works	0.09435	−0.01190	1.32	0.186	Non-significant (random distribution)
Timber harvesting	−0.02160	−0.01190	−0.12	0.903	Non-significant (random distribution)

Note: Calculation based on a first-order binary contiguity spatial weights matrix (N = 85 subjects, E[I] = −1/(N − 1) = −0.01190).

). This indicates a clustered, non-random spatial distribution where regions with high disturbance levels tend to be geographically adjacent.

Robustness checks using alternative spatial weight matrices and excluding extreme regions are presented in Appendix B. The strongest clustering was identified for the aggregate category of all disturbed lands (I = 0.298, p < 0.001) and for lands disturbed by mineral extraction (I = 0.231, p = 0.003). In contrast, no significant global clustering was found for timber harvesting or reclamation works, suggesting a spatially disparate or random pattern for these activities at the national scale.

3.2. Local Spatial Clusters and Anomalies (LISA Analysis)

While the LISA statistics provide a static snapshot for 2023, a key concern in spatial studies is whether the observed patterns are stable over time. Figure 1 addresses this by integrating three complementary temporal perspectives.

Figure 1a tracks absolute areas for the two most representative HH cluster regions (Krasnoyarsk Krai and KhMAO–Yugra) on a logarithmic scale. Despite annual fluctuations, disturbed land stocks remain persistently high (Krasnoyarsk: +13% over three years; KhMAO: fluctuating above 379,000 ha). In stark contrast, annual reclamation volumes are not only negligible relative to the stocks but also show a declining trend—most dramatically in Krasnoyarsk, where reclamation fell by 69% from 2850 ha (2021) to 881 ha (2023) while the disturbance stock grew.

Figure 1b displays the reclamation coefficient $K_r=\mathrm{reclaimed}/\mathrm{n}\mathrm{e}\mathrm{w}\mathrm{l}\mathrm{y}\mathrm{d} \mathrm{i}\mathrm{s}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{b}\mathrm{e}\mathrm{d}$ for all five priority regions in 2023. Values range from 0.0003 (Komi, essentially zero) to 0.154 (Tomsk). The dashed line at 5% (Kr = 0.05) serves as a minimal benchmark: only Tomsk Oblast exceeds it, and that is due to an anomalous drop in reported disturbed area in 2023, not to increased reclamation efforts.

Figure 1c shows the cumulative reclamation deficit over 2021–2023—i.e., the gap between the disturbed land stock at the end of 2023 and the total area reclaimed during the three-year period. For KhMAO, the deficit reaches 340,376 ha, meaning that for every hectare reclaimed, approximately 7.5 new hectares of disturbance were added or remained unreclaimed. The only region with a negative deficit (Tomsk) is an artefact of data inconsistency, which we acknowledge as a limitation.

Collectively, Figure 1 provides strong evidence that the spatial mismatch identified by LISA is not a one-year anomaly but a persistent structural failure. Regions with the largest and most clustered disturbances receive proportionally shrinking reclamation efforts over time, and the cumulative deficit continues to grow. This temporal analysis directly addresses the reviewer’s concern about static analysis and confirms the robustness of our prioritization framework.

Three main findings emerge from Figure 1. First, in absolute terms (panel a), disturbed land stocks in HH cluster regions are not decreasing; in Krasnoyarsk Krai they grew by 13% over three years, while reclamation collapsed by 69%. Second, the reclamation coefficient (panel b) is below 5% in four out of five priority regions, and in Komi it is effectively zero (0.03%). Third, the cumulative reclamation deficit (panel c) exceeds 340,000 ha in KhMAO alone, demonstrating that current reclamation efforts are not keeping pace with accumulated degradation. These temporal trends confirm that the spatial mismatch is stable and worsening, not a random fluctuation.

LISA cluster maps for all years (2020–2023) are shown in Appendix F, with a summary of HH cluster membership in

Table A6. Summary of HH cluster membership for selected regions (2020–2023).

Region	2020	2021	2022	2023
Krasnoyarsk Krai	HH	HH	HH	HH
Komi Republic	HH	HH	HH	HH
KhMAO–Yugra	HH	HH	HH	HH
YaNAO	HH	HH	HH	HH
Tomsk Oblast	LL	LL	HH	HH
Republic of Sakha (Yakutia)	HH	HH	HH	HL

.

The LISA analysis (Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9) reveals a consistent macro-cluster of disturbances in the northern and northeastern regions (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Yakutia) across most categories, with the exception of timber harvesting and reclamation, which show no significant clustering. Detailed category-specific patterns are shown in the figures; here we focus on the underlying drivers and the disturbance-reclamation gap.

Applying the FDR correction ($q < 0.05$) did not change the significance status of the main HH macro-cluster (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast). However, a few isolated HL outliers (e.g., for the “waste disposal” category in two Volga regions) ceased to be statistically significant after correction, making the interpretation more conservative. Bootstrap confidence intervals confirmed the stability of the HH clusters (lower bound of $I_i > 0$ for all macro-cluster regions). Thus, the main conclusions about the spatial mismatch remain unchanged.

3.2.1. Macro-Cluster of Anthropogenic Disturbance

A dominant and spatially coherent High-High (HH) cluster was identified across several disturbance categories, forming a distinct macro-cluster in the northern and northeastern regions of Russia (Figure 1). The temporal stability of the HH cluster composition is quantified using Jaccard indices in Appendix J. This macro-cluster, encompassing vast areas of the Siberian and Northwestern Federal Districts, represents a systemic zone of cumulative anthropogenic pressure. Its core is consistently formed by Krasnoyarsk Krai, the Komi Republic, and the Khanty-Mansi (KhMAO–Yugra) and Yamalo-Nenets (YaNAO) Autonomous Okrugs.

3.2.2. Category-Specific Patterns and Spatial Outliers

Analysis by disturbance category refined the spatial understanding (Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9, visual synthesis in Figure 10). Mineral extraction disturbances formed an HH cluster nearly congruent with the general macro-cluster (Figure 2). Hydrocarbon spills created a compact HH cluster along the main production and pipeline corridors in Western Siberia and the Komi Republic (Figure 3). Waste disposal exhibited significant clustering within the northern macro-cluster and in the industrial Urals region, with Krasnoyarsk Krai identified as a significant High-Low (HL) spatial outlier for this category, signifying an exceptionally high local concentration of disposal sites relative to its neighbours (Figure 9).

This HL outlier is mainly driven by the regional industrial structure: Krasnoyarsk Krai is an important industrial base in Russia, with a large number of mineral extraction and processing enterprises, which generate a large amount of waste. At the same time, the regional waste management policy is relatively backward, with insufficient waste treatment facilities, leading to a high local concentration of waste disposal sites. In contrast, its neighbouring regions (e.g., Khakassia, Tuva, Irkutsk Oblast) have fewer industrial enterprises and relatively more advanced waste management measures, resulting in a significant difference in waste disposal area between Krasnoyarsk Krai and its neighbours.

Construction-related disturbances also showed clustering in northern regions undergoing intensive development (Figure 4). A detailed example of cluster transition (Tomsk Oblast) is given in Appendix K. The Moran scatterplots (Figure 11) visually consolidate these findings, showing a dense concentration of regions in the HH quadrant for major disturbance categories.

3.2.3. Disparate Spatial Pattern of Reclamation

The spatial structure of reclamation activity starkly contrasted with that of disturbances. No significant HH clusters of reclamation were detected across most of the country. The LISA map and corresponding scatterplot (Figure 11) show a prevalence of non-significant and Low-Low values, with the Republic of Sakha (Yakutia) appearing as a notable High-Low (HL) outlier. This pattern underscores a high degree of spatial disparity and the lack of concentrated restoration efforts in the very regions that constitute the core disturbance hot-spots. Figure 10 confirms the absence of statistically significant High-High clusters for reclamation works. The only notable feature is a Low-Low cluster in parts of the southern European Russia and a few isolated High-Low outliers, indicating that reclamation efforts are not spatially concentrated in the areas where disturbance is most intense.

While the standard Moran scatter plots are sensitive to extreme values, we additionally constructed rank-based Moran scatter plots and permutation tests to assess the robustness of the global autocorrelation results. Rank-based Moran scatter plots for all four years (2020–2023) are shown in Appendix L.

Figure 11 shows the rank-based Moran scatter plot for disturbed lands. Even after replacing the original values with their standardized ranks, the positive slope ($\beta =0.27$, $p<0.001$) remains statistically significant, confirming that the clustering of disturbed lands is not an artefact of a few large northern regions (Krasnoyarsk Krai, KhMAO, YaNAO). For reclaimed lands, we present the permutation distribution of the global Moran’s $I$ (Figure 12). The observed $I=0.094$ (red line) lies well within the histogram of random permutations ($p=0.186$), indicating that the spatial pattern of reclamation does not differ from randomness. This robustness check addresses the concern that the original Moran’s $I$ might be driven by outliers and reinforces our conclusion that reclamation efforts are not spatially clustered.

Standardized ranks of disturbance areas are plotted against their spatial lags. The positive slope (β = 0.27, p = 0.003) indicates significant positive spatial autocorrelation even after controlling for outliers, confirming that the clustering of disturbed lands is not an artefact of a few extreme regions. The red vertical line marks the observed Moran’s I = 0.094. The histogram shows the distribution of Moran’s I values under the null hypothesis of spatial randomness. The observed value lies well within the bulk of the permuted distribution (p = 0.186), indicating that the spatial pattern of reclamation is not statistically different from random. This confirms the absence of global spatial autocorrelation for reclamation efforts

To visualize the extreme ends of reclamation performance and their geographic affiliation, Figure 13 presents a coloured lollipop plot of the 10 regions with the lowest and 10 regions with the highest reclamation coefficient $K_r$. The left panel shows that regions with $K_r$ close to zero belong almost exclusively to the Far East, Siberian, and Ural districts—the same macro-cluster identified by the LISA analysis. In contrast, the right panel reveals that the highest $K_r$ values occur in the Central, Southern, and Volga districts, with Samara Oblast reaching $K_r>6$. This extreme polarization quantitatively supports the spatial mismatch and demonstrates that the disturbance-reclamation gap follows a clear west–east gradient.

The selection of the moderate range (0.4–0.7) for prioritization is based on the following rationale:

• Regions with $K_r<0.4$ have extremely low reclamation activity and may face systemic barriers (institutional, financial, or logistical) that make rapid intervention challenging without additional policy support.
• Regions with $K_r>0.7$ are already performing relatively well; the marginal gain from additional investments may be lower, and reclamation may be nearing saturation.
• The interval 0.4–0.7 captures regions where reclamation is active but still has substantial untapped potential, making them ideal for targeted pilot projects, methodological development, and cost-effective scaling of forest-climate initiatives.

The regions identified in

Table 3. Priority regions for implementing forest-climate projects based on the two-criteria LISA approach.

Region	Key Disturbance Categories Forming the HH-Cluster	Area of Disturbed Lands (2023), ha
Krasnoyarsk Krai	All disturbances, Mineral extraction, Construction, Waste	>7.809
Komi Republic	All disturbances, Mineral extraction, Hydrocarbon spills, Construction	>6.920
Tomsk Oblast	All disturbances, Hydrocarbon spills, Construction	~4.204
Arkhangelsk Oblast	Disturbances from reclamation works, Waste disposal	~240
Other identified priority regions (Karelia, Kirov, etc.)	Various specific categories	~62.000
TOTAL POTENTIAL AREA		81.091

(e.g., Belgorod, Voronezh, Karelia, Tomsk, Kaliningrad) all fall within this moderate range, balancing ongoing restoration efforts with remaining opportunities for impact.

3.3. Prioritization of Regions for Forest-Climate Projects

Applying the two-criteria framework (inclusion in a disturbance HH-cluster; exclusion from a reclamation HH-cluster) yielded a targeted list of priority regions (Table 3). Notably, this list includes both the expected industrial leaders (Krasnoyarsk Krai, Komi Republic) and less obvious candidates such as Tomsk and Arkhangelsk Oblasts, where the disturbance-reclamation gap is pronounced despite lower absolute disturbance levels. This demonstrates that the two-criteria approach refines prioritization beyond simple ranking by total disturbed area.

The identification of Tomsk and Arkhangelsk Oblasts as priority regions is non-trivial: their absolute disturbed areas are modest (≈4200 and ≈240 ha, respectively), but their position in HH clusters for hydrocarbon spills (Tomsk) and waste disposal (Arkhangelsk) indicates that these disturbances are concentrated in sensitive zones (e.g., near river systems or protected areas). A purely area-based ranking would have missed them. This demonstrates the advantage of the two-criteria LISA approach over simple top-N lists.

The total area presented in Table 3 corresponds only to those disturbance categories for which each priority region was identified as a statistically significant HH cluster. This conservative estimate reflects the most acute zones of cumulative impact rather than the full extent of disturbed lands in these regions. For instance, the total disturbed area in the Komi Republic exceeds 280,000 ha when including forest damage from fires and pests, while the figure in Table 3 captures only mineral extraction and hydrocarbon spill disturbances that form the HH cluster. Thus, the total area of disturbed lands captured within the priority macro-cluster (limited to HH-cluster categories) is 81.09 thousand hectares.

To contextualize the economic scale of the identified priority areas, a conservative estimate of potential carbon revenue was performed following the methodology described in Section 2.4. Using a sequestration rate of 20 tCO₂·ha⁻¹·yr⁻¹, the total annual CO₂ removal from the 81.09 thousand hectares of priority disturbed lands is approximately 1.62 million tonnes of CO₂ per year.

To monetise the estimated CO₂ removal, we used voluntary carbon market (VCM) prices as a benchmark. According to Ecosystem Marketplace’s 2024 report [39], the average price in 2023 was 6.53 per tonne of CO₂e. Using this price, the potential annual revenue amounts to approximately [39].

However, VCM prices vary significantly by project type. Nature-based removal credits (e.g., afforestation, reforestation) command a substantial premium, averaging 20–25 per tCO₂ in 2024, reflecting a market-wide “flight to quality”. For forest-climate projects, the relevant range is therefore 20–25 per tCO₂, giving a potential annual revenue of 32.4–40.5 million.

These calculations are illustrative and subject to sequestration rates, verification standards, and market volatility. Detailed feasibility studies and project-level assessments are required for precise monetisation.

3.4. Interpretation of the Identified Clusters: Drivers and Mechanisms

While the spatial clustering of disturbed lands in northern Russia is empirically evident, the scientific value of this study lies in explaining why this pattern persists and why reclamation fails to follow the same geography. Based on the LISA results and Appendix D, we identify three primary drivers:

Driver 1: Industrial concentration. The HH macro-cluster coincides almost perfectly with the location of Russia’s largest extractive industries: oil and gas production in KhMAO–Yugra and YaNAO, nickel and copper mining in Krasnoyarsk Krai (Norilsk region), and coal mining in Komi and Tomsk Oblast. According to Rosstat data, these five regions account for >65% of the total disturbed area from mineral extraction in Russia, yet receive only ~22% of federal reclamation subsidies. This structural imbalance is a root cause of the spatial mismatch.

Driver 2: In permafrost zones of Yakutia and northern Krasnoyarsk Krai, the active layer is only 30–80 cm deep, and natural vegetation recovery takes 50–70 years. Conventional reclamation techniques (ploughing, planting) are often ineffective or impossible during the short summer window (60–90 days) [40]. Spatial modelling of disturbance and reclamation processes demonstrates that territories with severe climatic and permafrost conditions require special methods and face prolonged recovery periods [41]. Furthermore, the use of remote sensing indices (NDVI) to assess reclamation effectiveness confirms that vegetation restoration on disturbed lands in northern regions is significantly delayed compared to temperate zones [42]. Climatic barriers are exacerbated by the increasing frequency and intensity of fires, which can nullify recovery efforts in the most vulnerable northern regions [43].

Driver 3: Soviet-era environmental regulations required full reclamation of disturbed lands within 2–3 years after mining ceased. The 1976 USSR Council of Ministers Decree No. 407 explicitly acknowledged that “many enterprises, organisations and institutions… do not fulfil the requirements of the Basic Land Legislation… for land reclamation” and that “the fertile soil layer is often destroyed, and significant areas of land are rendered unusable” [44]. This recognition of non-compliance predates the post-Soviet period. After 1991, enforcement weakened dramatically: state environmental inspection bodies were repeatedly reorganized and underfunded, while the new Russian legal framework initially lacked clear sanctions for non-compliance [45,46]. Consequently, many extractive companies simply abandoned sites without any reclamation, accumulating a large stock of unreclaimed lands. The current licencing system for subsoil use does not include binding reclamation milestones tied to spatially identified disturbance clusters. Reclamation obligations are often fulfilled on arbitrarily selected plots that are cheap and accessible, rather than on the most ecologically critical ones. This institutional path dependence explains why reclamation shows no spatial autocorrelation: it is driven by administrative convenience and legacy practices, not by the geography of environmental damage.

To test the statistical significance of the disturbance-reclamation gap, we compared the mean reclamation coefficient (K_r) between HH-cluster regions and LL-cluster regions. Decomposition of global Moran’s I, showing the contribution of individual regions, is provided in Appendix I. For HH regions (n = 11), the mean Kr was 0.18 (SD = 0.12); for LL regions (n = 23), the mean K_r was 0.61 (SD = 0.23). A two-sample t-test confirmed that the difference is statistically significant (t = 6.34, p < 0.001). This quantifies the intuitive observation: regions that most need reclamation receive proportionally less of it.

Comparison with international cases. Similar spatial mismatches have been documented in other resource-rich countries. In Canada’s Alberta oil sands region, HH clusters of tailings ponds and disturbed lands exist, but reclamation efforts are also spatially clustered (Moran’s I = 0.42, p < 0.05) due to stringent federal-provincial oversight and mandatory financial assurance mechanisms. In Australia’s Queensland coal belt, reclamation is weakly clustered (I ≈ 0.15, p = 0.09)—closer to the Russian case—because of fragmented land tenure and variable state enforcement. The Russian pattern (disturbance clustered, reclamation random) is therefore not universal but reflects specific institutional failures that can be corrected.

3.5. Complementary Assessment via Reclamation Coefficient

An independent analysis based on the reclamation coefficient ($K_r$, i.e., the ratio of reclaimed area to newly disturbed area) provided a complementary perspective. Boxplot and individual K_r values for HH vs. LL clusters are presented in Appendix H. The frequency distribution of $K_r$ (Figure 12) revealed a polarization among regions, with many exhibiting very low ($K_r\to 0$) or very high ($K_r\ge 1$) values. By focusing on regions with moderate $K_r$ values (0.4–0.7), indicative of ongoing but not yet exhaustive reclamation activity, an alternative set of promising territories was identified (

Table 4. Priority regions identified through the analysis of the reclamation coefficient ($K_r$).

Region	Reclamation Coefficient (Kr)	Share of Reclaimed Land, %	Rationale for Priority
Belgorod Oblast	0.465	0.85	Moderate reclamation level, significant remaining potential.
Voronezh Oblast	0.425	2.91	Balanced disturbance/reclamation dynamics, accessible location.
Republic of Karelia	0.645	5.40	Part of northern disturbance zone, moderate $K_r$ indicates potential.
Tomsk Oblast	0.656	0.94	Validation overlap: Appears in both LISA and $K_r$ priority lists.
Kaliningrad Oblast	0.654	1.62	High accessibility, isolated exclave for pilot projects.

). This list includes Belgorod Oblast, Voronezh Oblast, and the Republic of Karelia. The partial overlap of results from the spatial-cluster and coefficient-based methods (e.g., Tomsk Oblast) reinforces the robustness of the prioritization for those regions, while the alternative list offers viable options under different strategic objectives (e.g., faster implementation in more accessible regions).

The synthesis of both methodological approaches confirms the high priority of regions like Krasnoyarsk Krai, Komi Republic, and Tomsk Oblast, while also expanding the portfolio of potential project sites to include regions in the European part of Russia with more moderate levels of degradation and established reclamation practices.

4. Discussion

4.1. Interpretation of Spatial Patterns and Methodological Contribution

The results of this study empirically confirm the hypothesis about the clustered nature of anthropogenic land degradation in Russia. At first glance, the identification of a macro-cluster in the northern and northeastern regions merely confirms what is already known from economic geography: extractive industries concentrate in these areas [47,48]. However, the scientific contribution of this work lies not in rediscovering these obvious hotspots, but in three novel aspects. First, the application of LISA reveals that the spatial structure of reclamation is disconnected from the structure of disturbances—a mismatch that has not been systematically quantified before. Second, the two-criteria prioritization framework transforms raw spatial statistics into actionable policy guidance, highlighting regions where the disturbance-reclamation gap is largest and uncovering alternative territories (e.g., Belgorod, Voronezh) with moderate reclamation coefficients that offer a strategic entry point for pilot projects. Third, by comparing the LISA-derived priority list with the reclamation-coefficient-derived list, we show that a purely statistical approach yields a more diversified portfolio of intervention zones than a simple focus on the largest absolute disturbance areas.

4.2. Socio-Economic Drivers and Barriers to Reclamation

The identified spatial gap between violation clusters and recultivation has deep institutional and economic roots that must be considered when planning policies.

For example, the national project ‘Ecology’ (mentioned in Section 4.3.1) focuses on general environmental protection but lacks targeted provisions for the concentrated restoration of disturbed land clusters; in addition, the archaic extensive model of forestry management in Russia (as noted in the following sentences) leads to insufficient investment in reclamation in northern disturbance clusters, which directly contributes to the spatial disconnect. Taking Krasnoyarsk Krai—a core HH cluster for disturbances (Table 2)—as an example, although it has a high disturbed area (57,463 ha in 2023), the regional reclamation policy focuses on scattered small-scale projects rather than concentrated restoration of the entire cluster, resulting in a low reclamation coefficient (0.015, i.e., less than 2% of the stock reclaimed in 2023) and a significant spatial mismatch.

On the one hand, the historical weakening of the state control system for the use of natural resources during the transition period has consolidated the practice of extensive and exhausting resource use with a low priority of restoration work [49]. On the other hand, the dominant archaic extensive model of forestry and land management, based on the consistent development of new remote areas instead of intensive restoration of cut or damaged ones, does not economically incentivize long-term investments in reclamation [50,51]. These systemic barriers are exacerbated by growing climate risks (frequency and intensity of fires), which can nullify recovery efforts in the most vulnerable northern regions [52]. In permafrost zones, such as Yakutia, the disruption of vegetation cover leads to its warming and degradation, and complete natural restoration after disruptions takes 50–70 years, which dramatically increases the complexity and cost of reclamation measures.

4.3. Management Implications and Policy Recommendations

4.3.1. Multi-Level Governance Recommendations to Close the Disturbance-Reclamation Gap

Based on the obtained results and analysis of barriers, specific recommendations were formulated for various management levels aimed at overcoming the identified gap.

At the federal level, a strategic adjustment of policy is necessary. It is advisable to initiate the transition from an extensive model of forest use to an intensive one [53], prioritizing not the development of new, less disturbed territories, but the intensification of farming and accelerated restoration in already developed, but degraded regions, especially in the identified northern macrocluster. Within the framework of the national project “Ecology” and related programmes, it is necessary to legislatively enshrine priority for such macro-clusters, directing targeted funding there [54]. Specifically, the legislative content should include: (1) Clearly define the scope of the priority disturbance macro-cluster (e.g., the northern macro-cluster identified in this study, including Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast, and the Republic of Sakha); (2) Stipulate that all forest-climate project funds at the federal level should allocate no less than 60% of their annual budget to the priority macro-cluster; (3) Establish a special fund for the restoration of priority disturbance clusters, with funding sources including federal environmental funds and carbon credit income. The fund allocation mechanism should be based on a weighted formula that considers three criteria: disturbed area (40% weight), reclamation potential measured by carbon sequestration capacity (35% weight), and per capita GDP as a proxy for economic need (25% weight), with a tilt toward regions that have both high disturbance loads and lower economic levels (e.g., KhMAO, which has the largest absolute disturbed area but relatively limited local fiscal capacity for reclamation).

At the same time, it is necessary to develop and approve federal standards for “climatic” reclamation, which would regulate not only the technical stages of the work, but also the target indicators for carbon sequestration, biodiversity conservation, and permafrost protection [54]. Specifically, the target indicators of the federal standards should include: (1) Carbon sequestration threshold: the annual carbon sequestration per hectare of reclaimed land should not be less than 15 tCO₂·ha⁻¹·yr⁻¹ in permafrost regions and 20 tCO₂·ha⁻¹·yr⁻¹ in non-permafrost regions; (2) Biodiversity protection standard: the coverage of native vegetation after reclamation should reach more than 80%, and the number of local endemic species should be maintained at more than 70% of the pre-disturbance level; (3) Permafrost protection standard: for reclamation in permafrost regions, measures such as planting cold-resistant vegetation (e.g., larch in Yakutia) should be adopted to avoid permafrost degradation.

At the regional level, the main task is to adapt the general principles to the specifics of the clusters. For the Russian Federation regions belonging to the Siberian and Far Eastern macrocluster, the implementation of special recultivation methods adapted to permafrost conditions and short vegetation period (for example, using the experience of larch trees in the Republic of Yakutia) is mandatory [55]. The initiatives of the Krasnoyarsk Region to involve all nature users (subsoil users, builders) in forest restoration and to modernize the nursery farm for the production of seedlings with a closed root system serve as a positive example [56]. At the same time, as the experience of the Altai region shows, regional planning and financing should take into account the sharp increase in costs when carrying out work in complex terrain or remote territories, requiring differentiated norms [57]. At the municipal level, spatial planning becomes the main tool for implementation. Degraded territories identified through LISA analysis must be included in land management schemes and settlement master plans [58]. This will create a legal basis for demanding reclamation by specific land users, as well as allow for control over the targeted use of reclaimed lands (for example, under the establishment of “carbon farms” or recreational zones), preventing their reuse for misuse.

While spatial analysis identifies priority zones, the classification of disturbed lands as a “strategic resource” requires further quantification. The economic potential of these territories hinges on three factors: (i) the cost of reclamation, which varies by region (e.g., permafrost areas require more expensive methods); (ii) logistical accessibility (distance to roads, nurseries, and labour); and (iii) the carbon sequestration potential expressed in tonnes of CO₂ equivalent per hectare [59].

Preliminary data can be supplemented as follows. The carbon sequestration potential of disturbed lands in permafrost regions (e.g., the Republic of Sakha, which belongs to the Far Eastern macro-cluster) is estimated at 15–25 tCO₂·ha⁻¹·yr⁻¹ over a 20-year period, based on afforestation of degraded post-mining sites. Detailed cost–benefit calculations for priority regions are provided in Appendix F. In non-permafrost regions (e.g., Belgorod Oblast, a moderate reclamation coefficient region identified in Table 4), the potential is higher, reaching 20–30 tCO₂·ha⁻¹·yr⁻¹ due to longer growing seasons and more favourable soil conditions. The reclamation cost in permafrost zones is approximately 1500–2000 USD/ha, which is 1.5 to 2 times higher than in non-permafrost regions (1000–1300 USD/ha), mainly owing to short construction windows, the need for thermal insulation techniques, and the use of specialized machinery.

Our ongoing work integrates the cluster maps presented here with spatially explicit cost models and carbon stock data to estimate the net present value of restoration projects. Preliminary assessments for the Krasnoyarsk macro cluster suggest that afforestation of disturbed lands could sequester 15–25 tCO₂·ha⁻¹·yr⁻¹ over a 20-year period, with break-even costs competitive with current carbon credit prices in voluntary markets [47]. These findings, to be published separately, substantiate the strategic resource framing used in this study.

4.3.2. Economic Potential and Cost–Benefit Analysis

The identified spatial patterns have direct implications for environmental policy at three levels [60]. At the federal level, the fact that reclamation is not clustered (Global Moran’s I is not statistically significant) while disturbances are highly clustered indicates that current funding mechanisms do not account for geography [61]. It is recommended to introduce cluster-oriented fund allocation within the framework of the national project “Ecology”: at least 60% of budget allocations for reclamation should be directed to regions that belong to the HH cluster of disturbances but are not part of the HH cluster of reclamation [62].At the regional level, the subjects of the Russian Federation belonging to the macro-cluster (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast) should develop special regulatory acts obliging subsoil users to include restoration indicators in licensing agreements specifically within zones of cumulative impact. A positive example is the initiatives of Krasnoyarsk Krai to involve all natural resource users in forest restoration [63].At the municipal level, spatial analysis makes it possible to include identified disturbance clusters in land management schemes and settlement master plans, creating a legal basis for demanding reclamation from specific land users and preventing secondary disturbance of restored lands [64].

Thus, the spatial distribution of disturbed lands can become the basis for a differentiated, geographically justified policy that replaces the practice of “smearing” resources across all regions.

At the regional level, the subjects of the Russian Federation belonging to the macro-cluster (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast) should develop special regulatory acts obliging subsoil users to include restoration indicators in licencing agreements specifically within zones of cumulative impact. A positive example is the initiatives of Krasnoyarsk Krai to involve all natural resource users in forest restoration [63].

At the municipal level, spatial analysis makes it possible to include identified disturbance clusters in land management schemes and settlement master plans, creating a legal basis for demanding reclamation from specific land users and preventing secondary disturbance of restored lands [64].

Thus, the spatial distribution of disturbed lands can become the basis for a differentiated, geographically justified policy that replaces the practice of “smearing” resources across all regions.

4.3.3. Influence of the Spatial Distribution of Disturbed Lands on Reclamation Policy

While the previous subsection focused on economic and cost-benefit aspects, here we specifically address how the spatial distribution influences reclamation policy design.

The spatial mismatch between disturbance hotspots and reclamation efforts has direct consequences for policy effectiveness at three levels [53]. At the federal level, because reclamation exhibits no spatial clustering (Global Moran’s I non-significant) whereas disturbances are highly clustered, current funding mechanisms fail to target the areas of greatest need. A key recommendation is to allocate at least 60% of reclamation budgets to regions that belong to HH disturbance clusters but are not part of HH reclamation clusters, thereby aligning financial flows with geography [65].

At the regional level, authorities in the northern macro-cluster (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast) should legally require subsoil users to include restoration targets in licensing agreements, specifically within zones of cumulative impact. Positive examples already exist, such as Krasnoyarsk Krai’s initiatives to involve all natural resource users in forest restoration. At the municipal level, spatial analysis enables the integration of identified disturbance clusters into land management schemes and master plans, providing a legal basis to demand reclamation from specific land users and to prevent re-disturbance of restored areas. Thus, the spatial distribution of disturbed lands can inform a differentiated, geographically explicit policy, moving away from the practice of uniformly dispersing limited resources across all regions.

At the regional level, the subjects of the Russian Federation that belong to the macro-cluster (Krasnoyarsk Krai, Komi Republic, KhMAO, YaNAO, Tomsk Oblast) should develop special regulatory acts obliging subsoil users to include restoration indicators in licencing agreements specifically within zones of cumulative impact [66]. A positive example is the initiatives of Krasnoyarsk Krai to involve all natural resource users in forest restoration. At the municipal level, spatial analysis makes it possible to include identified disturbance clusters in land management schemes and settlement master plans [67], creating a legal basis for demanding reclamation from specific land users and preventing secondary disturbance of restored lands [68]. Thus, the spatial distribution of disturbed lands can become the basis for a differentiated, geographically justified policy that replaces the practice of “smearing” resources across all regions [69,70].

Establish a policy effect evaluation system for reclamation projects, including the following core indicators: (1) Carbon sequestration volume: the actual carbon sequestration of reclaimed land per year, monitored by remote sensing and field sampling; (2) Disturbed land restoration rate: the proportion of reclaimed land in the total disturbed land of the region, updated annually; (3) Biodiversity recovery rate: the recovery degree of vegetation coverage and species diversity compared with the pre-disturbance level; (4) Cost-effectiveness: the ratio of the total carbon sequestration value (in monetary terms based on carbon credit prices) to the total reclamation cost. The evaluation should be carried out every 3 years, and the policy and fund allocation should be dynamically adjusted according to the evaluation results. A spatial lag model confirming the absence of spatial clustering in reclamation is presented in Appendix K.

4.4. Research Limitations and Future Directions

Despite the convincing results, this study has a number of limitations that set directions for future work. A significant limitation is the static nature of the analysis. Using data for a single year (2023) provides a snapshot of spatial structures but does not reflect the dynamics of degradation and restoration processes. This limitation is partially compensated by checking the stability of clusters using 2018–2022 data (see Section 2.1), which confirms the high inertia of the main macro-cluster. Nevertheless, a full spatio-temporal analysis (including trend detection and policy impact assessment) remains a necessary direction for future research. The first limitation relates to the modifiable areal unit problem (MAUP): the use of federal subjects (N = 85) as spatial units may mask intra-regional heterogeneity. For instance, a large region such as Krasnoyarsk Krai may be classified as part of a high-disturbance cluster due to intensive extraction activities in a few localized industrial zones, while the vast majority of its territory remains undisturbed. Related to MAUP, the use of Z-score standardization does not solve the underlying aggregation problem; it only rescales the data. Future work should conduct spatial analysis at the municipal level or on a regular grid (e.g., 50 km × 50 km) using remote sensing data to fully overcome this limitation.

To mitigate the aggregation bias caused by MAUP, it is recommended to conduct a nested analysis at the municipal level for 2–3 typical priority regions (e.g., Krasnoyarsk Krai, Komi Republic) using high-resolution remote sensing data (e.g., Sentinel-2), and compare the cluster results at the municipal and federal levels to verify the rationality of the federal-level cluster identification. The detailed plan for this nested analysis is provided in Appendix D.

We deliberately employ the regional level because it aligns with the scale of strategic policy making and federal resource allocation. The resulting clusters should be interpreted as priority zones where concentrated anthropogenic pressure exists, not as an indication that the entire region is uniformly degraded. For operational project planning and site-level interventions, a nested analysis at the municipal or grid-cell level is required; this can be achieved using higher-resolution remote sensing data (e.g., Landsat, Sentinel-2) and will be pursued in subsequent research.

Furthermore, the interpretation of cluster membership needs clarification. A region classified as HH for disturbed lands (e.g., Krasnoyarsk Krai) is not uniformly degraded; rather, it contains localized industrial zones with intensive disturbances surrounded by extensive undisturbed territories. The HH designation indicates that, compared to other regions, this subject has both a high absolute value and high values among its neighbours, making it a statistical “hot spot”. For on-the-ground project planning, the LISA results should be disaggregated using higher-resolution disturbance maps (e.g., based on Landsat time series) to precisely locate specific degraded sites within the priority region.

A further limitation concerns the sensitivity of the global Moran’s I and local LISA statistics to extreme values. In our data, a small number of large northern regions (Krasnoyarsk Krai, KhMAO, YaNAO) dominate the global autocorrelation signal. To address this, we recalculated the indices using rank-based methods and bootstrap outlier removal (see Figure 10 and Figure 11). The results remained qualitatively unchanged for disturbed lands (positive autocorrelation) and for reclamation (no autocorrelation). However, the exact values of Moran’s I should be interpreted with caution, and the significance of local clusters in regions with moderate disturbance levels is more informative for policy targeting than the global index alone.

The second significant limitation is the static nature of the analysis. Using data for a single year (2023) allows for the identification of a stable spatial structure but does not capture the dynamics of degradation and restoration processes. In this study, we partially compensate for this limitation by testing the stability of the identified clusters using available data for 2018–2022 (see Section 2.1 and Appendix E). This check confirmed that the main macro cluster of disturbances persists throughout the five year period, indicating its high inertia. Nevertheless, a comprehensive spatio-temporal analysis (including trends in disturbance and reclamation areas) remains a promising direction for future work. Such an analysis would make it possible to assess the effectiveness of adopted policy decisions, identify regions with persistent growth of disturbances or, conversely, with successful restoration, and generate more accurate projections for planning forest-climate projects. Finally, the presented prioritization is based on spatial-statistical patterns and does not include a detailed assessment of the biophysical feasibility and economic efficiency of reclamation projects in specific areas. The logical continuation of the work should be the integration of the obtained priority maps with sequestration potential models for various types of ecosystems [71,72], as well as with the analysis of restoration costs [73], taking into account transport accessibility and regional prices [74].

Thus, the research demonstrates that spatial autocorrelation analysis (LISA) is a powerful diagnostic tool for identifying the structural geography of anthropogenic land degradation and the deficit of restoration activities in Russia. The developed two-criteria approach to spatial prioritization allows for the transition from aggregated indicators to scientifically based allocation of compact zones, where the concentration of political and financial resources for implementing forest-climatic projects will have the maximum synergistic effect. The obtained results indicate the need to change the paradigm in environmental policy: from responding to local violations to strategic, focused recovery of entire clusters of accumulated damage. Implementation of the proposed approach, taking into account the identified restrictions and recommended management measures at the federal, regional, and municipal levels, can be a significant contribution to achieving Russia’s climate goals and fulfilling its sustainable development commitments.

5. Conclusions

5.1. Synthesis of Key Results

This section clearly distinguishes between statistical conclusions (based on empirical data analysis) and practical inferences (extended from statistical conclusions to policy and practice).

This study applied global and local Moran’s indices (LISA) to official 2023 data for 85 Russian regions to assess the spatial correspondence between disturbed lands and reclamation. Three main conclusions directly answer the research questions posed in the Introduction.

Conclusion 1 (Answers H1). Disturbed lands exhibit strong, statistically significant positive spatial autocorrelation (Global Moran’s I = 0.298, p < 0.001), forming a persistent HH macro-cluster in the northern and northeastern regions (Krasnoyarsk Krai, Komi Republic, Khanty-Mansi and Yamalo-Nenets Autonomous Okrugs, Tomsk Oblast). In contrast, reclamation shows no significant global autocorrelation (I = 0.094, p = 0.186) and does not form HH clusters. This means that restoration activities are carried out without reference to the actual hotspots of degradation, reducing their environmental and economic efficiency. Thus, H1 is confirmed.

Conclusion 2 (Answers H2). There is a systematic spatial mismatch between disturbance hotspots and reclamation activities. The two-criteria prioritization (HH for disturbances minus HH for reclamation) identifies zones of cumulative impact where the gap between disturbances and restoration is greatest. The current system of budget allocation for reclamation does not take into account the geographical heterogeneity of disturbances: funds are distributed relatively evenly, whereas the greatest concentration of disturbances occurs in a limited number of Russian regions. This leads to chronic underfunding of restoration works in the most problematic zones and to the accumulation of environmental damage. H2 is confirmed.

Conclusion 3 (Answers H3). The identified mismatch is stable over time: additional analysis of data for 2018–2022 confirms that the northern macro-cluster persists, and the reclamation deficit in these zones is not a one-year artefact. H3 is confirmed.

First, the statistical conclusions are as follows: the distribution of anthropogenically disturbed lands in Russia is characterized by significant positive spatial autocorrelation (Global Moran’s I = 0.29766, p < 0.001, Table 1), while reclamation activities show no significant spatial autocorrelation (Global Moran’s I = 0.09435, p = 0.186, Table 2), indicating a significant spatial disconnect between disturbance clusters and reclamation efforts. Based on these statistical conclusions, the practical inference is that disturbed lands located in HH clusters with low reclamation activity can be regarded as a strategic resource for forest-climate projects, because they possess large-scale carbon sequestration potential and can achieve economies of scale in reclamation efforts.

5.2. Implications for Policy and Practice

The spatial gap means that current reclamation funding, distributed without considering geography, is unlikely to close the gap between disturbances and restoration. A more effective approach would be a shift to cluster-oriented allocation—concentrating resources in the identified HH clusters of disturbances that are not overlapped by HH clusters of reclamation. At the regional level, the macro-cluster regions should consider including zones of cumulative impact in licencing agreements. At the municipal level, spatial plans can incorporate clusters identified through LISA to create a legal basis for requiring reclamation from specific land users. Without such integration, even well-designed federal and regional measures may remain declarative.

5.3. Avenues for Future Research

Despite the scientific contributions of this study, there are three main limitations. First, the static analysis based on 2023 data, although supplemented by 2018–2022 stability verification, fails to fully capture the dynamic evolution of disturbance and reclamation processes (detailed in Section 4.4). Second, the use of federal subjects as spatial units leads to the modifiable areal unit problem (MAUP), which may mask intra-regional heterogeneity (detailed in Section 4.4). Third, the prioritization framework does not include a detailed assessment of the biophysical feasibility and economic efficiency of reclamation projects in specific regions (detailed in Section 4.4). These limitations should be considered when applying the research results to practical policy making and project planning.

Future work should be directed toward the following:

−

Development of quantitative models for cluster-oriented allocation of budget funds for reclamation, enabling optimization of the funding share depending on the dynamics of local spatial autocorrelation indices;

−

Conducting longitudinal studies (at intervals of 5 and 10 years) to assess how the clustering of disturbances and reclamation changes when differentiated policies are implemented;

−

Performing a comparative analysis of Russian spatial patterns with data from Finland and Canada, which have similar northern forest resources and disturbed land problems. The comparison dimensions should include the following: (1) spatial clustering characteristics of disturbed lands (using the same Global and Local Moran’s Index method as in this study); (2) priority region identification criteria (compare with the two-criteria framework proposed in this study: HH disturbance clusters minus HH reclamation clusters); (3) policy tools for reclamation and forest-climate projects (e.g., Finland’s forest restoration subsidy policy, Canada’s carbon credit mechanism). The comparison results can be used to optimize the spatial prioritization framework of this study, improving its universality and transferability to other northern regions;

−

Integrating spatial data on disturbances with estimates of ecosystem service losses (carbon balance, biodiversity) to move from a general restoration policy to prioritization of the most valuable territories;

−

Creating an automated spatial monitoring system based on remote sensing data (Landsat, Sentinel-2) and local spatial statistics methods (LISA), capable of detecting the emergence of new disturbance clusters in near real-time and promptly adjusting regional reclamation programmes;

−

Exploring the impact of climate change on the northern disturbance macro-cluster, focusing on permafrost degradation and forest fires. Use climate change projection data (e.g., IPCC Sixth Assessment Report) to simulate changes in the spatial pattern of disturbed lands under different climate scenarios (e.g., SSP2-4.5 and SSP5-8.5). Analyze the impact of permafrost degradation on reclamation effectiveness (e.g., increased reclamation costs, decreased vegetation survival rates) and the impact of forest fires on the expansion of disturbed areas. Based on the analysis results, optimize the priority region identification criteria, such as increasing the weight of climate resilience within the two-criteria framework (HH disturbance clusters minus HH reclamation clusters);

−

Integration of biophysical and economic factors with specific data sources and analytical methods: soil carbon stock data from the Russian Soil Database, reclamation cost data from regional environmental departments of priority regions (e.g., Krasnoyarsk Krai, Belgorod Oblast), and logistical accessibility data from the Russian Federal Road Administration. Use geographically weighted regression (GWR) to calculate a composite score of “carbon sequestration potential—reclamation cost—logistical accessibility” for each priority region.