Shifts in River Flood Patterns in the Baltic States Between Two Climate Normals

Abstract

River spring and flash floods are highly dependent on variations in meteorological conditions. In the Baltic States, substantial changes in air temperature and precipitation have been observed between the two most recent climate normal periods (1961–1990 and 1991–2020). Therefore, changes in the magnitude of spring and flash floods across different hydrological regions between these periods were analyzed to better understand shifting hydrological patterns. Daily flow data from 1961 to 2020 were obtained from 68 water gauging stations on 55 rivers. The Pettitt and Mann–Kendall tests, as well as Sen’s slope estimator, were applied to analyze the time series of flood maximum discharges. The most pronounced negative trends in spring and flash floods were observed in Lithuanian rivers, with the magnitude of these trends gradually weakening toward Latvia and Estonia. The maximum flood heights (h_MAX) generally declined during 1961–2020, particularly in Lithuania and western Latvia. Spring flood data showed the most significant decrease, particularly during 1991–2020, when h_MAX declined on average by 0.14 mm/year in Lithuania and 0.05 mm/year in Latvia. Flash floods exhibited smaller declines, also concentrated in 1991–2020. In the major rivers (Nemunas, Neris, and Daugava), peak discharges of both floods declined consistently throughout the study period.

1. Introduction

Floods are natural phenomena that cannot be entirely prevented [1]. Human society has a complex yet essential relationship with rivers and their floods [2]. River floodplains, valued for their scenic beauty and fertile soils, have long attracted people to live and farm there. As long as river floods remain within certain thresholds and do not disrupt human activities, they are viewed favorably and regarded as important ecosystem service providers, a part of the river’s natural variability necessary for healthy riverine and floodplain habitats. Recent research suggests that river floodplains provide multiple ecosystem services—including nutrient cycling, groundwater recharge, and biodiversity support—that might be highly undervalued [3,4]. However, when floods become unpredictable, inundate large territories, and exceed their typical limits, they are classified as natural disasters. Flooding has historically been one of the primary natural causes of fatalities and infrastructure damage throughout recorded history. Each year, millions of people are affected by devastating floods, facing significant losses and challenges [5,6,7]. Over the past century, floods have been the second-deadliest natural disaster, surpassed only by droughts [8]. Thus, river flooding represents a complex phenomenon, encompassing both benefits and risks.

It is common to view the increasing frequency of extreme hydrological events as an indicator of climate change. Since river floods typically result from heavy rainfall, rapid snowmelt, rain-on-snow melt, or ice jam events, it is clear that precipitation and air temperature changes significantly influence the potential severity of floods. However, extensive floods do not occur each year, and when they do, they may not affect all river systems uniformly. While numerous studies have been conducted on the impact of a changing climate on extreme hydrological events [7,9,10], and the role of climatic factors on the formation of major floods is undeniable, doubts remain that climate change is the sole cause of the growing flood-related losses. The effect of climate change on flood hazards is multifaceted and depends on the specific mechanisms driving river flood events [11]. River floods triggered by rainfall and snowmelt are controlled by processes in three key compartments [12]. The first is the atmosphere, where rainfall is produced, and energy conditions for snowmelt and evaporation are set. The second is the catchment area, including land surfaces, soils, and aquifers, where water runs off or infiltrates the ground. The third is the river system, where runoff is transported downstream, accumulating water from multiple catchments. Changes in any of these compartments can significantly influence flood characteristics. The most direct and noticeable impact of human activities on flood formation occurs specifically at the catchment level. There is a growing recognition that irresponsible human activities along riverbanks significantly contribute to flood losses [13]. The invasion of floodplains and the increasing exposure of assets at risk distort an objective assessment of flood magnitude dynamics [14,15].

The interaction of multiple previously discussed flood drivers determines the type of river flood. Although many flood studies focus on annual maximum flows, they often do not distinguish the types of floods characterized by these flow peaks. Meanwhile, categorizing the annual maximum flow data into specific flood types with distinct features (such as origin, seasonality, duration, intensity, etc.) provides an opportunity to analyze these extreme flow data more thoroughly. The author [14] suggested that one key distinction between flood types is determined by the size of the affected area and the duration of the precipitation event that triggers the flood. As a result, two main categories emerged: extensive, long-lasting floods and local, sudden floods. Based on the causative mechanisms, [16] proposed the following flood types: long-rain floods, short-rain floods, flash floods, rain-on-snow floods, and snowmelt floods. Later, some of these suggested flood types were merged. In this way, snow-related flood events (snowmelt floods and rain-on-snow floods) formed one category termed ‘snow-related floods’, and those of synoptic origin (long rain and short rain floods) were categorized as ‘synoptic floods’. In contrast, the flash flood type remained as it is [17]. These are just a few studies exploring approaches to classifying flood events. To date, no single unified classification system exists [18].

Extensive research reveals that severe floods are a major concern, and flood patterns and intensities vary widely across the world regions [9,19,20,21,22]. In recent decades, the impact of climate change on river flooding has become increasingly evident in the Baltic States—Estonia, Latvia, and Lithuania [23,24]. These countries, which have highly seasonal climatic conditions, are now experiencing noticeable shifts in flood patterns. Based on the 80–90 years of river runoff data series, the joint studies published by hydrologists from the Baltic countries [23,24,25] found a significant decrease in spring floods over the last century and the very beginning of the 21st century. Although two main types of floods—spring and flash floods—are historically distinguished in these countries, most river flood studies are based on annual maximum [26,27,28] or spring maximum flows [24,25,29]. Flash floods remain a largely understudied phenomenon, with limited research currently available [30].

To effectively assess changes in extreme hydrological variables, it is crucial to have access to long-term data series. These datasets enable researchers to detect trends, identify anomalies, and better understand the dynamics of water systems over time. In line with this, the World Meteorological Organization (WMO) recommends analyzing changes in abiotic parameters based on established climate normals to provide a reliable context for interpreting environmental changes. Comparing these periods may help identify changes in flood patterns caused by long-term climate change, rather than short-term climate fluctuations. This supports focusing on these two periods to analyze flood changes in the face of climate change. To date, no studies have examined spring and flash floods across the rivers of the Baltic States using a standardized methodology. To assess the impact of climate change, aiming to standardize and harmonize data across the Baltic States rivers, we used the historical base period of 1961–1990, along with the most recent 30-year period (1991–2020).

Therefore, to better understand shifting hydrological patterns and support effective climate adaptation, the current study analyzed changes in spring and flash floods in the Baltic States (Estonia, Latvia, and Lithuania) over the two climate normal periods: 1961–1990 and 1991–2020.

2. Materials and Methods

The Baltic States, consisting of Estonia, Latvia, and Lithuania, are situated in Northeastern Europe along the eastern coast of the Baltic Sea. The Baltic States cover a relatively small area, with a total territory of 175,117 km². Geographically, this region encompasses diverse landscapes such as coastal areas, forests, and plains. Hydrologically, the region is characterized by a complex network of fluvial systems, including major rivers like the Daugava and Nemunas, as well as numerous lakes and wetlands. The Baltic Sea plays a significant role in shaping the hydrological cycle and climatic conditions of the surrounding region, influencing local weather patterns. The climate is humid and relatively cold. The hydrological regime of Baltic rivers is influenced not only by climatic factors such as air temperature and precipitation but also by geomorphology, geology, soil composition, lakes, wetlands, and land use, all of which modify the overall nature of the water regime.

The authors [31] analyzed geographical and hydrometeorological characteristics such as the percentage of lake, forest, and wetland cover, the average density of river network, topography, lithology, long-term temperature, precipitation, snow cover duration, and river runoff data in the Baltic States. The variability and patterns of the listed characteristics formed the foundation for hydrological regions across the States. Ten distinct hydrological regions were identified—four in Latvia and three each in Estonia and Lithuania—primarily based on the dominant river-feeding sources: snowmelt, precipitation, and groundwater (Figure 1). The western regions of Lithuania (LT-W) and Latvia (LV-W), located near the Baltic Sea, fall within the marine climate zone, where precipitation is the primary source of river recharge. In contrast, eastern Latvia (LV-E), south-eastern Lithuania (LT-SE), and eastern Estonia (EE-E) represent the continental zones of the Baltic States, where river systems are predominantly fed by snowmelt and subsurface flow. These rivers exhibit a relatively uniform annual discharge pattern, typical of most Eastern European rivers, characterized by pronounced spring floods driven by snowmelt. The remaining hydrological regions display more localized and variable river discharge characteristics.

To investigate whether the dominant flood mechanisms had a seasonal and regional pattern, the analysis was performed separately for the cold and warm periods and ten hydrological regions. Therefore, the time series of maximum discharges (Q_MAX) was divided into two blocks of equal length: November–April and May–October, and then the highest values within each block were considered the peaks of spring floods and the peaks of flash floods, respectively. These dates were not selected arbitrarily. A review of the literature revealed that, in the studied region, spring floods typically occur between November and April, while summer flash floods are most common from May to October. This seasonal pattern is supported by regional hydrological and climatic studies in the Baltic region and other countries [24,25,26,27,32]. Using daily flow data from 1961 to 2020 collected at 68 water gauging stations (WGS) (Figure 2: 23 in Estonia, 22 in Latvia, and 23 in Lithuania) across 55 rivers of varying sizes, we extracted maximum flows for both the cold and warm periods (Appendix A,

Table A1. Main hydrological characteristics of the investigated rivers.

No.	River-WGS	Catchment Area, km2	Average of Spring Flood QMAX		Average of Flash Flood QMAX
			1961–1990	1991–2020	1961–1990	1991–2020
	Major rivers
1.	Narva-Vasknarva	47,807	482	494	509	497
2.	Daugava–Daugavpils	64,500	2318	1931	1329	1187
3.	Nemunas-Druskininkai	37,400	738	477	354	295
4.	Nemunas-Nemajūnai	42,900	854	558	421	351
5.	Nemunas-Smalininkai	81,200	1925	1360	890	707
6.	Neris-Vilnius	15,200	378	228	171	147
7.	Neris-Jonava	24,500	685	458	293	251
	Eastern hydrological region (EE-E)
8	Tagajõgi-Tudulinna	240	25.9	23.1	11.6	10.6
9	Kääpa-Kääpa	264	9.02	8.63	5.29	4.47
10	Pedja-Tõrve	739	42.3	36.9	23.4	19.8
11	Põltsamaa-Pajusi	1082	35.7	30.5	23.4	18.6
12	Emajõgi-Tartu	7849	123	127	99.9	102
13	Emajõgi-Rannu-Jõesuu	3366	33.2	38.7	38.6	41.6
14	Ahja-Ahja	897	29.0	27.1	14.2	13.3
15	Väike-Emajõgi-Tõlliste	1045	50.9	46.9	27.7	22.9
	Northern hydrological region (EE-N)
16.	Purtse-Lüganuse	785	52.3	39.0	28.3	21.7
17.	Kunda-Sämi	421	16.3	14.4	11.5	8.72
18.	Valgejõgi-Vanaküla	308	17.5	15.6	10.0	7.85
19.	Pudisoo-Pudisoo	124	7.38	6.00	4.45	3.26
20.	Leivajõgi-Pajupea	84.4	5.73	4.51	3.04	1.93
21.	Keila-Keila	636	33.5	31.5	17.3	14.7
22.	Vihterpalu-Vihterpalu	478	30.3	27.4	13.4	11.4
	Western hydrological region (EE-W)
23.	Kasari-Kasari	2642	204	181	89.1	63.3
24.	Prandi-Tori	284	15.4	13.3	7.90	6.10
25.	Navesti-Aesoo	1049	58.7	55.0	26.8	20.9
26.	Halliste-Riisa	1881	107	99.5	44.9	36
27.	Pärnu-Tahkuse	2047	136	131.4	64	49.8
28.	Pärnu-Oore	5131	303	272	135	107
29.	Lõve-Uue-Lõve	143	9.79	8.07	2.93	2.52
	Eastern hydrological region (LV-E)
30.	Rēzekne-Griškāni	518	20.7	16.1	6.69	9.59
31.	Dubna-Višķi	917	16.3	15.0	12.5	11.7
32.	Dubna-Sīļi	2070	70.4	57.2	31.3	33.5
33.	Aiviekste-Aiviekstes HES	8720	238	209	149	137
	Northern hydrological region (LV-N)
34.	Vaidava-Ape	411	21.3	20.8	13.4	10.7
35.	Tirza-Lejasciems	593	33.8	30.8	15.3	14.3
36.	Gauja-Velēna	710	42.9	43.8	16.5	16.7
37.	Gauja-Valmiera	6320	213	217	117	107
38.	Amata-Melturi	298	32.4	31.3	13.8	14.2
39.	Gauja-Sigulda	8670	282	302	160	147
40.	Salaca-Mazsalaca	2260	56.6	62.0	42.6	41.9
41.	Salaca-Lagaste	3190	131	139	69.5	60.9
42.	Lielā Jugla Zaķi	623	44.2	40.6	19.9	18.7
43.	Ogre-Lielpēči	1670	126	108	50.6	53.3
	Central hydrological region (LV-C)
44.	Mūsa-Bauska	5390	260	177	49.7	45.0
45.	Lielupe-Mežotne	9470	491	349	117	108
46.	Svēte-Ūziņi	611	25.7	19.2	6.15	6.16
47.	Bērze-Baloži	883	40.2	31.4	14.1	9.85
	Western hydrological region (LV-W)
48.	Venta-Kuldīga	8390	557	407	167	107
49.	Irbe-Vičaki	1860	74.8	67.2	30.6	24.4
50.	Bārta-Dūkupji	1700	179	150	60.4	34.1
	South-eastern hydrological region (LT-SE)
51.	Merkys-Puvočiai	4300	102	71.1	48.3	50.2
52.	Ūla-Zervynos	679	24.1	15.2	10.8	8.67
53.	Verknė-Verbyliškės	694	33.6	24.7	10.6	12.6
54.	Strėva-Semeliškės	234	4.12	3.26	3.07	2.47
55.	Žeimena-Pabradė	2600	45.5	38.8	35.5	31.3
56.	Šventoji-Anykščiai	3570	126	93.4	53.5	44.4
57.	Šventoji-Ukmergė	5380	200	145	79.2	67.9
	Central hydrological region (LT-C)
58.	Nemunėlis-Tabokinė	2740	196	147	53.6	48.5
59.	Mūša-Ustukiai	2280	157	88.6	27.1	19.7
60.	Šušvė-Šiaulėnai	162	15.9	8.16	4.08	2.58
61.	Dubysa-Lyduvėnai	1070	78.2	46.1	18.1	14.8
62.	Šešuvis-Skirgailai	1880	183	140	34.1	29.6
	Western hydrological region (LT-W)
63.	Venta-Papilė	1560	119	65.6	26.0	19.1
64.	Venta-Leckava	4020	258	183	76.1	52.4
65.	Bartuva-Skuodas	617	93.2	68.7	32.0	15.9
66.	Jūra-Tauragė	1660	251	228	76.2	72.4
67.	Akmena-Paakmenis	314	50.4	47.5	14.6	13.6
68.	Minija-Kartena	1220	150	132	54.2	44.2

). The data were obtained from the hydrological yearbooks of the Estonian Meteorological and Hydrological Institute, Latvian Environmental, Geology and Meteorology Centre, and Lithuanian Hydrometeorological Service. Before publication in the yearbooks, these data passed quality control, and their quality is the responsibility of the institutions mentioned above. The rivers were selected based on their data availability and quality, focusing on those unaffected or minimally affected by significant flow regulations.

In this study, the Pettitt test, the Mann–Kendall (MK) test, and Sen’s slope estimator were employed to analyze trends and detect potential change points in the time series of maximum discharges of spring and flash floods. The non-parametric nature of all these tools makes them particularly suitable for hydrological data, which often exhibit non-normality.

Change points in the 60-year flood data series were identified using the Pettitt test [33]. This rank-based, non-parametric test is well-suited for detecting abrupt changes or shifts in a dataset without assuming any specific distribution. The Pettitt test is considered sensitive to breaks near the middle of a time series [34]. It was employed to locate the most likely year in which a significant alteration in the flood regime may have occurred. Based on the Mann–Whitney statistic, the test compares all possible partitions of the time series to locate the point of maximum deviation. Since it relies on the ranks of the elements of a series rather than the values themselves, the ranking approach also implies that it is less sensitive to outliers than the other tests [35]. The test statistic is calculated as:

$$Z_k=2{\sum }_{i=1}^k{R}_i-k\left(n+1\right),k=1,\dots ,n$$

(1)

If the statistic Z_K = max Z_k is near the year k = K, then a change point occurs in year K. The critical values of the test are given in Pettitt [33]. This test was performed at the 5% (strong) and 10% (weak) significance levels.

The Mann–Kendall test [36,37] is another valuable and widely used non-parametric method for detecting trends in long-term data, particularly in hydrological and climatological studies. This statistical test is well-regarded for its robustness to non-normal data distributions and relative insensitivity to outliers. It detects the presence of a monotonic trend (increasing or decreasing) without requiring a linear assumption. This test was employed to evaluate temporal trends in the time series of maximum flood discharges.

The Mann–Kendall test statistic S is calculated according to [36,37]:

$$S={\sum }_{i=1}^{n=1}{\sum }_{j=i+1}^{n}sign(x_j-x_i);sign\left(x_j-x_i\right)=\left\{\begin{array}{c}+1\left(x_j-x_i\right)>0\\ 0\left(x_j-x_i\right)=0\\ -1\left(x_j-x_i\right)<0\end{array}\right.$$

(2)

A positive S value indicates an increasing trend, while a negative value shows a decreasing trend. The variance (S) of the time series is estimated to obtain the Z value [36,37]:

$$Var\left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-\sum _{i=1}^m{t}_i(t_i-1)(2t_i+5)}{18}}$$

(3)

The normal Z test statistic is computed by the equation [35,36]:

$$Z={\displaystyle \frac{S\pm 1}{Var{\left(S\right)}^{1/2}}}$$

(4)

This equation uses S − 1 if S > 0, S + 1 if S < 0, and Z is 0 if S = 0. A positive value of Z defines an upward trend, while a negative value indicates a downward trend.

Possible positive and negative trends were evaluated at two levels of significance: very significant at a 95% probability (α = 0.05) and weakly significant at a 30% probability (α = 0.30). Trends with weak significance (30%) were considered general trends, while only those with 95% probability were recognized as strong positive or negative significant trends.

To enable comparison among rivers from different catchments, each river’s total maximum discharge (Q_MAX) was converted into an equivalent water height (in millimeters) over the catchment area [38]:

$$h={\displaystyle \frac{Q_{max}·t}{1000·A}}$$

(5)

where t—time, s (per day 86.4 × 10³), A—catchment area, m².

A non-parametric linear Sen’s slope estimator [39] was used to quantify the magnitude of the detected trends. This method is widely utilized to identify the slope of a trend line in long-term hydrological time series. It calculates the median of all possible slopes between each pair of points, providing a robust and reliable estimate of the true slope. This method is advantageous when the data contains outliers or does not follow a normal distribution (which is common in hydrological datasets). Sen’s estimator assumes a linear trend but does not require the residuals to be normally distributed.

The following equation is used to estimate each individual slope (h_i) [39]:

$$h_j={\displaystyle \frac{x_j-x_k}{j-k}}$$

(6)

where i = 1, 2, 3, … N and j > k, where x₁, x₂, x₃ … x_j, x_k, … x_n are the data values; and the median of N values of h_j series can be obtained as Sen’s estimator of slope.

Visual summaries of flood peak data were provided as a ridgeline plot. This technique, also known as a ridge plot, displays the distribution of a numerical variable across multiple groups by stacking overlapping density plots, creating a ‘mountain range’ effect. It enables an easy comparison of the shape, spread, and central tendency across different groups or periods. The x-axis shows the investigated variable, while each ridge along the y-axis represents a group, with ridge width indicating the density of the data.

3. Results

3.1. Detection of Change Points in Flood Time Series

This study was dedicated to analyzing changes in river maximum discharges during the warm and cold periods of the year over a 60-year period. Since a comparison of the first thirty years with the second thirty-year period was intended, we first needed to test if a shift in the data occurred at the junction of these periods. To detect potential abrupt changes, the Pettitt test was applied. Since the calculated p-values were less than 0.05, the analyzed time series were considered nonhomogeneous. Figure 3 illustrates detected shifts in maximum discharge patterns for selected Lithuanian rivers over 1961–2020. In the case of spring flood discharges, the identified change points revealed a shift around 1990 or slightly later in most of the rivers. Unlike in the smaller rivers from different hydrological regions (where the years of change points had a certain variation), a shift was detected in the major Lithuanian rivers in 1988 or 1990, as for flash flood discharges, change-point years varied quite significantly. A group of rivers selected from the central hydrological region (LT-C) exibited a slightly different pattern, with the break points detected in 1998 or 2005, i.e., considerably later than in most other rivers. In major rivers, change points in flash flood events were more scattered than those in spring flood discharges. Therefore, patterns of change-point dates were more uniform in the time series of spring flood maximum discharges. In most cases, these shifts aligned closely with the boundary between the two selected periods. Therefore, the period from 1961 to 2020 was divided into two 30-year segments: the first corresponding to the climate normal period of 1961–1990, and the second to the most recent climate normal period of 1991–2020.

However, in the remaining states (Latvia and Estonia), change points in the maximum discharge data series were generally not detected. Significant shifts occurred only in limited number of cases. In Latvian rivers, these changes occured in 1972, 1988, 1989, 1994, or 1996, while in Estonian rivers in 1972, 1977, or 1991. The results of this analysis for the flood data in Estonia and Latvia are presented in Appendix B, Figure A1.

This study is based on data from governmental monitoring carried out by authorized institutions, ensuring a high level of data reliability. The data analysis was performed following a unified methodology. It should be taken into account that the studied rivers are subject to some degree of anthropogenic influence, which may be reflected in the results. Additionally, the rivers differ in terms of catchment size, types of feeding sources, and meteorological conditions (see Appendix A and Appendix D), all of which are addressed and clarified in the article.

3.2. Trends in Flood Maximum Discharges

Linear trends in the magnitudes of spring and flash flood maximum discharges (Q_MAX) were calculated and analyzed for the periods 1961–2020, 1961–1990, and 1991–2020 across ten hydrological regions of the Baltic States.

When comparing trend diagrams for the entire study period (1961–2020), a tendency toward a decline in flood peaks emerged (Figure 4, Appendix C,

Table A2. Statistical characteristics (p-value and Sens-slope) of flood h_MAX in the studied rivers.

No.	River-WGS	1961–2020		1961–1990		1991–2020
		p-Value	Sen’s Slope	p-Value	Sen’s Slope	p-Value	Sen’s Slope
Major rivers
1.	Narva-Vasknarva	0.254	0.002	0.021	0.016	0.844	−0.002
2.	Daugava–Daugavpils	0.012	−0.022	0.139	−0.032	0.544	−0.016
3.	Nemunas-Druskininkai	<0.0001	−0.015	0.077	−0.024	0.292	−0.007
4.	Nemunas-Nemajūnai	<0.0001	−0.017	0.108	−0.023	0.134	−0.010
5.	Nemunas-Smalininkai	0.000	−0.016	1.000	0.000	0.090	−0.015
6.	Neris-Vilnius	<0.0001	−0.024	0.019	−0.040	0.181	−0.010
7.	Neris-Jonava	0.000	−0.023	0.708	−0.012	0.072	−0.013
Eastern hydrological region (EE-E)
8.	Tagajõgi-Tudulinna	0.329	−0.030	0.556	−0.032	0.422	0.045
9.	Kääpa-Kääpa	0.452	−0.007	0.803	0.007	0.817	0.005
10.	Pedja-Tõrve	0.062	−0.025	0.069	−0.058	0.532	0.014
11.	Põltsamaa-Pajusi	0.182	−0.009	0.363	−0.025	0.261	0.015
12.	Emajõgi-Tartu	0.557	0.002	0.112	0.013	0.292	0.009
13.	Emajõgi-Rannu-Jõesuu	0.001	0.007	0.011	0.017	0.218	0.006
14.	Ahja- Ahja	0.151	−0.015	0.363	−0.031	0.392	−0.024
15.	Väike-Emajõgi-Tõlliste	0.137	−0.020	0.276	−0.039	0.775	−0.008
Northern hydrological region (EE-N)
16.	Purtse- Lüganuse	0.014	−0.039	0.668	0.019	0.301	−0.029
17.	Kunda-Sämi	0.475	−0.007	1.000	0.000	0.443	0.018
18.	Valgejõgi-Vanaküla	0.243	−0.016	0.454	0.027	0.630	−0.013
19.	Pudisoo- Pudisoo	0.071	−0.026	0.318	−0.053	0.972	−0.002
20.	Leivajõgi-Pajupea	0.092	−0.024	0.344	0.037	0.943	0.001
21.	Keila- Keila	0.433	−0.009	0.422	0.030	0.556	−0.014
22.	Vihterpalu-Vihterpalu	0.499	−0.010	0.708	0.013	0.957	−0.003
Western hydrological region (EE-W)
23.	Kasari-Kasari	0.172	−0.025	0.830	0.012	0.556	−0.024
24.	Prandi-Tori	0.153	−0.019	0.789	−0.012	0.943	0.003
25.	Navesti-Aesoo	0.429	−0.012	0.762	−0.015	0.803	0.015
26.	Halliste-Riisa	0.452	−0.009	0.498	0.020	0.721	−0.015
27.	Pärnu-Tahkuse	0.433	−0.013	0.592	0.024	0.261	−0.056
28.	Pärnu-Oore	0.256	−0.016	0.858	0.013	0.858	−0.011
29.	Lõve-Uue-Lõve	0.035	−0.035	0.695	−0.026	0.972	−0.001
Eastern hydrological region (LV-E)
30.	Rēzekne-Griškāni	0.021	−0.024	0.605	−0.011	0.253	−0.033
31.	Dubna-Višķi	0.200	−0.006	0.443	−0.006	0.830	−0.001
32.	Dubna-Sīļi	0.057	−0.020	0.318	0.029	0.064	−0.040
33.	Aiviekste-Aiviekstes HES	0.028	−0.012	0.344	−0.017	0.432	−0.013
Northern hydrological region (LV-N)
34.	Vaidava-Ape	0.755	−0.003	0.292	−0.041	0.077	0.068
35.	Tirza-Lejasciems	0.214	−0.023	0.392	−0.042	1.000	0.000
36.	Gauja-Velēna	0.223	−0.021	0.120	−0.061	0.301	−0.047
37.	Gauja-Valmiera	0.389	−0.007	0.762	−0.009	0.218	−0.020
38.	Amata-Melturi	0.355	−0.022	0.568	−0.037	0.929	−0.009
39.	Gauja-Sigulda	0.949	−0.001	0.986	−0.001	0.318	−0.019
40.	Salaca-Mazsalaca	0.389	0.005	0.486	0.013	0.708	−0.007
41.	Salaca-Lagaste	0.823	0.003	0.592	0.011	0.301	−0.027
42.	Lielā Jugla Zaķi	0.575	−0.007	0.580	0.010	0.139	0.043
43.	Ogre-Lielpēči	0.014	−0.039	0.521	−0.032	0.212	−0.040
Central hydrological region (LV-C)
44.	Mūsa-Bauska	0.006	−0.038	0.225	−0.066	0.943	−0.002
45.	Lielupe-Mežotne	0.024	−0.032	0.943	0.005	1.000	0.001
46.	Svēte-Ūziņi	0.036	−0.022	0.010	−0.113	0.972	0.004
47.	Bērze-Baloži	0.124	−0.017	0.301	0.037	0.432	−0.017
Western hydrological region (LV-W)
48.	Venta-Kuldīga	0.000	−0.045	0.986	0.001	0.046	−0.063
49.	Irbe-Vičaki	0.400	−0.006	0.193	0.036	0.762	−0.004
50.	Bārta-Dūkupji	0.159	−0.034	0.886	0.005	0.121	−0.065
South-eastern hydrological region (LT-SE)
51.	Merkys-Puvočiai	0.001	−0.015	0.708	−0.007	0.656	−0.004
52.	Ūla-Zervynos	0.000	−0.030	0.292	−0.031	0.354	−0.018
53.	Verknė-Verbyliškės	0.001	−0.041	0.972	−0.004	0.008	−0.082
54.	Strėva-Semeliškės	<0.0001	−0.010	0.498	−0.004	0.080	−0.011
55.	Žeimena-Pabradė	0.007	−0.008	0.592	−0.006	0.153	−0.008
56.	Šventoji-Anykščiai	0.003	−0.024	0.748	−0.007	0.080	−0.025
57.	Šventoji-Ukmergė	0.007	−0.025	0.844	−0.009	0.454	−0.014
Central hydrological region (LT-C)
58.	Nemunėlis-Tabokinė	0.017	−0.046	1.000	0.000	0.218	−0.047
59.	Mūša-Ustukiai	0.000	−0.064	0.532	−0.061	0.276	−0.028
60.	Šušvė-Šiaulėnai	<0.0001	−0.113	0.630	−0.036	0.001	−0.109
61.	Dubysa-Lyduvėnai	0.000	−0.047	0.721	0.014	0.169	−0.037
62.	Šešuvis-Skirgailai	0.063	−0.035	0.253	−0.050	0.605	−0.028
Western hydrological region (LT-W)
63.	Venta-Papilė	<0.0001	−0.073	0.253	−0.051	0.080	−0.048
64.	Venta-Leckava	0.000	−0.049	0.789	−0.012	0.020	−0.068
65.	Bartuva-Skuodas	<0.0001	−0.140	0.643	−0.046	0.000	−0.186
66.	Jūra-Tauragė	0.114	−0.049	0.901	−0.004	0.023	−0.208
67.	Akmena-Paakmenis	0.296	−0.038	0.175	0.113	0.112	−0.158
68.	Minija-Kartena	0.021	−0.055	0.301	0.065	0.009	−0.150

and

Table A3. Statistical characteristics (p-value and Sens-slope) of flash flood h_MAX in the studied rivers.

No.	River-WGS	1961–2020		1961–1990		1991–2020
		p-Value	Sen’s Slope	p-Value	Sen’s Slope	p-Value	Sen’s Slope
Major rivers
1.	Narva-Vasknarva	0.914	0.000	0.066	0.012	0.175	−0.007
2.	Daugava–Daugavpils	0.117	−0.009	0.363	−0.015	0.292	−0.016
3.	Nemunas-Druskininkai	0.009	−0.004	0.253	−0.006	0.735	−0.001
4.	Nemunas-Nemajūnai	0.005	−0.005	0.232	−0.005	0.402	−0.004
5.	Nemunas-Smalininkai	0.001	−0.006	0.372	−0.006	0.218	−0.006
6.	Neris-Vilnius	0.015	−0.005	0.269	−0.006	0.175	−0.005
7.	Neris-Jonava	0.041	−0.005	0.915	−0.001	0.193	−0.006
Eastern hydrological region (EE-E)
8.	Tagajõgi-Tudulinna	0.808	−0.003	0.681	0.019	0.887	−0.005
9.	Kääpa-Kääpa	0.262	−0.005	0.412	0.015	0.354	−0.012
10.	Pedja-Tõrve	0.326	−0.009	0.886	−0.006	0.872	−0.004
11.	Põltsamaa-Pajusi	0.067	−0.009	0.432	−0.010	0.544	−0.006
12.	Emajõgi-Tartu	0.858	0.000	0.212	0.008	0.454	−0.007
13.	Emajõgi-Rannu-Jõesuu	0.034	0.004	0.061	0.009	0.721	0.002
14.	Ahja- Ahja	0.838	0.001	0.080	0.023	0.225	−0.015
15.	Väike-Emajõgi-Tõlliste	0.301	−0.008	0.830	−0.007	0.181	−0.019
Northern hydrological region (EE-N)
16.	Purtse- Lüganuse	0.597	−0.006	0.318	0.000	0.292	−0.033
17.	Kunda-Sämi	0.293	−0.007	0.605	0.008	0.858	−0.001
18.	Valgejõgi-Vanaküla	0.093	−0.012	0.498	−0.022	0.301	−0.019
19.	Pudisoo- Pudisoo	0.214	−0.010	0.943	−0.003	0.617	−0.009
20.	Leivajõgi-Pajupea	0.035	−0.021	0.175	−0.050	0.335	−0.026
21.	Keila- Keila	0.345	−0.008	0.097	−0.038	0.915	0.003
22.	Vihterpalu-Vihterpalu	0.475	−0.007	0.159	−0.037	0.443	0.012
Western hydrological region (EE-W)
23.	Kasari-Kasari	0.358	−0.011	0.817	−0.015	0.498	0.014
24.	Prandi-Tori	0.118	−0.012	0.929	−0.005	0.335	−0.018
25.	Navesti-Aesoo	0.151	−0.010	0.872	−0.004	0.775	−0.007
26.	Halliste-Riisa	0.124	−0.012	0.803	−0.003	0.830	−0.003
27.	Pärnu-Tahkuse	0.118	−0.014	0.972	0.001	0.318	−0.025
28.	Pärnu-Oore	0.191	−0.010	0.682	−0.013	0.887	−0.001
29.	Lõve-Uue-Lõve	0.646	−0.003	0.721	0.009	0.199	−0.019
Eastern hydrological region (LV-E)
30.	Rēzekne-Griškāni	0.100	0.006	0.120	0.013	0.038	−0.038
31.	Dubna-Višķi	0.264	−0.004	0.630	−0.005	0.748	−0.003
32.	Dubna-Sīļi	0.853	−0.001	0.630	0.012	0.164	−0.021
33.	Aiviekste-Aiviekstes HES	0.069	−0.008	0.193	−0.018	0.125	−0.022
Northern hydrological region (LV-N)
34.	Vaidava-Ape	0.596	−0.005	0.344	0.023	0.695	0.010
35.	Tirza-Lejasciems	0.479	−0.006	0.655	−0.012	0.232	−0.027
36.	Gauja-Velēna	0.583	−0.004	0.556	−0.010	0.382	−0.023
37.	Gauja-Valmiera	0.386	−0.005	0.521	−0.010	0.292	−0.017
38.	Amata-Melturi	0.536	−0.014	1.000	0.000	0.101	−0.079
39.	Gauja-Sigulda	0.183	−0.006	0.682	−0.007	0.193	−0.014
40.	Salaca-Mazsalaca	0.764	−0.002	0.943	0.002	0.630	−0.006
41.	Salaca-Lagaste	0.418	−0.004	0.748	−0.009	0.592	−0.009
42.	Lielā Jugla Zaķi	0.134	−0.015	0.363	−0.028	0.087	−0.056
43.	Ogre-Lielpēči	0.475	−0.008	0.344	−0.023	0.205	−0.042
Central hydrological region (LV-C)
44.	Mūsa-Bauska	0.358	−0.003	0.617	0.005	0.475	−0.008
45.	Lielupe-Mežotne	0.111	−0.006	0.735	−0.003	0.372	−0.008
46.	Svēte-Ūziņi	0.429	−0.003	0.803	0.003	0.544	−0.007
47.	Bērze-Baloži	0.290	−0.004	0.120	0.023	0.656	−0.004
Western hydrological region (LV-W)
48.	Venta-Kuldīga	0.115	−0.009	0.363	0.021	0.153	−0.018
49.	Irbe-Vičaki	0.731	−0.001	0.139	0.019	0.972	0.001
50.	Bārta-Dūkupji	0.067	−0.016	0.486	0.032	0.556	−0.015
South-eastern hydrological region (LT-SE)
51.	Merkys-Puvočiai	0.314	−0.002	0.803	−0.001	0.309	−0.006
52.	Ūla-Zervynos	0.022	−0.008	0.682	−0.006	0.643	−0.004
53.	Verknė-Verbyliškės	0.207	−0.004	0.580	−0.006	0.083	−0.012
54.	Strėva-Semeliškės	<0.0001	−0.008	0.001	−0.016	0.121	−0.006
55.	Žeimena-Pabradė	0.021	−0.005	0.803	−0.002	0.134	−0.008
56.	Šventoji-Anykščiai	0.063	−0.007	0.844	0.004	0.159	−0.012
57.	Šventoji-Ukmergė	0.103	−0.006	0.972	−0.001	0.181	−0.012
Central hydrological region (LT-C)
58.	Nemunėlis-Tabokinė	0.046	−0.011	0.261	−0.018	0.193	−0.022
59.	Mūša-Ustukiai	0.256	−0.004	0.143	0.018	0.112	−0.013
60.	Šušvė-Šiaulėnai	0.026	−0.016	0.521	0.021	0.143	−0.028
61.	Dubysa-Lyduvėnai	0.134	−0.006	0.443	0.011	0.218	−0.014
62.	Šešuvis-Skirgailai	0.200	−0.007	0.901	0.002	0.199	−0.012
Western hydrological region (LT-W)
63.	Venta-Papilė	0.296	−0.005	0.199	0.022	0.052	−0.022
64.	Venta-Leckava	0.149	−0.007	0.205	0.018	0.164	−0.018
65.	Bartuva-Skuodas	0.142	−0.019	0.326	0.052	0.544	−0.015
66.	Jūra-Tauragė	0.583	−0.007	0.402	0.039	0.464	−0.021
67.	Akmena-Paakmenis	0.904	−0.001	0.104	0.083	0.125	−0.061
68.	Minija-Kartena	0.293	−0.014	0.138	0.076	0.148	−0.046

). The data of spring floods, especially in Lithuanian rivers, demonstrated particularly pronounced negative trends. Additionally, a tendency toward weaker negative trends—or the absence of any trend—could be observed moving northward. It should be noted that one river flood data showed an upward trend: a weak positive trend was identified in the Emajõgi River in the eastern region of Estonia. In general, considerably more insignificant trends were found in Estonian and northern Latvian rivers compared to those in Lithuania. This pattern may be explained by decreasing air temperatures (especially during the winter season) from south to north, i.e., moving from Lithuania to Estonia. Colder and more severe winters in the north (Estonia) allow the formation of both snow and ice sheets, the melting of which results in more consistent snowmelt-driven floods. Meanwhile, in Lithuania and a large part of Latvia (LV-W, LV-C, LV-E), winters tend to be milder, with the frequent above-freezing temperatures that inhibit snow cover accumulation. As a result, the potential for snowmelt-induced flooding is significantly reduced.

When the two climate normal periods were analyzed separately, the tendencies of either no significant trends or only weak positive or negative trends in spring floods were identified across all hydrological regions of the Baltic States in the first climatic normal period (1961–1990). Notable exceptions included the Svēte River (LV-C), which exhibited a strong positive trend, and the Emajõgi River (EE-E), which showed a strong negative trend. The changes in winter and spring seasonal air temperatures over 1961–1990 did not demonstrate significant trends; consequently, the trends in spring floods were also insignificant or weak.

During the second time slice (1991–2020), both strong and weak negative trends in flood peaks were observed in rivers across Lithuania and Latvia. In contrast, in Estonia and a large part of Latvia, no trends or only weak ones in spring floods were found. Overall, weak negative trends in spring flood peaks predominated across all three Baltic States during this period. Weak positive trends in maximum flows were identified in only two hydrological regions. Notably, the EE-E hydrological region stood out, where two rivers (Põltsamaa and Emajõgi) demonstrated weak positive trends in spring floods. This could be attributed to long-term fluctuations in the water regime, where a period of low water abundance that began in 2011 gradually transitioned into a more water-rich phase. Such hydrological cycles, often spanning several decades, can significantly influence flood frequency and magnitude.

A slightly different trend was observed in the case of flash floods (Figure 4, Appendix C, Table A2 and Table A3). Over the full study period 1961–2020, weak negative trends in flash floods predominated across all states. Only a few rivers (the Rezekne from LV-E; the Emajõgi from EE-E) showed upward trends in flood peaks. During the first climate normal period (1961–1990), flash flood patterns varied more widely. In many hydrological regions, rivers exibited no trend. However, each country had hydrological regions where some rivers displayed weak positive flash flood trends. For instance, in the LT-W region, four out of six studied rivers had an increasing flash flood trends. During the second climate normal period (1991–2020), the analysis of maximum discharges revealed either negative or no trends across all hydrological regions. The highest number of weak negative trends was observed in Lithuania, while the fewest was recorded in Estonia.

3.3. Changes in Maximum Flood Height Slopes

The Mann–Kendall test identified only tendencies toward increases or decreases in flood peak magnitudes. Maximum discharges of the rivers were found to be strongly dependent on the catchment size. To compare flood parameters between small and large rivers, the maximum flood height (h_MAX, mm) was calculated for all hydrological regions and all time slices considered in this study (Figure 5, Appendix C, Table A2 and Table A3). Changes in maximum flood height were estimated using a slope estimator [32]. In Figure 5, box plots show the maximum, minimum, 25th, and 75th percentiles as well as the means of selected time series of maximum flood height slopes for each hydrological region. Box plot magnitudes were calculated in mm as changes in h_MAX per year over the selected periods (1961–2020, 1961–1990, and 1991–2020).

For the period 1961–2020, the median of height slopes was negative in all hydrological regions (Figure 5, Appendix C, Table A2 and Table A3), with values ranging from −0.05 to −0.01 mm/year. The greatest decreases of h_MAX slopes, along with the highest variability, were observed in the Lithuanian hydrological regions (LT-W and LT-C) and one Latvian region (LV-W). In contrast, changes in h_MAX slopes were negligible in the remaining regions of Latvia and Estonia over the 60 years. The analysis of maximum spring flood discharges also confirmed that the most significant negative trends were characteristic of the hydrological regions of Lithuania.

In 1961–1990, median values of h_MAX slopes differed significantly from those estimated for 1961–2020. In many hydrological regions, the median of h_MAX slopes was approaching zero or even becoming positive. This indicates that during the first period of climate normal, the magnitude of spring floods was almost unchanged, or snowmelt floods had an increasing trend (in EE-W and EE-N). However, in the second period of climate normal (1991–2020), the trend of negative slopes medians was again evident. The highest variations and changes in snowmelt flood heights were found in the western hydrological region of Lithuania. The median height slope was −0.15 mm/year in 1991–2020.

The analysis of flash flood heights revealed smaller changes in the slopes of median values than those for spring flood heights (Figure 5, Appendix C, Table A2 and Table A3). There were insignificant changes in h_MAX slopes in 1961–2020, with the rate of decline ranging between 0.00 and 0.01 mm per year. From 1961 to 1990, no significant changes were observed in flash flood heights in most hydrological regions, except in western Lithuania (LT-W), where the median height slope was 0.5 mm per year. In 1991–2020, there was a very slight decrease in median height (up to 0.02 mm per year). Overall, changes in h_MAX slopes did not display clear patterns in the Baltic States during the three investigated periods.

3.4. Distribution Patterns of Maximum Flood Height

Visual summaries of flood data from different hydrological regions of the Baltic States over two climate normal periods (1961–1990 and 1991–2020) were presented using ridgeline plots (Figure 6).

To enable comparison, the total maximum discharge volume of each river was converted to an equivalent water height (in millimeters) over the catchment area. A general comparison of the density distribution patterns revealed that spring floods typically produced greater maximum heights than flash floods, both in terms of mean and peak values. When comparing the two periods, a decrease in maximum height values was observed across most of the studied time series. Specifically, for spring floods in all regions, the location of peaks in the density plots indicated a decline in both maximum values and data variance. Also, the data range along the x-axis narrowed. Ridgeline plots for flash flood data generally showed similar tendencies: lower mean and maximum values, along with narrower data ranges in the second period. However, exceptions were observed. In the eastern region of Latvia, the mean slightly increased. Additionally, in the south-eastern region of Lithuania and the central and eastern regions of Latvia, higher maximum values appeared in the second thirty-year period. In these regions, the data showed a slight tendency toward bimodality within the range of high values. Overall, the flash flood height plots exhibited sharper peaks and lower variability around the mean.

In Estonia, in the selected climate normal periods, the highest mean height values of spring floods were recorded in the western region. In contrast, the northern hydrological region exhibited the highest means of flash flood data. The lowest mean and maximum values for both flood types were observed in rivers located in the eastern region.

In Latvia, in both studied periods, the highest means of spring flood heights were observed in the western region, while according to the same criterion, the most intense flash floods were in the northern hydrological region. The lowest means and maximum values of the spring flood heights were identified in the eastern region, while in the case of flash floods, they were the smallest in the central and eastern regions.

In Lithuania, both flood types exhibited a west-to-east decline in mean values, maximum values (except for spring flood data from the central region in the first climate normal period), and the data range (with the same exception) over time. In the recent thirty-year period, steeper data curves and narrower ridges indicated reduced variability, suggesting more consistent flood magnitudes. The plots derived from spring and flash flood data in Lithuania’s western region were notably flatter than those of other regions. This flatness implies greater variability in flood intensity, suggesting that floods in these areas were more unpredictable.

Some density plots (e.g., spring flood plots for the Lithuanian western region in the second period, the Latvian western region in the first period, the Estonian northern region in the second period) exhibited varying degrees of bimodality. The presence of two peaks may indicate that the data within these groups behaved differently and potentially originated from two distinct subgroups.

3.5. Variability of Spring and Flash Floods in the Major Rivers

In this study, the variability of spring and flash floods in the major rivers of the Baltic States was analyzed separately. Trends in maximum flood events were calculated for the entire 60-year period as well as for each climate normal period (

Table 1. Trends in spring and flash floods (1961–2020, 1961–1990, and 1991–2020) in the major rivers of the Baltic States.

River	Spring Floods			Flash Floods
	1961–2020	1961–1990	1991–2020	1961–2020	1961–1990	1991–2020
Narva
Daugava
Nemunas
Neris

Strong negative trend, strong positive trend, weak negative trend, weak positive trend, no trend.

). For Lithuanian rivers (specifically the Nemunas and Neris), they revealed consistent patterns across all analyzed time slices and for both flood types: strong negative trends were identified for the full period (1960–2020), weak negative trends during the second climate normal period, and no significant trends during the first climate normal period. Some similarities to the Nemunas and Neris were also observed in the case of the Daugava River in Latvia. Its flood peaks showed downward trends (of strong significance for spring flood data and weak for flash floods) if calculated for the entire study period. However, when assessed separetely for the two climate normal periods, the spring flood data demonstrated the presence of weak negative trends in 1961–1990, and no trends in 1991–2020. In contrast, flash flood data displayed opposite tendencies. The variation in flood peaks of the Narva River from Estonia had a markedly different character. The trend analysis revealed an increase in flood peaks in the first climate normal period, with spring floods showing a statistically significant rise. Additionally, spring flood maximum discharges increased over the entire study period. Overall, when examining the climate normal periods individually, no trend or weak negative trends could be detected (except for the Narva River). The Narva River represents a specific case, as the corresponding water gauging station (WGS) is situated near the river’s source, where the flow is predominantly controlled by the hydrometeorological regime of Lake Peipsi, including ice conditions, water level, wind direction and wind setup. The lake’s water level has exhibited positive trends throughout the entire period, which explains the observed positive trends in flood peak patterns.

During the periods 1961–2020, 1961–1990, and 1991–2020, negative trends of h_MAX slopes of spring and flash floods were identified for the Daugava, Nemunas, and Neris rivers (Figure 7). The biggest negative median of height slopes was found for spring floods in the Daugava River. Variation pattern of this parameter in the Narva River was different from that of other major rivers. In 1961–2020, median values of h_MAX slopes remained close to zero for both spring and flash floods. Of all the major rivers, only the h_MAX slopes of the Narva River floods demonstrated an increase (0.01–0.02 mm per year) during the first period of climate normal (1961–1990). Overall, the analysis of flood peak trends and h_MAX slopes revealed consistent patterns in the major rivers of the Baltic States throughout the three study periods.

Ridgeline plots of flood maximum height data from major rivers reflected a general trend of lower flash flood peaks compared to spring floods and a decrease in maximum discharges over time (when two periods were compared) based on mean values, mirroring patterns observed in river groups across different hydrological regions (Figure 8).

The plots obtained from spring and flash flood data of the Daugava River were distinguished by their flatness, which may imply greater variability in flood peak magnitudes.

4. Discussion

Although numerous detailed national-scale studies have been devoted to studying river flood events in search of clear signs of climate change, no consistent pattern has been identified: in some regions, river floods are increasing; in others, they are decreasing; and in some, no clear trend is evident [9,15,40,41]. One possible reason for this inconsistency is the reliance on annual peak flows, which may have different flood-generating mechanisms if they occur in different seasons [42,43]. In such cases, trend analysis becomes challenging and may fail to reveal clear or meaningful tendencies. Accurately identifying patterns in maximum flood peaks enhances the ability to project their future behavior and mitigate potential impacts. Since floods in the Baltic States may occur at any time of the year due to different drivers, flood peaks were analyzed separately for the cold period (November–April) and the warm period (May–October), and were attributed to spring and flash floods, respectively. This approach allowed to group maximum flood discharges primarily driven by snowmelt during the cold period of a year and those mainly dependent on rainfall during the warm period.

The derived maximum discharge data covered a common 60-year period, encompassing two standard climate normal periods: 1961–1990 and 1991–2020. An initial goal of the study was to assess whether there was a shift in the magnitudes and patterns of flood peaks between the two thirty-year periods. Using a similar approach (the same two climate normal periods), significant changes were detected in the hydrometeorological parameters of the Curonian Lagoon in Lithuania [44], surface mass balance and summer temperature of Northern Hemisphere glaciers [45], the near-surface temperature in Scandinavia [46], the number of regional extreme heat events in summer and spring in the Apennine Mountains (in Italy) [47]. This approach is new in river flood change analysis. It was hypothesized that flood peak patterns would differ between the studied periods, and the results partially supported this expectation.

Analysis of flood peaks in the rivers of the Baltic States over the entire observation period (1961–2020) revealed more or less pronounced negative trends. In the case of spring floods, the trends were more significant than for flash floods. However, in the first 30-year period, magnitudes of both flood maximum discharges showed mostly no trends or general trends with prevailing increasing tendencies. On the contrary, the flood peaks observed during the second period exibited quite a considerable number of negative trends (with some exceptions in the case of spring floods). Ridgeline plots illustrated that spring floods typically reached greater maximum heights than flash floods; in most cases, maximum height values tended to be smaller in the second climate normal period for both floods.

Trends were compared across the states and different hydrological regions characterized by distinct runoff-generating conditions to identify regional flood patterns. The flood trend analysis indicated a tendency toward weaker negative trends—or the absence of trends—moving northward. Therefore, the estimated changes were examined in relation to climatic parameters (air temperature and precipitation) (Appendix D). The results suggested that snowmelt floods may have been strongly influenced by cold-season air temperatures. The difference in cold-season temperatures between the two climate normals was similar across all Baltic States (1.5–1.6 °C) (

Table A4. Air temperature changes in the Baltic States during two climate normal periods (according to https://climateknowledgeportal.worldbank.org/ (accessed on 3 July 2025)).

Month or Period	The Baltic States
	EE 1961–1990	EE 1991–2020	LV 1961–1990	LV 1991–2020	LT 1961–1990	LT 1991–2020
Jan	−6.4	−4.1	−5.8	−3.6	−5.5	−3.3
Feb.	−6.2	−4.4	−5.3	−3.4	−4.7	−2.8
Mar.	−2.4	−1.0	−1.4	0.2	−0.6	1.0
Apr.	3.8	5.2	4.9	6.4	5.9	7.3
May	10.2	10.7	11.2	11.8	12.1	12.6
Jun.	14.6	14.7	14.9	15.3	15.5	16.0
Jul.	16.5	17.7	16.5	18.0	16.8	18.3
Aug.	15.6	16.9	15.9	17.3	16.3	17.8
Sep.	11.1	12.2	11.6	12.6	12.1	13.0
Oct.	6.3	6.4	6.7	6.9	7.1	7.3
Nov.	1.0	1.8	1.4	2.1	1.8	2.5
Dec.	−3.6	−1.7	−3.2	−1.8	−2.8	−1.5
Annual	5.0	6.2	5.6	6.8	6.2	7.3
Warm period	12.4	13.1	12.8	13.6	13.3	14.2
Cold period	−2.3	−0.7	−1.6	0.0	−1.0	0.5

). However, the lowest temperatures were recorded in Estonia (average temperatures for 1961–1990 and 1991–2020 were −2.3 °C and −0.7 °C, respectively), while the highest values were observed in the southernmost state, Lithuania (−1.0 °C and 0.5 °C, respectively). The shift from negative to positive temperatures between the climate normals may have contributed to the significant negative trends in spring floods observed in Lithuania. In contrast, in Estonia, average cold-season temperatures remained negative in both periods, which may explain the highest number of statistically insignificant trends in spring floods. Previous research [48] has confirmed that in the Baltic States, air temperature is a key factor governing snow cover duration and determining the type of precipitation (rain or snow), which in turn forms the river runoff regime. This research also reported the decrease in the number of days with snow cover in the period 1961–2015. A decrease in spring flood peaks in the Baltic States was detected in the other studies [23,24,25,28] as well, although these were based on different time slices rather than standard climate normal periods.In the case of flash flood formation, the amount and distribution of rainfall are considered to play the most critical role [49]. A comparison of the average warm-season precipitation (

Table A5. Changes in precipitation in the Baltic State during two climate normal periods (according to https://climateknowledgeportal.worldbank.org/ (accessed on 3 July 2025)).

Month or Period	The Baltic States
	EE 1961–1990	EE 1991–2020	LV 1961–1990	LV 1991–2020	LT 1961–1990	LT 1991–2020
Jan.	38.4	49.5	38.9	49.4	40.6	49.2
Feb.	29.2	38.7	28.4	39.8	29.5	40.0
Mar.	31.0	34.7	34.4	37.5	36.7	38.8
Apr.	35.1	34.3	39.1	36.7	41.6	37.8
May	41.5	44.5	47.9	50.5	52.8	54.6
Jun.	50.7	66.9	59.6	68.4	68.5	67.2
Jul.	69.8	65.9	77.1	75.6	79.7	84.8
Aug.	78.3	78.2	76.9	75.8	76.1	75.3
Sep.	73.8	59.4	71.5	62.1	67.0	59.5
Oct.	64.9	73.6	60.9	71.5	56.4	64.5
Nov.	64.9	62.6	63.6	57.9	62.5	53.7
Dec.	52.2	55.5	52.5	54.1	52.9	52.4
Annual	630	664	651	679	664	678
Warm period	379	389	394	404	400	406
Cold period	251	275	257	275	264	272

) did not reveal significant differences between the two climate normals. For the periods 1961–1990 and 1991–2020, the values were as follows: in EE—379 mm and 389 mm; in LV—394 mm and 404 mm; in LT—400 mm and 406 mm, respectively.

Previous studies [50,51], covering the periods 1966–2015 and 1950–2019, respectively, reported increases in monthly and annual precipitation, as well as in the amount of precipitation during rainy periods in the studied region; however, most of these trends were statistically insignificant. It should also be noted that extreme rainfall events, which often lead to flash flooding, tend to occur irregularly and affect random areas, most often in urban regions. Therefore, it was impossible to identify consistent patterns in precipitation-related trends observed across hydrological regions in the Baltic States. The major rivers were analyzed separately because their catchments extend across multiple hydrological regions and their runoff comes from several tributaries. Although the magnitude of their flood peaks was expected to reflect broader change patterns beyond individual hydrological regions, the ridgeline plot analysis revealed tendencies similar to those found in smaller rivers, attributed to particular hydrological regions, i.e., smaller peaks for flash floods and either a decline or no statistically evident change in maximum peak values over time, with exception of the Narva River. The hydrological regime of this Estonian river is highly dependent on the hydrometeorological regime of Lake Peipsi [52].Consequently, the lake regime should be analyzed in conjunction with the river’s flow dynamics to achieve a more comprehensive and accurate assessment.

Similar negative trends in both spring and flash flood peaks of the Nemunas River have been identified by other researchers [30] as well. Additionally, studies in Eastern Europe [53] revealed that negative trends were significantly influenced by catchment size—the larger the catchment, the more pronounced the decline. Including large rivers in studies like the present one is important and valuable, as at the scale of large river catchments, the effects of land use changes (e.g., deforestation and urbanization) on floods are typically minimal [12]. Therefore, flood trends in these catchments may better reflect the impact of climate change rather than anthropogenic alterations.

The outcomes of the present study highlighted two key aspects. First, the patterns of flash and spring flood maximum discharges in the Baltic States differ. This suggests that our approach, which distinguishes between different types of flood events, is likely to yield more accurate results than analyses based solely on annual maximum discharges. Our study provides a more detailed understanding by accounting for the distinct driving forces behind each flood type. Even greater differences might have emerged if the classification had been based on a more in-depth analysis of flood-generation processes—such as verifying whether a spring flood peak was indeed driven by snowmelt—rather than on time-based grouping alone [54]. The second key aspect is that splitting the flood peak time series into 30-year periods did not reveal substantial trends, although declining trends over the 60-year period were considerably more pronounced. The complexity of the relationship between flood events and climate variability confirms the presence of other critical local factors shaping flood patterns [15,55]. The observed bimodality in flood peak data suggests that, even within the same hydrological regions, river flood regimes may be governed by different formative factors. The absence of consistent shift points in flood peak series at the boundaries of climate normal periods in the Baltic States implies that climate variables may not exert a direct linear influence. This also helps explain the limited effectiveness of trend analyses based on 30-year climate normals. As climatic and hydrological cycles do not always coincide, such short-term periods may fail to capture the full variability of hydrological extremes. Longer hydrological records are likely to improve the detection of significant trends and may provide greater insight into patterns of flood clustering, such as flood-rich or flood-poor periods [12,56].

River flooding in a changing world represents a complex and multifaceted phenomenon. The magnitude and associated risk of individual flood events are governed by specific formation mechanisms, which are influenced by a combination of natural processes and anthropogenic factors. While some drivers are relatively well understood, others remain insufficiently explored. Substantial knowledge gaps persist, posing challenges for accurately assessing and preparing for future extreme flood events. Therefore, continued research is essential to better understand the processes behind flood generation and enhance our ability to predict such events under changing environmental conditions. The lessons learned from the present study, recognizing its limitations and shortcomings, should help guide more effective and reliable future efforts on the subject.

5. Conclusions

Statistical analysis of spring and flash flood peaks in the Baltic States rivers over two climate normal periods shows the following changes. In the studied hydrological regions, both spring and flash flood peaks generally decreased throughout 1961–2020. These trends were strongest in Lithuanian rivers and weakened or disappeared moving north to Estonia and northern Latvia, reflecting regional climate differences that influence river hydrology.

The trends in maximum flood height (h_MAX) were mostly negative during 1961–2020, especially in Lithuanian and western Latvian rivers. Spring floods decreased the most, particularly in the second climate normal period (1991–2020), with h_MAX declining on average by 0.14 mm/year in Lithuania and 0.05 mm/year in Latvia. Flash floods showed smaller decreasing trends, especially in 1991–2020, with h_MAX dropping on average by 0.01 to 0.02 mm/year.

In major Baltic rivers such as the Nemunas, Neris, and Daugava, both spring and flash flood peaks generally decreased during 1961–2020. These declines were most pronounced in Lithuanian rivers during 1961–2020 and weaker in 1991–2020, with h_MAX dropping up to 0.02 mm and 0.01 mm per year, respectively. The Narva River in Estonia is an exception, showing a significant increase in spring flood peaks in the first period (up to 0.01 mm per year), with overall trends less clear due to Lake Peipsi’s unique hydrology.

Although this paper aimed to investigate river flood discharges in relation to changes in climatic factors (climate normals), peak flow variations were also influenced by a range of other, often more difficult to quantify, drivers—including land use change, the installation of hydraulic structures, and similar anthropological factors—since the studied rivers are not entirely natural systems.