Extinction behaviour for competing continuous-state population dynamics

This paper investigates a Lotka-Volterra type model of two competing populations driven by Brownian motions and spectrally positive $\alpha$-stable random measures, establishing nearly sharp conditions under which one of the populations becomes extinct.

The Gist

Imagine a bustling ecosystem with two competing species, let's call them Team Alpha and Team Beta. They live in the same territory, eat the same food, and constantly fight for resources. Sometimes, they help each other a little, but mostly, they are rivals.

This paper is a mathematical investigation into a very specific question: Under what conditions will one of these teams completely die out (go extinct), and under what conditions will they just get very small but never quite disappear (extinguish)?

The authors, Jie Xiong, Xu Yang, and Xiaowen Zhou, built a complex mathematical "simulator" to predict the fate of these populations. Here is a breakdown of their work using simple analogies.

1. The Game Board: A Noisy, Jumpy World

In real life, population growth isn't smooth. It's messy.

• The Smooth Part: Think of the daily growth of a population as a gentle, rolling hill. This is modeled by Brownian motion (like a drunk person walking in a straight line but stumbling slightly).
• The Jumpy Part: Sometimes, a disaster happens (a drought, a disease, a sudden boom). The population doesn't just slide down; it teleports to a new level. This is modeled by Poisson random measures (sudden, sharp jumps).
• The Competition: The two teams affect each other. If Team Alpha gets too big, it might crush Team Beta. This is the "two-way interaction."

The authors created a set of rules (equations) that describe how these two teams grow, shrink, jump, and fight in this noisy world.

2. The Two Fates: "Extinction" vs. "Extinguishing"

The paper makes a crucial distinction between two ways a population can fail:

• Extinction (The Hard Stop): The population hits zero exactly. It's like a bank account hitting $0.00 and closing the account forever. The species is gone.
• Extinguishing (The Fade Out): The population gets infinitely close to zero but never actually touches it. It's like a candle flame that gets smaller and smaller, flickering near the wick, but never fully goes out. It's technically still alive, but functionally, it's a ghost.

3. The Main Discovery: The "Power" of the Fight

The authors found that the fate of the teams depends heavily on the shape of the competition. They looked at "exponents" (powers) in their equations. Think of these powers as the intensity of the rivalry.

• The "Safe Zone" (Non-Extinction):
  If the competition is "mild" (mathematically, if the interaction power is high enough, specifically $\ge 1$), the teams are safe. Even if they fight hard, they will never hit zero. They might get small, but they will always bounce back. It's like two boxers who are tired but have enough stamina to keep standing.

• The "Danger Zone" (Extinction):
  If the competition is "aggressive" (the power is low, specifically $< 1$), things get dangerous.

  • Scenario A: If the rivalry is just right (a specific balance of how fast they grow vs. how hard they fight), there is a 50/50 chance (or some probability between 0 and 1) that one team dies out. It's like flipping a coin; sometimes they survive, sometimes they don't.
  • Scenario B: If the rivalry is too intense compared to their ability to grow, they are guaranteed to die out (probability = 1). The system is rigged against them.

4. The "Magic Ratio" Trick

How did they figure this out? They didn't just run simulations; they used a clever mathematical trick.

Imagine you are watching the two teams. Instead of watching them separately, you watch the ratio of Team Beta to Team Alpha.

• If Team Beta is huge compared to Alpha, the ratio is high.
• If Team Beta is tiny, the ratio is low.

The authors created a "test function" (a mathematical lens) that looks at this ratio. They asked: "If the ratio gets very high, does the system push it back down, or does it keep running away?"

• The "Escape Artist": In some cases, they found a way to prove that even if the population gets tiny, there is a mathematical "safety net" that prevents it from hitting zero.
• The "Sinking Ship": In other cases, they proved that once the population gets small, the forces of competition and noise push it inevitably toward zero.

5. Why Does This Matter?

You might ask, "Who cares about two imaginary populations?"

This math applies to real-world biology and finance:

• Biology: Will a rare species survive in a changing climate? Will a predator wipe out its prey? This paper gives scientists a checklist to predict if a species is doomed or if it has a fighting chance.
• Finance: In economics, "populations" can be companies or market sectors. If two companies compete, will one go bankrupt (hit zero), or will they just shrink to a niche size? This model helps predict market crashes or the survival of startups.

Summary

Think of this paper as a survival guide for competing species in a chaotic world.

• The Good News: If the competition isn't too brutal (the "powers" are high enough), life finds a way. The species will survive, even if they get small.
• The Bad News: If the competition is too fierce relative to their growth, they are likely to vanish completely.
• The Twist: Sometimes, the outcome is a coin toss. It depends on the exact starting numbers and the specific "flavor" of the noise in the environment.

The authors didn't just guess; they built a rigorous mathematical proof to show exactly where the line is drawn between survival and extinction.

Technical Summary

1. Problem Statement

The paper investigates the extinction and "extinguishing" behaviors of a two-dimensional stochastic population model representing two competing species. The system is modeled by a pair of Stochastic Differential Equations (SDEs) driven by Brownian motions and spectrally positive $\alpha$-stable random measures.

The specific system (Equation 1.1) describes the dynamics of two populations, $X_t$ and $Y_t$, with two-way interactions (mutual competition). The equations are:
$$\begin{aligned} X_t &= X_0 - \int_0^t [a_1 X_s^{p_1+1} + \eta_1 X_s^{\theta_1} Y_s^{\kappa_1}] ds + \int_0^t \sqrt{2a_2 X_s^{p_2+2}} dB_1(s) + \int_0^t \int_0^\infty \int_0^{a_3 X_s^{p_3+\alpha_1}} z \tilde{N}_1(ds, dz, du), \\ Y_t &= Y_0 - \int_0^t [b_1 Y_s^{q_1+1} + \eta_2 Y_s^{\theta_2} X_s^{\kappa_2}] ds + \int_0^t \sqrt{2b_2 Y_s^{q_2+2}} dB_2(s) + \int_0^t \int_0^\infty \int_0^{b_3 Y_s^{q_3+\alpha_2}} z \tilde{N}_2(ds, dz, du). \end{aligned}$$
Key Definitions:

• Extinction: The process hits 0 in finite time ($\tau_0^- < \infty$).
• Extinguishing: The process converges to 0 as $t \to \infty$ but never hits 0.
• Context: Previous work (e.g., [22]) focused on one-sided interactions. This paper addresses the more complex two-sided (mutual) interaction case, which is significantly more challenging due to the coupling of the drift and jump terms.

2. Methodology

The authors employ a rigorous probabilistic approach based on Chen's criteria for boundary classification of Markov processes, adapted for two-dimensional systems.

• Martingale Approach: The core technique involves identifying appropriate test functions ($g(x,y)$) and analyzing the action of the infinitesimal generator ($L$) of the process on these functions.
  • To prove non-extinction (probability of hitting 0 is 0), they use test functions that diverge to infinity near the boundary (e.g., logarithmic or negative power functions) and show $Lg \leq Cg$.
  • To prove extinction (probability of hitting 0 is 1), they use test functions bounded near the boundary (often involving exponential and power functions) and show $Lg \geq Cg$.
• Ratio Analysis: A novel contribution is the analysis of the ratio $U_t = Y_t X_t^{-\beta}$ (or similar power ratios). By studying the behavior of this ratio, the authors derive conditions under which the extinction probability is strictly between 0 and 1.
• Comparison Theorem: A pathwise comparison theorem (Proposition 3.2) is established to handle the coupling between the two processes, allowing the authors to bound the system's behavior.
• Irreducibility: The paper proves that the process can reach any open subset of the positive quadrant with positive probability (Proposition 3.3), which is crucial for extending local results (small initial values) to global results (any initial values).

3. Key Contributions and Results

A. Dichotomy of Extinction vs. Non-Extinction (Theorem 1.2)

The paper establishes a sharp condition for whether the populations can ever reach zero:

• Result: $P(\tau_0^- < \infty) = 0$ if and only if $\theta_1 \geq 1$ and $\theta_2 \geq 1$.
• Implication: If the interaction exponent $\theta_i \geq 1$, the competition term is not strong enough to drive the population to zero (even if the intrinsic growth is negative). If $\theta_i < 1$, there is a positive probability of extinction.
• Significance: This extends the known result for one-dimensional nonlinear CSBPs to the two-dimensional interacting case, showing that the $Y^{\kappa_1}$ term in the interaction does not alter the extinction threshold for $X$ when $\theta_1 \geq 1$.

B. Conditions for Extinction with Probability 1 (Theorem 1.4)

Under the assumption that $\theta_1 \geq 1$ and $0 \leq \theta_2 < 1$(meaning$Y$is susceptible to extinction but$X$is not), the paper provides sharp conditions for$Y$to become extinct with probability 1 ($P(\tau_0^-(Y) < \infty) = 1$).

• Conditions: Extinction occurs if the interaction powers and coefficients satisfy specific inequalities involving the minimum powers of the intrinsic dynamics ($p, q$) and the interaction exponents ($\kappa_2, \theta_2$).
  • Specifically, if the "competition pressure" (determined by $\theta_1, \kappa_2, q$) is sufficiently high relative to the "recovery potential" (determined by $p$), extinction is certain.
• Critical Cases: The paper distinguishes between non-critical cases (determined by powers) and critical cases (where coefficients $a_i, b_i, \eta_i$ matter).

C. Conditions for Extinction with Probability in $(0, 1)$ (Theorem 1.3)

This is a major novelty. The paper identifies conditions where the extinction probability is strictly between 0 and 1.

• Mechanism: This occurs when the interaction is strong enough to cause extinction with some probability, but not strong enough to guarantee it.
• Method: The authors construct a specific test function involving the ratio $Y X^{-\beta}$ and use a new criterion (Proposition 2.5) to show that for small initial values, the process can "escape" to infinity (or rather, avoid 0) with positive probability.
• Conditions: The conditions involve inequalities like $\theta_1 < 1 + \frac{\kappa_2(q-\kappa_1)}{q+1-\theta_2}$ or $p < \frac{\kappa_2 q}{q+1-\theta_2}$.

D. Technical Lemmas and Tools

• Lemma 3.4 & 3.9: Provide precise estimates for the generator applied to exponential-type test functions, handling the singularities near zero and the jump integrals.
• Proposition 2.5: A new criterion for showing $P(\tau_0^- < \infty) < 1$ by analyzing the ratio of processes, which is a significant methodological advancement over previous one-dimensional techniques.

4. Significance

1. Theoretical Advancement: This is one of the first systematic studies of boundary behaviors (extinction/extinguishing) for two-dimensional interacting stochastic population models with jumps. It bridges the gap between one-dimensional CSBPs and complex multi-species systems.
2. Sharp Conditions: The paper provides "nearly sharp" conditions, distinguishing clearly between cases where extinction is impossible, certain, or probabilistic. This is crucial for understanding the long-term survival of competing species in stochastic environments.
3. Methodological Innovation: The adaptation of Chen's criteria to two dimensions, particularly the use of ratio-based test functions and the rigorous handling of the irreducibility of the 2D process, offers a template for analyzing other multi-dimensional SDE systems.
4. Ecological Relevance: The results clarify how the interplay between intrinsic growth rates, competition intensity, and environmental noise (both Gaussian and jump) determines the survival of competing populations. It highlights that even if one species is robust ($\theta \geq 1$), the other can still face a non-trivial risk of extinction depending on the specific power laws of interaction.

5. Limitations and Open Problems

• Critical Cases: The paper explicitly leaves the "critical cases" (where the inequalities in Theorems 1.3 and 1.4 become equalities) as an open problem.
• Positive Interactions: The current analysis focuses on negative interactions (competition). The authors note that extending these results to positive interactions (mutualism) would be more challenging due to the potential for explosion.

In summary, the paper provides a comprehensive mathematical framework for analyzing the survival of competing populations in a stochastic environment, offering precise thresholds for extinction and introducing powerful new tools for multi-dimensional stochastic analysis.